www.techrepublic.com/article/inside-the-raspberry-pi-the-story-of-the-35-computer-that-changed-the-world/

Not a bad article about the founding of the Raspberry Pi phenomenon.

a $3 wifi module and micro-controller. it is a cheap and reliable solution for small automation projects.

>>16
>Intel ME™ inside™
>>1405
>proprietary jewberry
meh, use Olimex

>>1407
We're all well-aware of the botnet around here, and of striving to protect against it anon. Have anything better to contribute than simple grousing? At least post pics of good SBCs you approve of ITT if nothing else.

>>1406
I think I remember seeing that from the site before anon. Thanks for the pdfs especially.

>>1410
As for microcontrollers, there's the HiFive1, for example.

SoCs, there currently exists the HiFive Unleashed (which costs a lot) and there's going to be lowRISC

>>1429
thanks anon.

>>1430
www.lowrisc.org/
www.sifive.com/boards/hifive1

neat little DIY raspi 'laptop' project https://invidious.snopyta.org/watch?v=tDVWA3wdStY

Beaglebone has produced a new product they are touting as being for AI. Dual Arm Cortex-A15 microprocessors w/ dual C66x floating-point VLIW DSPs. I assume that's only 2 DSPs, but I'm not sure yet. TI has been in this game all the way from the very beginning of silicon-based computing, and they are arguably the world leader in DSP systems, so yea. It's probably a very powerful little SBC, maybe the best atm. Not sure yet about general support for common libs like TensorFlow or PyTorch yet, so YMMV. https://beagleboard.org/ai www.ti.com/product/AM5729

>>3722 >update: It looks like the board is capable of doing Computer Vision directly in hardware. >"embedded-vision-engine (EVE) cores supported through an optimized TIDL machine learning OpenCL API with pre-installed tools." http://downloads.ti.com/mctools/esd/docs/tidl-api/example.html If so, then to my knowledge this is the very first SBC to do so, at least for the inexpensive consumer device market. Probably means this would be a very nice dedicated board for running the robowaifu's eyes & vision.

>>3722 I believe this is the proper user guide for the DSP subsection of the chip.

>>3724 This one is probably a closer fit for our needs tbh.

TI's Deep Learning SDK (presumably) used on the board. http://software-dl.ti.com/processor-sdk-linux/esd/docs/05_00_00_15/linux/Foundational_Components_TIDL.html kek, this is turning into a cluster. I'm going to redo these posts into some kind of order. :^)

>>1431 Through a train of links via this post, I found this open-source hardware FPGA microcontroller board. They already have a working prototype stage ATP. This is the smallest-scale one of a planned family of these for their platform. https://github.com/pulp-platform/pulpissimo https://pulp-platform.org/ Also, there seems to be a growing movement for opensauce h/w as exemplified by lowRISC in Cambridge. https://www.lowrisc.org/blog/2019/10/ibex-on-fpga-get-stuff-executed/ >>1429

>>3727 >kek, this is turning into a cluster. I'm going to redo these posts into some kind of order. :^) ehh, I'm just going to dump everything here now and then see about deleting/cleaning it up into a couple of orderly posts later. more important to collect the into one place than to make it pretty at the same time. >demo of programming the AI https://beagleboard.org/p/175809/tidl-on-beaglebone-ai-1ee263 https://invidious.snopyta.org/watch?v=v3_hNBJbqME https://invidious.snopyta.org/watch?v=1XTIi9Z_8x4 >programming the beaglebone PRU https://beagleboard.org/pru https://invidious.snopyta.org/watch?v=PZSxRHO59AI https://invidious.snopyta.org/watch?v=ZNz45v9Uesg https://invidious.snopyta.org/watch?v=y8-gwthzf-c

>>3729 >the beaglebone PRU https://markayoder.github.io/PRUCookbook/

https://ftp.openbsd.org/pub/OpenBSD/6.7/armv7/ https://ftp.openbsd.org/pub/OpenBSD/6.7/armv7/INSTALL.armv7

Beaglebone AI repo https://github.com/beagleboard/beaglebone-ai

>>3733 Beaglebone AI reference manual https://github.com/beagleboard/beaglebone-ai/wiki/System-Reference-Manual I'll see about turning this into a single pdf file later.

Projects that utilize the PRUs (2x available on the Beaglebone) Project Description Type Link -LEDscape BeagleBone Black cape and firmware for driving a large number of WS281x LED strips Code Library https://github.com/osresearch/LEDscape Documentation and example projects http://trmm.net/LEDscape -LDGraphy Laser direct lithography for printing PCBs Code library and example project https://github.com/hzeller/ldgraphy/blob/master/README.md -PRdUino This is a port of the Energia platform based on the Arduino framework allowing you to use Arduino software libraries on PRU. Code Library https://github.com/lucas-ti/PRdUino -DMX Lighting Controlling professional lighting systems Project http://beagleboard.org/CapeContest/entries/BeagleBone+DMX+Cape/ Tutorial http://blog.boxysean.com/2012/08/12/first-steps-with-the-beaglebone-pru/ Code Library https://github.com/boxysean/beaglebone-DMX -Interacto A cape making BeagleBone interactive with a triple-axis accelerometer, gyroscope and magnetometer plus a 640 x 480/30 fps camera. All sensors are digital and communicate via I²C to the BeagleBone. The camera frames are captured using the PRU-ICSS. The sensors on this cape give hobbyists and students a starting point to easily build robots and flying drones. Project 1 http://beagleboard.org/CapeContest/entries/Interacto/ Project 2 http://www.hitchhikeree.org/beaglebone_capes/interacto/ code library https://github.com/cclark2/interacto_bbone_cape -Replicape: 3D Printer Replicape is a high end 3D-printer electronics package in the form of a Cape that can be placed on a BeagleBone Black. It has five high power stepper motors with cool running MosFets and it has been designed to fit in small spaces without active cooling. For a Replicape Daemon that processes G-code, see the Redeem Project Project http://www.thing-printer.com/product/replicape/ Code Library https://bitbucket.org/intelligentagent/replicape/ -PyPRUSS: Python library PyPRUSS is a Python library for programming the PRUs on BeagleBone (Black) Code Library http://hipstercircuits.com/pypruss-a-simple-pru-python-binding-for-beaglebone/ -Geiger The Geiger Cape, created by Matt Ranostay, is a design that measures radiation counts from background and test sources by utilising multiple Geiger tubes. The cape can be used to detect low-level radiation, which is needed in certain industries such as security and medical. Project 1 http://beagleboard.org/CapeContest/entries/Geiger+Cape/ Project 2 http://elinux.org/BeagleBone/GeigerCapePrototype Code library https://github.com/mranostay/beaglebone-telemetry-presentation -Servo Controller Foosball Table Used for ball tracking and motor control Project http://www.element14.com/community/community/knode/single-board_computers/next-gen_beaglebone/blog/2013/07/17/hackerspace-challenge--leeds-only-pru-can-make-the-leds-bright Tutorial https://docs.google.com/spreadsheet/pub?key=0AmI_ryMKXUGJdDQ3LXB4X3VBWlpxQTFWbGh6RGJHUEE&output=html Code library https://github.com/pbrook/pypruss -Imaging with connected camera Low resolution imaging ideal for machine vision use-cases, robotics and movement detection Project & Code library http://www.element14.com/community/community/knode/single-board_computers/next-gen_beaglebone/blog/2013/08/18/bbb--imaging-with-a-pru-connected-camera -Computer Numerical Control (CNC) Translator Smooth stepper motor control; real embedded version of LinuxCNC Tutorial http://www.buildlog.net/blog/2013/09/cnc-translator-for-beaglebone/ Tutorial http://bb-lcnc.blogspot.com/p/machinekit_16.html -Robotic Control Chubby Project http://www.youtube.com/watch?v=dEes9k7-DYY Code library https://github.com/cagdasc/Chubby1_v1 SpiderBot Project http://www.youtube.com/watch?v=JXyewd98e9Q Reference http://www.ti.com/lit/wp/spry235/spry235.pdf -Software UART Soft-UART implementation on the PRU of AM335x Code library & Reference http://processors.wiki.ti.com/index.php/Soft-UART_Implementation_on_AM335X_PRU_-_Software_Users_Guide -Deviant LCD PRU bit-banged LCD interface @ 240x320 Project http://www.beagleboard.org/CapeContest/entries/DeviantLCD/ Code library https://github.com/cclark2/deviantlcd_bbone_cape -Nixie tube interface code library https://github.com/mranostay/beagle-nixie -Thermal imaging camera Thermal camera using Beaglebone Black, a small LCD, and a thermal array sensor project & Code library https://element14.com/community/community/knode/single-board_computers/next-gen_beaglebone/blog/2013/06/07/bbb--building-a-thermal-imaging-camera -Sine wave generator using PWMs Simulation of a pulse width modulation Project & Reference http://elinux.org/ECE497_BeagleBone_PRU Code library https://github.com/millerap/AM335x_PRU_BeagleBone -Emulated memory interface ABX loads amovie into the Beaglebone's memory and then launches the memory emulator on the PRU sub-processor of the Beaglebone's ARM AM335x Project https://github.com/lybrown/abx -6502 memory interface System permitting communication between Linux and 6502 processor Project http://elinux.org/images/a/ac/What's_Old_Is_New-_A_6502-based_Remote_Processor.pdf Code library https://github.com/lybrown/abx -JTAG/Debug Investigating the fastest way to program using JTAG and provide for debugging facilities built into the Beaglebone Project http://beagleboard.org/project/PRUJTAG/ -High Speed Data Acquistion Reading data at high speeds Reference http://www.element14.com/community/community/knode/single-board_computers/next-gen_beaglebone/blog/2013/08/04/bbb--high-speed-data-acquisition-and-web-based-ui -Prufh (PRU Forth) Forth Programming Language and Compiler. It consists of a compiler, the forth system itself, and an optional program for loading and communicating with the forth code proper. Compiler https://github.com/biocode3D/prufh -VisualPRU VisualPRU is a minimal browser-based editor and debugger for the Beaglebone PRUs. The app runs from a local server on the Beaglebone. Editor and Debugger https://github.com/mmcdan/visualpru -libpruio Library for easy configuration and data handling at high speeds. This library can configure and control the devices from single source (no need for further overlays or the device tree compiler) Documentation http://users.freebasic-portal.de/tjf/Projekte/libpruio/doc/html/index.html Library (German) http://www.freebasic-portal.de/downloads/fb-on-arm/libpruio-325.html -BeagleLogic 100MHz 14channel logic analyzer using both PRUs (one to capture and one to transfer the data) Project http://beaglelogic.net -BeaglePilot Uses PRUs as part of code for a BeagleBone based autopilot Code Library https://github.com/BeaglePilot/beaglepilot -PRU Speak Implements BotSpeak, a platform independent interpreter for tools like Labview, on the PRUs Code Library https://github.com/deepakkarki/pruspeak >sauce processors.wiki.ti.com/index.php/PRU_Projects

>>3736 >>3723 I just realized there are 4x PRU and 4x EVE on each board. BeagleBone Black mechanical and header compatibility TI AM5729 processor featuring 2x A15 CPU, 2x C66 DSP, 4x M4 MCU, 2x SGX544 3D, GC320 2D, IVA-HD, 4x PRU and 4x EVE 1GB RAM and 16GB on-board eMMC flash with high-speed interface USB type-C for power and superspeed dual-role controller; and USB type-A host Gigabit Ethernet, 2.4/5GHz WiFi, and Bluetooth microHDMI Zero-download out-of-box software experienceBeagleBone Black mechanical and header compatibility >sauce https://beagleboard.org/p/products/beaglebone-ai >>3722 >I assume that's only 2 DSPs, but I'm not sure yet. Correct.

This looks kind of cool tbh. I sort of want one of these. Not really a robowaifu thing, but it looks kind of fun. https://www.okdo.com/us/p/piper-computer-kit-2/

crossposted from the embedded learning thread. a grab-bag of technical documents for arduino microcontrollers. https://onlinedocs.microchip.com/

For anyone that can afford it this might be interesting. https://www.anandtech.com/show/15856/qualcomm-announces-rb5-robotics-platform-a-powerful-sbc

>>3851 That does look interesting Anon, thanks. A bit pricey atm for our purposes IMO, but still worth a look-see.

I suppose this is the most relevant thread atm on /robowaifu/ to post this. There has been a slow push for open sauce RISC chip designs & manufacturing. This movement has recently gotten a potentially big boost from Google what could possibly go wrong, Anon? with their Skywater PDK (process design kit). https://github.com/google/skywater-pdk https://en.wikipedia.org/wiki/Process_design_kit > via nano /g/4908

>>4128 >yfw the main tech company Google is fronting on this project starts the actual skynet

>>4128 related. https://invidio.us/watch?v=EczW2IWdnOM fair warning, this group he represents are hyper CoC-suckers

I asked about some infos about FPGAs on lain/diy. Even if many there might not like /robowaifu/, we could still pick their brains. I'm using all kinds of sites to ask stuff, and might link to it then. Must be careful not to dox myself, though wouldn't matter that much anyways. "Hi! I would like to know if it might make sense to preprocess images with an (Icestorm) fpga for interference in image recognition. Could I take images from a video stream when something changes beyond a certain treashold? Could I get a image e.g. every 10 frames in b/w and some others with a lower resolution? Would it make sense to do this with an FPGA, when the images then go to a SBC like Jetson Nano with a GPU anyways? Is it much more efficient? I would also like to compare images, like finding a difference, by checking for color changes in any area. The other thing is voice alternation, like from espeak or festival, then changing it. Could I develop these changes in a software like Audacity or train a neural network, and then import these changes into a FPGA toolchain?" https://lainchan.org/diy/res/2899.html#4795

>>4474 Sure, sounds like a good idea. There is plenty of good resources on lainchan, and a few smart ppl there. I would prefer us to be allies, for my part. Yes, don't dox yourself Anon.

>>4474 It depends on how fast the SBC's CPU and how much RAM you have. For initial testing you could probably just do the image processing on the SBC, where you can quickly test different algos. To prepare your design for FPGA will take longer, so you better have a working design first. The main advantage is you could offload a lot of the processing to the FPGA, leaving your SBC to do voice or something else. I've been learning about formal verification for FPGAs for the last few months here: https://zipcpu.com/tutorial/ I strongly recommend it if you have the time, before starting my testing was very basic XD

>>4487 >https://zipcpu.com/tutorial/ Not that anon, Anon, but thanks I was trying to remember that project name earlier today.

>>4488 You're welcome m8 ;)

>>4487 >The main advantage is you could offload a lot of the processing to the FPGA, leaving your SBC to do voice or something else. Hi, I didn't plan to offload interference to the FPGA since some of then newer SBC even have specialized Asics for that. I was thinking about using the FPGA for preparations, like only picking some pictures from a live feed, send them only to the SBC if something changes beyond a threshold. Or generaly only sending some pictures from the feed dependent on the situation and maybe doing some processing on them before doing so. I probably shouldn't get into FPGA now, have already enough other stuff on my table. I'm njust keeping a eye on it. Thanks for the link nevertheless.

>>4490 >have already enough other stuff on my table Very relatable XD

Here's a separate thread about alternative CPU designs >>4506, like Risc-V.

>>4511 It's a very good thread, thanks Anon.

>>4487 >FPGA I'm definitely interested in learning more about this. Will there ever be FPGAs inexpensive enough for the average anon to use them for his robowaifus I wonder? also, very-related X-link >>4969 >>4895

>>5150 Yes, I think so. Depends on how much money the "average anon" here has, but also the use case. Cheap and weak ones aren't so expensive.

offtopic guys but i recently bought a small fake phone with rather decent specs for just 1000 peso orange pi seems promising but they say its buggy i succesfully rooted the android 4.0 OS and took the phone apart lcd screen and camera is pretty decent depite the price how do i connect it to my arduino (will my device count as SBC)

>>5315 >how do i connect it to my arduino if it's got a usb connector, then you'd use a serial-to-usb connector. i think this type of connection was brought in our embedded group learning thread. >>367

There's now a cheaper version of Jetson Nano, with fewer RAM: https://youtu.be/UNCsNH3tdcg for around 60$, while the original is 99$ and the bigger brother 500$ or so.

>>5699 That's good news, thanks Anon.

Also, OpenCV is coming out with a kit, similar to Google TPU: https://opencv-ai-kit.backerkit.com/hosted_preorders We missed the campaign, when it was probably cheaper. Starts at 150$ now, same price as Google's Coral board. It uses OpenVino: https://youtu.be/kY9nZbX1DWM and aXelerate (own device) or Luxonis (Web API) https://docs.luxonis.com/tutorials/openvino_model_zoo_pretrained_model/ Test: https://youtu.be/dFh0rrKEVJc IMHO, even the smaller board is to big for a camera board, if one wants to build a humanoid with two eyes. It might fit into an eyeball, but probably won't. I'll hope there will be a version with two layers, half the size. Since it's open source, my hopes might not be futile. Anyone here compared Jetson Nano vs Beaglebone AI or other?

>>5705 Wow, also really good news. OpenCV is a great platform. Easy to use (relatively) and very powerful as well. And unlike many other libraries, is much more open and available to everyone, at the least it's not nearly so leveraged against commercial platforms as many others are.

>>5705 >if one wants to build a humanoid with two eyes. It might fit into an eyeball, but probably won't I have two JeVois cameras which are ~2cm flattened 'cubes'. They are light & self-cooled and run Linux+OpenCV right onboard.

>>5707 Oh wow, cool. I recalled to have watched a video about them quite a while ago. However, just realized my posting of OAK was rather OT, since we have a thread for cameras, vision, and control boards. I posted some extra infos there and crosslink it: >>5714

no posts for one month really BUMP im quite surprised nobody mentioned mini PC's like intel NUC speaking of mini pc is there a way to install linux on my H96 box (its just 1500 lol) modern windows boxes are quite expensive (about 7000 peso) also i accidentally broke my fake phone while unsoldering the csi camera ribbon cable (its not that severe) where can i buy one of these fake iphones (or android) (RPI camera and screen is too expensive) video related https://www.youtube.com/watch?v=GwhzГунт4Xg (i kinda like this channel)

>>7344 welcome, back happy thanksgiving. you're just in time to celebrate our 4th birthday here anon. >linux on h96 I don't see why not https://forum.armbian.com/topic/8210-linux-on-h96-max-rk3399-variants/ https://blog.nobugware.com/post/2019/h96-max-android-box-as-a-linux-server/ https://www.youtube.com/watch?v=nSCj1b6PbaQ

Anyone know anything/hear anything about Propeller processors? From their promo copy on the site, it seems to not need an OS to function properly. https://www.parallax.com/propeller/

>>8051 Thanks for posting this. It looks very interesting. It's some micro controller which special innovations. Also supports quite some programming languages to use it, one being a visual one for beginners, similar to Scratch. However, this ecosystem strikes me as rather expensive. Arduino is open source and available as cheap clones from China, this here seem to be tailored for education and ease of use. I guess it's good for that, but will be more expensive. I just figure that from a price of a sensor package of ~1600$, though I didn't look much further, might be a lot of them for a whole class or school. This whole concept might also be more relevant for small robots, where there isn't so much space for many controller boards and so the one needs to be more powerful.

Sorry I couldn't resist. :^)

Nano /g/ had an interesting post about what appears to be a PDA. Haven't checked into this yet, so caveat emptor. >What about this? Ignore the price for a second... >https://www.crowdsupply.com/sutajio-kosagi/precursor >Verifiable botnet-free. You can even build the gateware for the SoC yourself. A similar setup could be achieved with a custom pcb and probably just more ice40up5k's, at the cost of effort and bulk. >I'm interested in having a portable device that can use technologies free from the internet/infrastructure you pay for that is run by corporations. For example, https://www.meshtastic.org/ which is a mesh network using LoRa. If it's 'verifiable' botnet-free then that would certainly be a big step forward for freedom. aHR0cDovL25hbm9jaGFucXpheXR3bHlkeWtiZzVueGtneWp4azN6c3JjdHh1b3hkbWJ4NWpiaDJ5 ZHlwcmlkLm9uaW9uL2cvMzQ2MDUjcG9zdDM1NDE2Cg==

>>8146 The use of FPGAs has already been mentioned briefly here >>6861 in the alternative ISA thread. I don't expect ISA implementations on FPGAs to be a huge hit as you're paying an enormous premium for security or emulation cycle accuracy compared to exponentially more computing power a locked down but affordable mass produced SBC using ARM or x86-64 provides. Maybe as a password manager/crypto wallet/YubiKey that PDA project will find an audience.

>>8148 >Maybe as a password manager/crypto wallet/YubiKey that PDA project will find an audience. I think I understand. Well, I think it's probably going to be a common scenario to have some kind of home server/routing system serving as a protected gateway for our robowaifus out into the broader world, and to offload local computations. Maybe some kind of FPGA solutions can be a part of those. Thanks for the feedback Anon.

There are small devices which can run ML models coming from TinyML. Maybe this is usefull for some reflexes or chaper waifus. It coming from Nordic Semiconductor in a partnership with Edge Impulse. It seem to work for continuous motion recognition, responding to voice and recognizing sounds from audio. https://www.nordicsemi.com/News/2021/01/Edge-Inpulse-and-Nordic-partnership

>>8160 >BlueTooth IoT While vaguely similar to our use case, one of the things we're quite concerned with here on /robowaifu/ is botnet infestations in our waifus. The entire IoT world is famous at this stage for blatantly embracing just that (that supports these corpo's business models, after all). So, I'll look into this but with a grain of salt. But regardless, it's affirming to see groups focusing on issues we predicted here all the way back at our beginning. Hopefully innocuous systems will appear eventually. >tinyML Thanks I didn't know about them. Seems right up our alley if they are opensource. I'll be researching them and will post in one of the AI threads if it seems good. Thanks Anon. https://www.tinyml.org/home/

>>8163 >update: In pretty short order I've discovered enough opensource hardware and software to probably be able to start an 'AI-on-microcontrollers' thread here, primarily ones based on TensorFlow Lite for Microcontrollers for now. https://www.tensorflow.org/lite/microcontrollers Thanks again, stay tuned.

Here a comparison video between Teensy, Arduino and Arduino clones: https://youtu.be/75IvTqRwNsE - Teensy has some advantages, but cost 20$ and has a proprietory bootloader, also seems to be a bit harder to learn for beginners. I bought cheep Arduino clones, but haven't tested them yet, for lack of time. So I don't know if thy are relly bad. I think for prototyping they will most likely be good enough. ESPs are also an option, which are cheap but harder to get into.

>>8165 I don't get the beef he had with cheap arduino clones. Those are pretty much all I use and I found most of the errors come from my breadboard rather than the quality of the chips. ESP8266s are good, especially the D1 Minis. They only have a few digital pins but you can easily download their libraries in the Arduino IDE and are as close to plug and play, especially with the Blynk app that sets up the wifi for you. Though I want to get off proprietary eventually, it's still a good stepping stone. I've tried ESP32s but the only reason to use them is because they're fast enough to stream videos, thus most ESP combo chips are camera chips, I have a couple of ESP32-Cams heading my way. The absolute cheapest though is the STM32, in the ESP32 you had to hold the reset button while uploading, well in this one you have to change the jumper every time to boot mode from running mode. The STM32 was impossible to get running until I realized I was using the Windows Store version of Arduino instead of the simple download version. Remember the electrical engineer back on 8chan who taught lessons on AVR using Debian and register notations, even when what we were using already had ready libraries in Arduino IDE? Well crank that up a notch and what you get is literally multiplexed registers which only someone who slept on the STM32 datasheet can begin to comprehend. Thankfully there are some libraries of the common use cases already. In summary these are the use cases which I justify for the following: *arduino clone -- if you just want to flash an led or drive a DC motor *STM32 - if you want to drive a 1.8" display or have tons of interrupts (e.g. reading a radio control receiver) *ESP8266 - if you want to get into those app-controlled internet of things (there is a noticeable delay though from touching the virtual button and the LED lighting up) *ESP32 - similar to ESP8266, but the justification would be it's faster so better with cameras, so just get them combined with camera modules, they're just 5-10 bucks. The closest I can get to DIY wireless FPV without shelling out for those sketchy 5 Gigahertz Chinese things. Oh yeah, I have a Raspberry Pi and Jetson Nano but I consider them to be small form factor computer desktops rather than something I'd embed then forget.

>>8366 It would sure be nice if you taught everyone else here how to do this kind of stuff. Ever consider a /robowaifu/ teaching thread on these topics Anon?

Found out on nano/g about a new Raspberry Foundation's new Arduino microcontroller competitor called Pi Pico. >"$4 Raspberry Pi Pico board features RP2040 dual-core Cortex-M0+ MCU" https://web.archive.org/web/20210121070023/https://www.cnx-software.com/2021/01/21/raspberry-pi-pico-board-features-rp2040-dual-core-cortex-m0-mcu/

>>8437 related https://electronics360.globalspec.com/article/16251/video-raspberry-pi-launches-4-board-built-with-its-first-proprietary-mcu

Wiki dedicated to Arm-based IoT & embedded. Seems like it's been around for a number of years. https://wiki.friendlyarm.com/wiki/index.php/Main_Page https://github.com/friendlyarm

Simple-to-install RaspberryPi 3 & 4 images, with Ubuntu 20.04 LTS + ROS Noetic included & preconfigured. Probably the single easiest/fastest way to get started with ROS. https://learn.ubiquityrobotics.com/noetic_pi_image_downloads

(>>16420 >low-cost AI -related)

USB C powered SBC with integrated Arduino Leonardo. Thought it might be of interest. https://www.dfrobot.com/product-1728.html

>>16467 Very neat Anon, great idea they had to do a combo like that. Definitely a little pricey for what we're going for here in general, but overall probably a decent technical approach. A couple more-modest SBCs (RPi, etc), a handful of MCs and I'm guessing that would cover most of the onboard-compute needs of a entry-tier basic robowaifu model.

Just found out that Jetson Nano B01 seems to be back in supply and available for US$150. Xavier might be overkill to get started, but the Nano might be a good idea.

Should I learn Arduino if I wanted to get serious in robotics? I mean, do research, get a career, not just as a hobby. I heard Arduino is only good for hobbyists and I should learn more about microcontrollers, FPGAs if I want to go deep into robotics. Would Arduino be a waste of miney and time in that case?

>>19419 What do you mean by getting serious? If you study it, then you might not learn Arduino but go more into the basics, but could still pick it up for hobby projects. If you try to build your own projects and interacting with hobbyists it's probably unavoidable. Arduinos are still cheap, expensive compared to other micro-controllers, but easier to use.

>>19423 sorry, I meant if I decided to become a researcher on robotics, not a hobbyist. They probably don't use Arduino right? What do the big players in robotics like Tesla, BD use? I wanna learn those. I'mmore specifically interested in humanoid robots.

>>19424 > researcher on robotics, not a hobbyist. They probably don't use Arduino right? Ask them or look into some forums in regards to that topic. They might go for more basic knowledge or still doing Arduino or MicroPython if they want to focus on fast and high level development. Even more likely they're using ROS. Maybe ask some or the organizations mentioned here >>268

https://www.reddit.com/r/linuxhardware/comments/r60lcp/rk3588_is_3_times_faster_than_the_raspberry_pi_4/ >5-6 times faster in AI workload like OpenAI's Whisper >The December 17th, 2022 release of DietPi v8.12 comes with a new image for the NanoPi R6S and the Radxa ROCK 5B and a couple of improvements and bug fixes. https://dietpi.com/docs/releases/v8_12/ Video review of Orange Pi 5 (RK3588S): https://www.youtube.com/watch?v=fHQ1GpYRHPo >warning: RK3588S is 1/4th as fast in reading an NVME compared to RK3588 (Rock 5B) >good software support for Orange Pi 5 and renders videos even a bit faster than a Rock 5B >before watching: His reviews are focused on using those SBCs as desktops, not AI and robotics. >>19419 >>19424 Please stop cluttering this board with detailed questions on careers. This robowaifu board is dedicated to building robowaifus, not general career and educational advice. There are big subs on Reddit on robotics and many other websites with such info's exist. Here are not enough people to give you such specific advice. If you don't want to "waste time" learning Arduino, just learn general electronics from the scratch. The more, the better, I guess.

>>19428 >Watch out for some of the higher density drives. They can suck the juice. With my RPI-4s, I found that I had to stick with Samsung 500G T5s and T7s as the boot drive. Other SSDs would cause the RPIs to reboot, lockup, do funny things….. Of course you could plug in a powered USB 3.0 Hub and run your large drive from there. Always an option! https://hackaday.com/2022/12/27/dietpi-releases-8-12-with-support-for-the-rockchip-rk3588-soc/

We will need better sources and understanding to compare the more expensive SBCs. For example, the TOPS benchmark can be misleading. Also ... > "a decisive factor for many industrial applications is the long-term availability" ... High TOPS might also not be necessary, because of the limits of input data, like framerate of cameras. >Where a lot of TOPS are really needed is in “real-time critical” applications. Some SBC with a "i.MX 8M Plus"-based SBC might be well suited for robots at home, which also should have long term support. It also might have more support than more exotic NLUs (neural processing unit), because of a more widespread use, independent from the support of the company behind the SBC or NLU. >Excessive TOPS have no impact if the NPU is not supplied with enough data. >Do I Need an NPU? ... if there is a task that can be handled by “classic machine learning” (support vector machine, tree learner, …), the use of a CPU is sufficient in many cases. Picrel: High level speech seems to require 2-4 TOPS in a neural processing unit (NPU) https://towardsdatascience.com/when-tops-are-misleading-70b53e280c39 Ideal would be some channel like ExplainingComputers but for such boards.

For those still trying to use a Raspberry PI, here's a availability tracker: https://rpilocator.com/ - This is mostly as confirmation to use other alternatives, especially since we only need some server which might not even be (directly) connected to the internet. Or not at all if it controls the robowaifu or handles sensitive data. The software for the alternatives is sometimes more outdated, but this might or might not matter. >>19435 Another long article about NLUs, and how to compare them. This might not only be relevant for single board computers (SBCs) but also home servers. https://www.expedera.com/blog/2022/05/27/measuring-npu-performance/ There are various factors being relevant: >Frequency as a Variant in TOPS/W >Neural Network(s) Employed as a Variant in TOPS/W >Precision as a Variant in TOPS/W >Sparsity and Pruning as a Variant in TOPS/W >Process Node as a Variant in TOPS/W >Memory Power Consumption as a Variant in TOPS/W >Utilization as a Variant in TOPS/W

>>19428 >Please stop cluttering this board with detailed questions on careers. This robowaifu board is dedicated to building robowaifus, not general career and educational advice. That's my fault Anon, please pardon me. Now that you put it in that light I see your point. I'm the one that specifically directed him to ask in this thread. I should have sent him to our /meta instead, where such questions are perfectly reasonable & welcome.

>>19438 No problem, but I think I already answered him at least once. Or it were different anons, idk. Here seem to be more and more people asking for career advice. Most of us have no career in robotics (or at all) and don't try to have one, how should we know? It's not a good place to ask, especially when somebody already knows the company he want's to go for.

>>19435 >NLUs (neural processing unit), because of a more widespread use, independent from the support of the company behind the SBC or NLU. Lol. *NPU not NLU

>>19435 This processor has 6 TOPS https://www.rock-chips.com/a/en/products/RK35_Series/2022/0926/1660.html A decent low-cost option to try this SOC https://www.cnx-software.com/2022/11/11/orange-pi-5-most-affordable-rockchip-rk3588s-sbc/

>>19514 >orange pi 5 Seems pretty impressive Kiwi.

>>19514 >>19515 >This processor has 6 TOPS Yes it's mentioned in >>19428 - It looks quite interesting for running Whisper. But I would like to see how fast this actually is, and how we can optimize it's reaction time. There are quite some interesting new SBC coming out from time to time. Not sure if I should post an update every time I watch a video on one, or just see one available. >Orange Pi 5 slower data read speeds than the Rock 5B >Rock 5B: This isn't the Pi killer you're looking for Fast but expensive. https://www.youtube.com/watch?v=CxD_0q8tAdc it makes a good point in saying if you want to spend more than 100$/€, maybe consider a cheap mini PC. He obviously doesn't think of mobile systems like robots, but maybe the energy consumption isn't even that relevant compared to the actuators. >VisionFive 2: RISC-V Quad Core Low Cost SBC https://www.youtube.com/watch?v=ykKnc86UtXg - probably the closest to Raspi 3 in hardware for a good price, but lack of software support. Best current APU circa as fast as the GTX 1630: https://www.youtube.com/watch?v=_KfaRMDpijY The best cheap alternative to just have something like a Raspi seems to be Le Potato, if it's about having some SBC, not about speed and newest features. It's not a competitor to the Raspi 3 or even 4. Obviously also not good for AI workload. But it's barely more expensive than a Raspi Zero but I think it has better I/O connectors. I think for running some AI workload, especially in a MOBILE system, the NPU is important. Jetson Nano, Nvidia in general, Google Coral, and these RK3588 might be some good options. Price and availability matters. Otherwise, I already summed up the advice from Towardsdatascience here >>19435 and that the TOPS value is not the only factor to look at >>19435 and >>19436.

>>19516 Wealth of information Anon. Please keep it up when you feel inclined, it helps everyone here. >Le Potato Lel what a great name. :^)

>>19518 Things to look into: - How do these GPUs on these mini PCs compare to NPUs on single board computers? - Can we add some additional NPU to such a mini PC? - Testing Whisper and other models on some of these SBCs and mini PCs. - Comparing the newer RK3588 to the older but expensive one like Coral, Jetson Nano, etc. or find such comparisons. - Looking into more professional "i.MX 8M Plus"-based SBCs. ... >>19518 Thanks, but I think it's better to not report on every new SBC or alternatively just report on the ones which one has randomly encountered. Maybe better to report on those one actually buys and otherwise discuss general developments of the market and how these options relate to each other. I'm also looking forward to comparison videos, like the one ExplainingComputers has announced.

>>19522 >better to not report on every new SBC That said... MangoPi MQ-PRO: https://www.youtube.com/watch?v=m3LwGK-u8y0 MangoPi MQ-QUAD: https://www.youtube.com/watch?v=6WvxZRcMhc4 It might be a bit cheaper than Le Potato, but it's better in terms of many features (except maybe connectors). Also it's probably more available than Raspi Zero, faster than it, and it SUPPOSED TO BE only a bit more expensive or available for the same price (it isn't). MangoPi MQ-QUAD for 46€: https://www.aliexpress.com/item/1005004703545539.html vs https://rpilocator.com/ with Zero2 for 30€ While we're at it: VisionFive 2 for $128.95 USD https://ameridroid.com/products/visionfive-2 hahah, it was supposed to be $50-$60.

>>19522 >- How do these GPUs on these mini PCs compare to NPUs on single board computers? My primary concern about GPUs or other ancillary processors is their API. For instance it is (or at least was) difficult to capitalize on the RPi Broadcom GPU b/c AFAICT there wasn't a public API for using it. Obviously robowaifus have umpteen zillion different uses for matrix operations (GPU/NPU specialties) but you can't really take advantage of that hardware easiliy w/o a unified way to communicate with it.

>>19590 I think the Raspi/Broadcom thing was a special case. These NPUs are made to be used for DL models. Also, I think the Broadcom API for the Raspi got reverse engineered and open sourced at some point.

Any kind of pi is bad value versus just a phone. Prove me wrong.

>>19701 Smartphones don't just cost $40 like Le Potato or have a NPU while going for 70-100USD. Also phones are never secure, you would at least remove the UMTS module and then it would still be less open source and more likely be bugged from the start than a dev board. For the same reason, it might be hard to get a normal Linux or BSD running. Then connectors... where to plugin things? It's absurd.

>>19705 >Smartphones don't just cost $40 like Le PotatoNo they cost less lol I don't know what a NPU is though ngl. As for getting anything and everything to run on it you just root it.

>>19716 Isn't that a pretty limited set of devices today Anon? I mean for recent (much more-powerful) hardware. It's a delicate and tedious process, and therefore costly to perform.

>>19718 I think any pocco beats the potato in performance, price and features 2gb of ram right?

>>19720 What's a pocco though?

>>19721 Pocco cellphones

>>19726 Well I guess while you can install anything on it you wouldn't be able to use input on it afterwards. But there is stuff such as andronix and linux deploy https://play.google.com/store/apps/details?id=ru.meefik.linuxdeploy&hl=en&gl=US&pli=1 I think linux deploy requires root access

>>19726 >Pocco cellphones Chinese spy phones for some hundred euros. With sensors we most likely won't be able to access, and a lot of expensive things we won't need. Especially not everything in one device. I want open source firmware and GPIO pins.

>"The BreadBee is based on a relatively unknown IP camera SoC, the MSC313E, from a company called MStar," Palmer explains. "You might have never heard of MStar but you probably have one of their chips in your TV. There are no solutions yet that can run a real OS like Linux that can be integrated by hobbyists into boards from cheap PCB vendors that don't have show stopping limitations; i.e. the AllWinner V3s is very similar to the SoC used here but it only has one SPI controller that is lost as soon as you put SPI NOR on it. https://www.hackster.io/news/daniel-palmer-s-breadbee-is-an-ultra-compact-1ghz-arm-cortex-a7-sbc-with-on-board-ethernet-577cab543154 Let's see how well this is going to be supported. Seems to be a more powerful alternative to an Arduino or something similar, while still being very small.

>>20763 Very nice find, Anon. Here's a perspective view of the device. >

I just read about the ESP8266 and it's the newest microcontroller that is better than Arduino in every way. It even comes with micropython and is even more affordable.

This here is about software, but it shows that FPGAs might become more useful: >>21588 Related: https://digilent.com/reference/programmable-logic/zedboard/start >ZedBoard is a low-cost development board for the Xilinx Zynq-7000 all programmable SoC (AP SoC). This board contains everything necessary to create a Linux®, Android®, Windows®, or other OS/RTOS based design.

>Sipeed M3AXPI artificial intelligence vision AXERA AX620A 14.4 Tops NPU AI ISP low light night vision Linux development 2G RAM https://www.youyeetoo.com/products/sipeed-m3axpi-artificial-intelligence-vision-axera-ax620a-144-tops-npu-ai-isp-low-light-night-vision-linux-development-2g-ram?VariantsId=10633 Can also be found on AliExpress.

>>23398 That almost looks like they adapted a DDR4 memory slot as the host to the SoC daughterboard. Interesting, thanks Anon!

>>23399 Most sipeed boards are designed like that. They have some FPGA and RISC-V boards too in similar daughter board designs. I could imagine them being useful in a modular custom motherboard that incorporates all of them Sipeed Store https://a.aliexpress.com/_EJ3Vr4H

>>23398 >>23400 Thanks, looks very promising. I just wish we had an easy way to compare all such SBCs. Also, I recall that there was an argument how TOPs isn't a good criteria for comparison. I have to look into this again. 70€ (plus shipping) for something in that range is interesting anyways. I'm very tempted to directly order one. Once thing I would like to do with a board that allows for more than one camera, would be to have a way to use this for creating a somewhat 3D model of the world. Especially be able to know the distance of an object it recognizes. This will be absolutely crucial to understand the world.

>>23405 >Especially be able to know the distance of an object it recognizes. This is pretty much how so-called Stereo Depth Cameras work, using triangulation. BTW, a well-calibrated single camera can make rough distance estimates (within limits), particularly when the viewed objects have already-known textures. Eg; https://stackoverflow.com/questions/14038002/opencv-how-to-calculate-distance-between-camera-and-object-using-image Here are some practical solutions related to these topics NoidoDev, opensauce. https://docs.opencv.org/4.x/d6/d55/tutorial_table_of_content_calib3d.html >=== -add OpenCV hotlink -prose edit

>>23410 >A well-calibrated single camera can make rough estimates within limits it would have to be constantly bobbing its head from side to side like a pajeet to do triangulation, creepy af, would rather just pay for a second camera to avoid a pajeetfu

>>23410 Thanks. I guess it doesn't matter if the pictures come from two cameras. >>23413 LOL, I agree. I want two cams.

>>23413 >it would have to be constantly bobbing its head from side to side like a pajeet to do triangulation, creepy af, Lol. No not really. If you study the material, you can see that it works just as well for a still image. Motion isn't req'd. >>23416 >I want two cams. Agreed. Simply pointing out the possibilities. In fact auto-mesh generation has roughly the same demands/needs as depth-registration, and I find that tech far more exciting ATM. BTW, we're offtopic ITT. If you'd like to continue with it, please migrate over to our Vision thread (>>97). TIA.

>>23398 >>Sipeed M3AXPI Review (with strong eastern European accent): https://youtu.be/3q89nC_tHE0 - tldw;; Bad documentation and support, lot's of problems. You'll need C++ to work with it and write your own software to make it work, lol. Might be interesting for mass production, but not for hobbyists and devs. Data sheet for comparison: https://docs.google.com/spreadsheets/d/1BMj8WImysOSuiT-6O3g15gqHnYF-pUGUhi8VmhhAat4/edit#gid=0 This video here is about the version 2 (Maix-II SoM), also is a bit hard to understand and it's two years old: https://youtu.be/o2dXIO8-pvg but it shows more data for comparison. Based on what I gathered, some boards like this might be cheaper than the ones from Nvida or Khadas due to lower RAM, but good enough for many use cases. For example if you don't need to recognize many faces in a crowd.

>>23483 Nice research. Thanks!

Hailo offers an expensive but industrial alternative to Nvidia and Khadas, selling 10x10 SBCs for object recognition in up to 3 streams: https://up-shop.org/up-squared-pro-series.html plus the AI CORE with 2x Movidius Myriad X VPU: https://up-shop.org/default/up-ai-core-xm-2280-myriad-x-2.html (500+ Euros)

I'm anti raspberry pi because they're too expensive for what they are which is having less functionality than a used phone. Ive found something called ft323h which turns USB into gpio pins.

>>23617 > used phone > ft323h which turns USB into gpio pins Interesting, I wondered about this problem, but yes that sounds like a solution. I'm still skeptical, since you need something similar to the shell and pipes to make programs work together. Not sure if this works so well on Android. Anyways, so you want to go ultra cheap in every area. That's fine, not for everyone, but for some guys it might be the way to get a waifu. Just make sure your setup is independent from a specific phone. >raspberry pi because they're too expensive You know about ESP which is very often mentioned starting here >>1406 and also LePotato https://libre.computer/products/aml-s905x-cc/ (which is now maybe a bit overpriced as well, they went up a bit since they got more customers and buzz around their product, but still 30€ without the heatsink and Wifi, 35€ for the 2GB version).

>>23641 You can use the lgpio python module to control the pins and then do that via termux. Atleast that's the plan.

>>23641 >but for some guys it might be the way to get a waifu. I would suggest that will be the case for the vast majority of men who assemble a robowaifu on their own.

>>23641 Le potato is more reasonable, but they don't sell them around here and really they're not that powerful. They have 2gb of memory. We just need one old phone to the central unit which connect wirelessly to the computer to power the rest of the stuff. I however like the raspberry pi pico.

>Ampere Altra Dev Kit Not strictly SBCs but worth to keep an eye on: ARM servers by Ampere: https://www.ipi.wiki/products/com-hpc-ampere-altra More cores, means good for ML. Their top system has 128 cores (it's also costs more than 5k): https://www.ipi.wiki/products/ampere-altra-developer-platform?variant=42970872053922

Cortex-M (Blackpill) seems to be very power efficient, compared to Raspi Pico and ESPs, RISC-V variants are not more efficient but somewhere in the middle: https://www.youtube.com/watch?v=hdnONr-cXds

>>23927 Nice, thanks! Power efficiency will be key for production robowaifus.

Beagleboard has differentiated into a wide array of different boards while I was paying attention elsewhere hehe. :^) https://www.beagleboard.org/boards

>>24653 Thanks, these might be very interesting. These boards are also available. But we would have to look at one of the comparisons again >>19435 Related to the whole topic of SBCs is a new AI framework Tinygrad, which will increase support for other GPUs and accelerators than the ones from Nvidia. >tinygrad supports various accelerators out of the box, including CPU, GPU (OpenCL), C Code (Clang), LLVM, METAL, CUDA, Triton, and even PyTorch. This also seems to be about DSPs: >How can DSPs scale their ML use cases and gain a competitive advantage? https://www.telecomramblings.com/2022/03/how-can-dsps-scale-their-ml-use-cases-and-gain-a-competitive-advantage/ >Digital Signal Processor (DSP) https://semiengineering.com/knowledge_centers/integrated-circuit/ic-types/processors/digital-signal-processor-dsp/

>>24656 >But we would have to look at one of the comparisons again Yep I feel you on this one Anon. I would have expected there to be some man out there doing such comparisons regularly. I suppose the wide variety of board types and manufacturers would make managing such roundups a headache. Let's hope some objective comparisons happen regardless! Cheers. :^)

https://sipeed.com/licheepi4a https://youtu.be/1apoFXZ9ad8 >CPU: TH1520, 12nm, RISC-V 2.0G C910 x4 >RAM: 4 / 8 / 16 GB 64bit LPDDR4X-3733 >NPU: 4TOPS@INT8 1GHz, Tensorflow, ONNX, Caffe >GPU: 50.7GFLOPS, Fill 3168M pixels/s >155-206€

Let's get something out of the way. You can have an affordable waifu with an API or you can have a cost prohibitive waifu without an API. Even then it won't be able to run the best models such as koboldAI 20B models. It is wasteful and pointless.

>>24787 "If you don't have anything good to say, then don't say anything at all" my mother always told me.

>>24656 > GPUs and accelerators I can not say I know the answer to this but I suspect that if you buy commodity class general purpose processors then you could get more bang for the AI buck. Of course I readily admit I may be really wrong. I'm also talking about a specific class of computing. I'm not talking about high end, I mean comparing relative computing power of medium to low power devices.

>>24809 If you guys must have the AI not be in the cloud the best option would be a thin client at that point I think. Some thin clients come with a 1080ti last time I checked. There might be some that come with even better gpus now. I have an rtx 3060 and I've tried to use kobold AI with it and the stuff is capable of running is just not good. Hell I bought cloud instances with 4090 and not even those can run the 20b models. With this also the power requirements go way up. You can run your own model but realistically it'd be via an API.

>>24811 >If you guys must have the AI not be in the cloud We here will never recommend Anons run their waifu's 'mind' up on the Globohomo's cloud (or indeed have much of anything to do with them at all), and for good reasons : (>>20208). You want to use them for training/fine-tuning your own offline models for your waifus that's one thing. Actually depending on them for your waifu's very runtime operations themselves? No way, José. >=== -prose edit

>>24811 This thread is about SBCs (Single Board Computers), not about general debates on how to deal with compute. Most likely everyone will use SBCs, then servers at home with a VPN (tinc), and then on top of that we will have different ideas about how much connection to the net and what a "direct connection" even means when we already have a network of computers.

the reason its taken so long is because I was hesitant to buy a raspberry pi zero w, then I was hesitant to buy another sd card after the last one worked, then the last one worked but linux wouldn't boot all the way, so I hesitated some more cause the sd card with raspberry pi os cost roughly as much as the pi zero w itself... If it doesn't work when it arrives I'm going to light it on fire... I also had to wait for flux to solder the pins.

>>25095 >buying SD cards This is why it pays to make friends with your local recycling center (or sketchy phone repair store, if you're so inclined). You'd be amazed at how many people leave the microSD cards in their phones when they're thrown away in blissful ignorance of vultures like me who hoard whatever vaguely useful storage mediums we can find. They'll never know the fate I've saved them from by getting to their drives before any data stealing vultures could swoop in.

Not sure if this is the right board to ask this. Lately i was thinking about using wiper motors converted into servos and seen lot of videos where people rip apart standart model servos for board and potenciometer. While i ordered servos for prototyping a small proof of concept project i'd rather not waste them needlessly. If i wanted to buy just the boards instead of scraping perfectly fine tho small servo is there some keyword i could search for?

>>25099 The most common term I'm aware of is 'driver boards' or just 'drivers'. But today (and particularly in the context of building your own robowaifu from scratch) it seems to me that using a general purpose SBC/MC is the more reasonable approach. They're commonplace, cheap, flexible, and you're never going to have enough GPIOs 'laying around', right Anon? Good luck with your project research Anon. Good to hear from you again! Cheers. :^) >=== -minor edit

>>25098 You're absolutely right I expect, Greentext. I wish I had such a working relationship going myself.

>>25105 SBCs aren't really a replacement for motor driver boards. These boards are way cheaper or more expensive, but also maybe an add on to the SBC. Just look into some YouTube video. First you need to know what kind of motor it is (DC, BLDC (or the sub variant of stepper), then look for the right control boards, and first the explainer on some website or Youtube videos. We have a better thread for this, though: >>12810

>>25105 I've seen drivers before and they are generally h bridges with speed control which is far from servo functionality i need. I guess i could slap potenciometer or optical enkoder connected to analog input but i would quickly run out of those. Are input expansion modules a common thing?

>>25113 Ok, found meself some multiplexes and encoder. Gotta run the math if using them insted makes financial sense.

>>25111 Yep you're right NoidoDev. I even knew that but for some reason wasn't thinking clearly when I posted. >>25113 The other anon above is right. GPIO pins aren't meant to run a lot of power through them, merely signal-level current. As he pointed out 'motor driver board' is the correct keyword phrase. There are a variety of use-cases, which (I presume) your wiper motors will be straightforward brushed DC motors. Here's an example source link: https://www.pololu.com/category/11/brushed-dc-motor-drivers As to all the other ancillary needs, there are many possible ways to approach it. I prefer the Beaglebone Blue generally for prototyping, since robotics was the particular design goal behind that board's creation. For example with facial animation systems, very likely just two BBB's could directly run all the low-power servos needed for very sophisticated motions (say 15-20 smol servos), and could also accomodate the external encoders/sensors needed. And on top of that, they are general-purpose computers that can run Linux, support cameras, sound, networking, storage, etc. OTOH, its improbable they could run something as big as a hefty wiper motor all alone. But they could still be used to send the PWM signals needed, out to the drivers to run your wiper motors (though there are much cheaper alternatives to that for production designs). BTW, do you have a more 'personal' handle we can all refer to you by here, Anon? Cheers. :^) >=== -prose edit -add hotlink

>>25098 I don't know why the sd card with raspbi os burnt won't work. It makes no sense. But again I bought one that comes preinstalled with it and again if it doesn't work I'll return it and burn the raspberry pi zero w.

>>25142 I'm not too familiar with the Raspberry Pi platform, but shouldn't it be possible to just install the OS onto a blank SD card like a normal computer? Alternately, couldn't you download the OEM image and write it to an SD card with dd?

>>25142 >>25146 >couldn't you download the OEM image and write it to an SD card with dd? Yes, and I detailed the process here a while back. (>>5165).

>>25142 Aren't you the guy who tried to build a hasel actuator? You will be able to figure this out. Just don't us dd to overwrite your computer disk or so. Have backups or maybe use another program.

>>25111 I don't mean to be disrespectful anon. What i came for is not asking about motor but about additional boards that would come between said motor and microcontroler. Substantiate your claim that it should be in general actuator thread. >>25135 I'm useing PCA9685 for i did not intend to run power through Arduino anyway. End goal jhumanoid robot would use no less than 32 individual motors and goal of current endeavour is to build sufficient infrastructure to move joints w/o any added load and considering signal loss and power projection if boards and circutry should be centralized or disributed. Using servos or servo board modified wipers allows for much simpler command structure: arduino->PCA->servo as i do not need any additional feedback. while if i had to use feedback it would something look like: encoder->input_multiplex->arduino->PCA->motor driver->motor which adds lot of complexity and cable creep for no other benefit other than system could internaly monitor joint positions instead of ordering servos and leting them to do their work on their own. >do you have a more 'personal' handle we can all refer to you by here, Anon? I'm not big on using trips be it on thai crab wrestling or robo building imageboard but i may change it to "Podolek" if (You) prefer it.

>>25151 I don't want to sound like a schizo but I think I'm being targeted by hackers. It wouldn't be the first time. But I'm pretty persistent. I'm going to try to do it in a vm. I don't think they can hack their way into that.

>>25170 >PCA9685 >arduino->PCA->servo Yes that sounds like a great solution Anon. >thai crab wrestling or robo building imageboard Lol. Podolek sounds like a good handle, but we can just call you 'Anon', Anon and it seems like you'd prefer that. Good luck. Please keep /robowaifu/ abreast of your progress! Cheers.

>>25171 Heh, I doubt you're a bigger schizo sh*teposter than yours truly, Pete. I'd at least encourage you to stay grounded in some kind of objective reality though (for example, there's a ginormous weight of evidence now for the Globohomo and their machinations at the state & corporate level). >I'm going to try to do it in a vm. I'd recommend that as a first exploratory step for beginners Anon. Just ask for help and I'm sure some anons can at least point you in the general direction of the answers. Cheers.

>>25173 there is nothing to ask. Raspberry pi os writer is supposed to be straightforward and fool proof. Anyways other than that I used wireshark and just syn ddos attacked all that crap idc. I'm assuming the hacker prob ran out of vpns now.

>>25170 We don't use SBCs just for motor control, these are small general purpose computers. But your question is closer related to actuators. We very likely discussed these things in the actuator thread, where I would post it, so I would also look there for such info. This might also help: >Embedded Programming Group Learning Thread 001 >>367 >>25171 Make sure to write your most important data on an M-Disk. A SD card failing is not an indicator for being under attack, though.

>>25176 Fine, as you wish. Hope it's not as dead as it's been last few days.

>>25178 >Hope it's not as dead as it's been last few days. Be the future you want to see :^)

>>25099 >Cannibalize cheap servos for wiper motor servo Not worth it. Those boards are cheap chinese junk with terrible pots. Using a real microcontroller would be far batter. Assuming you are a beginner, starting with an Arduino and an easy tutorial will help you build the skills you want. https://www.instructables.com/Wiper-Motor-and-Arduino-Mega-Servo/

I got the raspberry pi zero w cause I thought it'd make things smoother. I have spent like a month altogether on that junk... My last video was like a month ago lol. I got it to run but it says no carrier for the wifi. I did not have any issues with the orange pi ironically but its gpio pin support sucks so I'm going to have to find a way to combine it with the esp and that's it. I don't know what I'm going to do with that raspberry pi I might throw it it in the trash I don't want to look at it anymore.

>>25207 You do know that Raspi has it's own forum with troubleshooting? https://forums.raspberrypi.com/viewtopic.php?t=313806

>>25188 >Those boards are cheap chinese junk with terrible pots. Alrightyou convinced me. Last thing i wanna do is troubleshhoot & run wires on those tiny boards.

I ran across something really interesting. I think. Apparently server centers are selling their used "Nvidia Tesla K40 12GB GDDR5 Graphics Card GPU Accelerator Card w/ Low Bracket". Look up the above in ebay. They're like $32 USD. The power of these things is immense and blows the living hell out of any sort of single board computer. Of course the trick with them is, 1. Programming a huge parallel processor will be difficult. 2. They don't have fans because data centers have a mass of cool air blowing through them. You will need one or a better solution some sort of pumped cooling.(there is apparently some solutions to this but I don't know exactly what that is yet) Some solutions to the programming are they have a library called CUDA. I see this is used in open source video upscale software. They use these for machine learning, engineering and...not sure what else. Maybe they also use them for general computing "if" they can code this mass of parallel processors. Other problems are you will have to have specific motherboards and you will have to get a monster beast power supply somewhere around 800W. Specifications 2880 CUDA GPU cores (GK110b) 4.2 TFLOPS single; 1.4 TFLOPS double-precision 12GB GDDR5 memory Memory bandwidth up to 288 GB/s PCI-E x16 Gen3 interface to system GPU Boost increased clock speeds Supports Dynamic Parallelism and HyperQ features Active and Passive heatsinks available for installation in workstations and specially-designed GPU servers The new GPU also leverages PCI-E 3.0 to achieve 4.2 TFLOPS!! Damn they are fast. And this sort of processing is exactly what you need for a waifu to process speech, movement and visual navigation and personal visual recognition. A ESP32 micro-controller has 600MIPS it would take 7,000 of them to equal this one card in power, roughly. At $8 a piece it would take $56,000 to equal it. I wonder if this is some one time deal where they dump a bunch on the market and no one yet has really found a way to use them. This is changing fast because people have already figured a way to run games with them. While not optimal for this the price of graphics cards being so high I bet soon the prices will go way up. I'm thinking about buying one of these even if I don't ever do anything with it. The power if I could ever figure a way use it would make it a steal. They could be used to train. I'm guessing most of us will use mostly pretrained open source models then add extra training for what we want. These would be slow compared to these huge data centers but if you let it run for a few weeks it could do the same and you don't have your data sent to some data center.

>>25396 I mentioned the K80 and the M40 before, it requires additional tinkering with the cooling to make it work. If you have barely any money then maybe the K40 is worth trying, otherwise it's better to use the K80 which has internally 2 GPUs each with 12GB. This said, these old GPUs also aren't well supported, so you might need to compile more basic programs like CUDA and some models might not work. The M40 is better in that regard, and it has 24GB. These are also not SBCs but for the use in a home server to which your waifu can connect. OT: They don't make much sense for gaming, since they need to much energy for what they do, might require overclocking and even more cooling, and the GPU prices came down. If I wanted to run the biggest Falcon model at home, then I would consider using the M40 (28 of them, ~1KW / 40-80€ per month energy costs, price for all seven servers: ~3-5K )

Yes it's obvious it''s not a single board computer but it fits here better than elsewhere and fits the basic idea of what SBC here are used for. I bought 2-K20's for under $30 plus shipping here, https://www.ebay.com/itm/314144513426 I may not ever do a single thing with them but...$30 for that kind of power...You never know. Some things they dump on the market surplus and they are super cheap then disappear. Some don't. If you look at the same seller, and others, they have HP 1200watt power supplies for $30 or less. There are break out boards where you can plug a card in to the PS for around $18. Some have cables, some not. People are using these for mining. I have an old computer I do nothing with. It has a slot that I think will take it. Get a PS and this older computer could be used to train. Maybe. The K20 is 60% of the K40. And no I'm not buying a K80 they're $160. This is just a whim. I'm not dogged serious about it but for $30 maybe I can piece something together if it comes up. Some K20 cards have passive cooling. I have no idea if the ones I bought do or not. I did see them online though. But I expect a fan would be good. They are designed for servers with good air flow. >=== -rm URI fingerprinting

>>25437 Okay, but make sure you know how to do take care of the cooling.

>>25396 >4.2 TFLOPS!! Nice. You'll need active cooling. Here's Nvidia's CUDA page. It's basically an extension of C programming (cf. >>20980) . https://developer.nvidia.com/cuda-toolkit >=== -add crosslink

As already mentioned in the meta-thread: The P40 (server GPU) also exists with 24GB. I didn't know. It's better than the M40. https://technical.city/en/video/Tesla-M40-24-GB-vs-Tesla-P40 - They need extra work with adding coolers or a cooler in a PCI slot.

This thread here becomes more and more a thread about GPUs. Maybe we should make one for that., or simply for AI related hardware topics in PCs and home servers. Right now these topics are spread over many threads, since in the past we didn't use this thread here for it. Anyways, I'm I the only one with a AMD APU in a PC? It seem not have been interesting for anyone that we can now run neural networks on such an APU. https://www.tomshardware.com/news/dollar95-amd-cpu-becomes-16gb-gpu-to-run-ai-software https://youtu.be/H9oaNZNJdrw https://youtu.be/HPO7fu7Vyw4 I didn't test it yet, but try to keep it in mind when I reboot. I need to change some Bios settings.

>>25759 I have a 5700G and it is honestly terrible for AI compared to a GPU. Yes, I can feed it heaps of RAM. It is too slow for it to matter though. My RTX 3060 undervolted and underclocked to 65 watts generates a Stable Diffusion image at 1024x768 with several LoRA's in 27 seconds. Considering a RTX 3060 can be had for 200 USD equivalent or even less if you're careful, it is hard to see the point

>>25769 One point is, that there are mini-PCs with APUs, which would fit into a robowaifu. That aside I'm more interested in trying it for some LLM. It could at least hold 16GB and some claim 64GB, but also if this works then sometimes only a few GB are missing in a PC/homeserver to run a model could come from the APU.

>>25770 LLM's do make sense for these systems. In which case, using LLaMAC++ (https://github.com/ggerganov/llama.cpp) is ideal. Those Zen cores are also very powerful and utilizing them with the iGPU can lead to good results. This hybrid approach is still being improved but, has great potential. I'm looking forward to further refinement in ROCm drivers. AMD has been making strides that have improved their performance in AI over the last year which should continue to benefit APU based systems.

Old chromebooks as alternative to SBCs: https://www.youtu.be/1qfSJxcgH5I (sometimes $99 for 10 of them, ofter cheaper). Of course, don't use an outdated version of Linux Mint like he might be in the video.

>>19428 More interesting development based around the RK3588's NPU. Useful Sensors have created a good library for using Transformer models on the NPU efficiently, called Useful Transformers. https://github.com/usefulsensors/useful-transformers Link to useful sensors https://usefulsensors.com/ Video demo of voice recognition and offline LLM response. https://youtu.be/K6qgN_QXH6M

>>26099 Hey that's probably a good idea, NoidoDev. >>26254 >Transformer models on the NPU efficiently Interesting.

There's progress in connecting (old) AMD GPUs to Raspis. This might become useful to run small models in mobile systems, I guess. https://youtu.be/BLg-1w2QayU - Work in progress, there are still issues.

>>26613 >There's progress in connecting (old) AMD GPUs to Raspis. Very cool. Lack of GPU dev for that Broadcom chipset has always been a shortcoming of the Pi. Maybe this will eventually turn into something important for this field. Appreciated, NoidoDev.

I've been thinking lately, Assuming that you're trying to build a full size humanoid body. Would it be better to have these single boards placed all over the body parts? Instead of having one powerful centralized board to micromanaged everything, your robowaifu will be a supercomputer, decentralized with many single boards throughout the humanoid body, Maybe this could solves the heat issue that generates from cpu? I'm not sure about the latency performance from this design though, I think there will be a hub somewhere that connects every single boards into one central board.

>>27359 There might be some merits to a system like you're describing, but there are some complicating factors. Firstly, each and every board will have to be powered one way or another. How they're powered will greatly affect the logistics of the setup. If they're powered by USB, then you have to either ensure that the central board can supply power to each peripheral board, or install a powered hub if that's not possible. Some might be powered by basic power outputs that exist on the central SBC. If so, you'll need to wire each one properly and ensure that the central board can even power everything. If not, you'll need a power supply. Similar problems exist with other powering schemes. You're adding a whole lot of cables that'll have to be managed. Depending on how the internal workings of the waifu are planned (and there are varying schools of thought here), this could get messy pretty quickly. You could opt for using PoE boards, ensuring that you only need one cable per board, but this is more expensive. You'll also need to install a PoE switch, which needs a whole bunch of power and may run hot if it's not cooled properly. Speaking of power, each board will likely have redundant components that will consume power without doing much. I doubt such a setup would consume less power than just having one board managing everything all the time (not even accounting for seperate power supply schemes). If you're planning on having a battery-powered waifu, this is a huge tradeoff to make. Every watt counts. If you're planning on having a tethered waifu (like me), then your main focus will be on placing as many components outside of the waifu as possible. It's better to have a beefier computer managing everything while it sits in its designated corner. If heat is your main concern, the motor controllers and power supply should be your main concerns anyway. All of that will put out far more heat than a laptop board when under significant load. Just think about how many motors you'll need for a waifu to work. Pretty much everyone here is cutting back on articulation in some way or another using various methods, but almost all of us still end up with relatively high figures because we want the most posability out of our waifus as we can manage. Seriously, think about it. Imagine any one physical activity you'd want to do with your waifu. Hugging? That'll need at least four points of articulation. A good hug? Triple that. Handholding? Add another five. Per hand. Don't forget the sensors, either. Even with a demultiplexer, you're looking at some hefty motor requirements unless you want your waifu to only be able to move one point of articulation at a time. Swapping out a single baked potato for a tray of fries won't change the fact that they're sitting in an oven.

>>27359 Yes there are a number of design tradeoffs for a distributed, onboard compute network for our robowaifus. I certainly increases the complexity from a systems-engineering perspective, but perhaps in some contexts the benefits are worth the pain! :D Good thinking, Anon. Thanks for sharing it. Cheers. :^) >>27360 Good, straightforward analysis Anon. >Swapping out a single baked potato for a tray of fries won't change the fact that they're sitting in an oven. Lel'd. :^) >=== -add'l reply -sp edit

>>27360 I like the idea of having all the internal electronic parts including the board fully submerged with water. I guess you might have seen a desktop pc motherboard running in dekstop casing that were filled with water. I forgot how the theory works. I think it has to do with electrolite like how an old car battery would run. Perhaps this technique could help cool down your robowaifu.

>>27359 Think about it slightly different. You want the heat to have a warm body. You have to either distribute electricity or the heat. Distributing the latter might only be possible through liquid, especially since it would have to go for a longer distance. I think it's probably easier to make a coupler in the joints for electricity than for liquid. >>27378 I think they're rather using some plant oil or some oil derivative. I don't think it's such a good idea. We won't need it and it would add difficulties. I think this is mainly useful for overclocking.

>Explore the groundbreaking Coral AI mini PCIe accelerator by Google, a game-changer for home automation and DIY projects, in my latest video where I integrate this innovative chip with the ZimaBoard and Zima Blade for superior performance and cost-effectiveness. Discover how this setup outperforms others, like the Raspberry Pi 5, in speed and thermal efficiency, and follow my journey from troubleshooting software issues to successfully running Frigate, an advanced home lab computer vision system. Learn how this affordable, under $100 setup can revolutionize your home tech projects! https://youtu.be/mdOEaNV8NXw

>>27359 It is most efficient to keep everything as centralized as possible. Remember, it is trivial to route power and data throughout a body. >Heat issue If this is a problem, your design has failed, make something better. I've designed and implemented several systems and heat has never been a problem because, I don't allow it to be. >>27360 >Power logistics This is an important factor. Minimize the range of voltages needed, ideally everything requires voltage within the standard discharge curve of your battery. Always strive to reach this ideal state. >USB Not recommended for internal components unless there is a need. For instance, an internal USB can be used for adding special components like SDR, GPS, Video Capture, and other special need components which don't make sense for a base model. >PoE Expensive, makes sense for charging her while transferring data. Great idea for an external connection. Internally, there's no reason not to just use wires. I trust you already knew all of this, just wanted to add clarity for other Anons. >>27378 >Submersion cooling Mineral oil is better for this. Still used for high performance computers. You need to actively filter the oil and replace it yearly. It is completely unnecessary and only ever makes sense as an aesthetic choice. >>27474 >Heating her skin Connecting her heat sinks to her skin is a great idea. Huge surface area to dissipate heat while making her feel more alive. >>27478 Looking forward to stronger accelerators. These are good to offload object recognition and/or word recognition for low compute platforms. Could be useful to slap into an old Thinkpad to give her eyes and ears.

>>27525 All very good points, and that's kinda what I was getting at in regards to his idea of putting small boards everywhere. Not all of them are designed to get their power the same way. I could have made that a bit more clear, though. Since mai waifu's going to be a tethered TENV system, I'm always thinking about how to make that specific kind of waifu work. However, I still maintain that the number-crunching is best kept remote, even (or especially) for a battery powered system. Modern day WiFi is fantastic, so I don't see much reason to have a larger computer using up space and power in your V1 waifu unless you actually plan on having her be especially mobile from the get-go. Just stick a desktop in the corner and let it chug away.

>>27526 >pic >tfw i have the original of that somewhere...

>>27525 >Mineral oil She will be able to piss and drink a fresh mineral oil as a part of its self routine maintenance cycle ^_^. Just wondering though, is there a reason to choose mineral oil instead of any other type of non conductive chemical substance?

>>27552 I don't see why you couldn't go with something else that's both electrically insulative and thermally conductive, but you'll have to be careful that whatever you use doesn't cause rot, corrosion, dissolution, or leave any film or substrate behind on the surfaces it's interacting with. At least, not enough to make any difference over a decade or so. One of these is bound to happen eventually when you have a liquid mechanically interacting with a surface. Also, be sure to watch out for your own health. Don't get anything toxic, even if you think she'll be well sealed (leaks happen...). All that being said, flushing might not even be necessary (at least, not on a regular basis) if you seal everything properly. The reason mineral oil coolers need to refilled is because the oil evaporates over time. Those systems are rarely completely sealed because they're only designed to operate upright. If your waifu is only ever going to be standing or sitting, then this is probably fine, but if you want her to operate in any orientation, then she'll need her cooling system to be perfectly sealed.

>>27555 You can add should use a water cooling line if you want liquid cooling. Mineral oil is simply ideal for immersion liquid cooling. They are entirely different concepts which should be considered as having nothing in common. Aside from a liquid carrying heat. If you'd like her to drink and pee, a soft line water cooling system is recommended. If you'll drink her pee, please incorporate a filter near her urethra.

Your waifu will be on the cloud or locahost and you will be happy

>Explore the groundbreaking Coral AI mini PCIe accelerator by Google, a game-changer for home automation and DIY projects, in my latest video where I integrate this innovative chip with the ZimaBoard and Zima Blade for superior performance and cost-effectiveness. Discover how this setup outperforms others, like the Raspberry Pi 5, in speed and thermal efficiency, and follow my journey from troubleshooting software issues to successfully running Frigate, an advanced home lab computer vision system. Learn how this affordable, under $100 setup can revolutionize your home tech projects! https://youtu.be/mdOEaNV8NXw >LattePanda's Mu is the latest entrant in the 'Pi Killer' battle, but it has a trick up its sleeve. https://youtu.be/GKGtRrElu30 SBC comparissions: https://github.com/geerlingguy/sbc-reviews

Well looking around at RISV micro-controllers and other MC and to my surprise they are now blowing out MC at 6 cents, 10 cents, etc. with 32 bit RISC and also Arm Cortex. I'm stupefied. I said I couldn't imagine how you could get touch, but if your MC has 10 or 16 touch inputs and cost 15 cents or less each. Well there it is. You have to get some conductive Mylar or whatever on both sides of something you could squeeze together or some sort of conductive material that when squeezed changes value. Not that this is easy but once you have it figured out you could mass produce it. I admit to being amazed. Here's write up on the, "...Arm Cortex-M0+ made by Puya, it’s less than 10 cents in moderate quantities..." https://jaycarlson.net/2023/02/04/the-cheapest-flash-microcontroller-you-can-buy-is-actually-an-arm-cortex-m0/ Maybe a good full handful of these would be all you needed for motion , actuator control(you still need MOSFETS for power) sensing and touch/pressure sensing. I found a link with their "high end" PY32F030F18P6TU 8KB 64KB FLASH TSSOP-20 Microcontrollers (MCU/MPU/SOC) ROHS 6 for $ 1.81 WOW!!!!!!!!!! 60+ for $ 13.76 https://www.lcsc.com/product-detail/Microcontroller-Units-MCUs-MPUs-SOCs_PUYA-PY32F030F18P6TU_C3018715.html If you had 120 or maybe 200 of these at around $30-$40 you have all the control you need. They also have this new RISC-V mentioned on the link above for low cost like less than 20 cents. Here's a few links talking about this(BTW the above part seems to have great support compared to some of the other Chinese offerings the article mentioned) https://hackaday.com/2024/01/04/ch32-risc-v-mcus-get-official-arduino-support/ https://www.cnx-software.com/2023/02/09/8-cents-for-an-arm-cortex-m0-microcontroller-meet-puya-py32-series-mcus/ https://hackaday.com/2023/03/02/a-ch32v003-toolchain-if-you-can-get-one-to-try-it-on/ https://hackaday.com/2023/04/19/risc-v-supercluster-for-very-low-cost/ I think the key to this is some sort of framework using vectors sent by the main CPU to these micro-controllers and letting them do all the compute and feedback sensing to preform the move.

>32872 >> We may devise some custom encoders if we can't find suitable ones for things like the knuckles, eyes, etc. I wish to repeat myself because, to me, it seems the absolute master genius, super-fi, bestest in the world idea for encoders for anything that moves.It's the guys idea on hackaday to send light through fiber optics, plastic fibers are super cheap, and then have the light vary due to movement. This would in turn be fed through another fiber optic back to a simple mass produced camera chip. So you could millions of inputs on one micro-controller with a camera. Called "Fibergrid" https://hackaday.io/project/167317-fibergrid#menu-details and some, rough,speculation about usage by me, >>24233

Another good link by the same guy with a run down on under $1 USD micro-controllers. https://jaycarlson.net/microcontrollers/

And even mre. A comparisom of linux based sngle board computers. This guy really good and the links I got from h im are, in my opinion, a must have for any sort of prototyping we need to do. https://jaycarlson.net/embedded-linux/

>>32880 >This guy really good and the links I got from h im are, in my opinion, a must have for any sort of prototyping we need to do. Yeah, its really good stuff Grommet. Thanks! He mentions the PY32 series [1] (including one for US10cents!). I really like that he has a set of C libraries for that platform already assembled together [2]. Cheers. :^) --- 1. https://jaycarlson.net/2023/02/04/the-cheapest-flash-microcontroller-you-can-buy-is-actually-an-arm-cortex-m0/ 2. https://github.com/jaydcarlson/py32-template

>>32874 I'll take another look at this. What we really need is an encoder/sensor that simply in the form-factor of being a fiber itself. That way we can simply run it up through the knuckles to obtain position information with no add'l muss & fuss (such as with rotary encoders, etc.) Know anything about such a form-factor sensor, Anon?

>>32874 Thanks for all the interesting comments, but especially for that. The idea to use light for robowaifus crossed my mind years ago, but this was just fantasizing, and then you posted this at some point already. This is really something we should keep in mind for sensors in the skin. Though, it would need to get quite small and maybe use the light spectrum. Idk, I guess there are reasons why this doesn't already exist. That said, I recently thought about if people could build their own optics based AI accelerator (GPU-like). Not for fitting into the PC but just some rather big drawer if necessary, but then good enough to run Llama 3. Just in case if the "Pausing AI"-crazies would be somewhat successful or just to discourage politicians in the first place. I thought about how to do that, and wondered how long it would take to calculate the color coming from a fibre since it would influence the frequency of change that could be allowed for transmitting information. Unfortunately, the camera he uses is rather big and he seems to only use the amount of light without analyzing the spectrum, which tells me this is the hard part. Which is was I thought. I wanted to use 50k pixels (220x220) and analyze the spectrum of each, haha. I don't really plan to pursue that, but maker projects like this one would be a place to look into it. So good to know it exists. Companies working on optics based computing would try to go very small. I think this approach even only helps if the hardware is farther away from each other, which is why this isn't being done. The distance would probably be good for cooling. The project I thought about would be more about having something we can build and repair. Bigger parts than regular chips, but still small enough to fit into a room and run one big LLM at any time.

>>32882 Why does it need to "run" through the knuckels? Can't it pick up light based on the position of a connection?

>>32893 Simple. Ease-of-manufacture. There's also the very-likely adjunct benefit of lowered-cost too (less parts, less mass, etc.) >pick up light based on the position of a connection Maybe you have something there, Anon. Can you explicitly & particularly define your idea of how 'position of a connection' might work in your concept? >=== -minor edit

>>32896 Related about buses, sensors and multiplexing in the engineering general thread: >>32900

Interesting chip, Sohu AI chip "...Most large language models (LLMs) use matrix multiplication for the majority of their compute tasks and Etched estimated that Nvidia’s H100 GPUs only use 3.3% percent of their transistors for this key task. This means that the remaining 96.7% silicon is used for other tasks, which are still essential for general-purpose AI chips..." They're showing lama70B throughput at high speeds. lama70b should be really powerful. https://www.tomshardware.com/tech-industry/artificial-intelligence/sohu-ai-chip-claimed-to-run-models-20x-faster-and-cheaper-than-nvidia-h100-gpus

>>32882 This fibergrid thing this guy thought up is in my opinion noble prize winning material. It's an idea that has been one of the most impressive out of the box thinking I've seen in a long time. I mean it makes it sooooo easy to monitor literally hundreds of thousands of sensors with one chip, a LED, fiber and a camera. There's all sorts of ways to run this. For joints I mentioned grays code but that's actually a little overboard, but good for "exact" placement. We don't need that precision. Something as simple as a slit that opens and closes as the joint rotates. One fiber feeds light and the return picks up the intensity based on how far the slit opens. Another, a piece of paper with coloring going from white to black in shade. Shine on it, then return the reflection. So may ways. Try as I might I have not come up with what I feel is a super satisfactory way to use this for pressure sensing. I have posted a few but, I admit they are a bit kludgy. I am still thinking about it and if anyone has any ideas about how to vary light with pressure please chime in. I just remebered one idea I had that just might work. Slit the fiber all the way down it's length. If pressed it would open the fiber and leak light. Have one set going down a limb or body part and many horizontal. Might work but...difficult to say. The beauty of this is you can use a camera to record the reflection and with a vertical and horizontal grid you get position. And you could do high detail, full body with a $9 camera, micro-controller combination and a few dollars of plastic fiber.

>>32882 >we can simply run it up through the knuckles to obtain position information with no add'l muss & fuss Right after I wrote the last comment I read yours again and maybe, just maybe, a split fiber would do this. One way, the simple way, to do this is to run the fiber down through the joint with a split at the joint then run it back up to the camera. What would really be slick is to put a reflector on the end of the fiber and have a beam splitter right at the light, camera interface. But...that adds so much complexity..(.just thinking out loud). "If" you could find a cheap, fast, way to do this, it would be very cool. As for the split fiber. Could be done with a template to split a certain depth and length in a jig. You would need to glue on a piece of black something, paper, plastic, cloth whatever to absorb light. You "might" have to have a sort of wedge shape above the fiber to force it open. haven't tried this. I know they have force sensing systems for fibers and I looked at it, a lot, but they were super complicated things. One thing good about this is the calibration could be automated. Bend the limb and read the output as it bends. Same with touch. Touch it all over while recording the light variance per force. So it wouldn't have to be perfect to work. I would suggest using multiple camera pixels inputs for each feedback light source so even if a few go bad you still can re-calibrate it without chunking the whole thing.

>>32918 > fibergrid thing... noble prize winning material. It's an idea that has been one of the most impressive out of the box thinking ... I don't understand this argument, what did I miss? Combining light and cameras, or optical sensors, is very obvious. If you just know data transfer through optical fibers exists, then how would you imagine that working? If you know about the concept or idea of a computer based on it, how else would it work? It's pixels or a matrix, intensity, the spectrum of light (color), and frequency of change on top of it. It's something very obvious.

>>32891 >I wanted to use 50k pixels (220x220) and analyze the spectrum of each, Shine it on a grating. The color will separate and you can measure not only the different spectrum's but the intensity. A cheap Fresnel lens would work. A simple dirt cheap way would be to shine the light on a regular music CD or optical dvd. Might have to move the CD around a bit to get the right effect you wanted. To be precise you are measuring the reflection off of a grated mirrored surface or ion the case of a Fresnel lens through it. You can buy Fresnel lenses that you stick in the back window of campers so they can see behind them. I'm not sure if that would work, but it might. I "know" a grating could work. You could make these with a razor blade and a piece of plastic and a little tricky scraping. Drawing Holograms by Hand http://amasci.com/amateur/hand1.html I wonder, I do not think pressure on a fiber will cause it to change light frequency. You know better or have heard of such? Please say so if you have. What would you use to change the light frequency?

>>32891 >wondered how long it would take to calculate the color coming from a fibre Instantly, or as fast as your light sensor works.

>>32920 >I don't understand this argument, what did I miss? Combining light and cameras, or optical sensors, is very obvious. It's brilliant. Using light and sensors has been around for a long time yes but using the individual pixels of a CCD camera and its preexisting software to multiply the amount of sensors to the thousands with one small chip by reading the reflective feedback from a variable light source, is...brilliant. It's so good. And if it's so obvious how come no one else thought of it? But he did. He has another crazy idea for robots. He takes an oscillating arm, or something, and feeds it through the whole entire robot and then uses some sort of clutches to grab the, whatever it is that moves, and that moves the muscles. It's the craziest most whacked thing ever but it's also...brilliant. He's built prototypes of these. And in fact if you tweak it a little... Say use a AC waveform, and then use magnetic amplifiers all throughout the body. The magnetic amplifer blocks the AC until you use a smaller signal to unblock it...well then...once unblocked it could be fed to a rectifier and provide DC to power a motor, solenoid, whatever. So you have one big ass AC flowing at all times. Only siphoning off the power to those muscles unblocked. And you wonder why I keep bleeting about magnetic amplifiers. That's why.

>>32921 > Fresnel lense Interesting. I recall these lenses vaguely from astronomy, for analyzing stars. Your ideas isn't bad, I was thinking more of a gradual filter while this one seems to do a hard cut-off, but I haven't looked into it. I thought more about some glass sphere or cylinder changing what spectrum it let's through. Maybe changing the color based on some flow of electricity, or something more indirect like vibrations. But I also don't want too many parts which could break. Anyways, this might actually be good enough, I mean better than nothing, though it would also need to be very small and this is where it probably would get tricky. That said, let's not forget this is more of a spontaneous thought experiment. >>32917 >Sohu AI chip Yeah, I came across this. But we can't buy it yet. I would like this for at home, or better as an SBC. Anyways, one of the comments from the liked news: >In theory, Intel seems to have shown that their Intel MESO (Magnetoelectric spin-orbit) logic, a post-CMOS technology related to spintronics, is much more amenable to both low-power logic and low-power neuromorphic computing. >For example, emerging Non-Volatile-Memory (NVM) / Persistent Memory MRAM is already based on spintronics phenomena >Therefore much more R&D ressources should be allocated to develop new manufactuting tools to improve and lower MRAM manufacturing cost, and then improve those tools to evolve to MESO manufacturing : this would be much, much more groundbreaking !!! https://www.techspot.com/news/77688-intel-envisions-meso-logic-devices-superseding-cmos-tech.html https://www.imec-int.com/en/press/imecs-extremely-scaled-sot-mram-devices-show-record-low-switching-energy-and-virtually >fibergrid sensors Btw, can we move the discussion of sensors out of the thread for computers which was even meant just for SBCs. Buses, sensors and multiplexing in the engineering general thread: >>32900 I fell for it myself again with the posting of >>32920 - Keeping the speculations on optical computing in this thread might be okayish, since it already moved beyond SBCs and small accelators to GPUs such.

Since we're doing hardware today: >A research paper from UC Santa Cruz and accompanying writeup discussing how AI researchers found a way to run modern, billion-parameter-scale LLMs on just 13 watts of power. That's about the same as a 100W-equivalent LED bulb, but more importantly, it's about 50 times more efficient than the 700W of power that's needed by data center GPUs like the Nvidia H100 and H200, never mind the upcoming Blackwell B200 that can use up to 1200W per GPU. >The work was done using custom FPGA hardware, but the researchers clarify that (most) of their efficiency gains can be applied through open-source software and tweaking of existing setups. Most of the gains come from the removal of matrix multiplication (MatMul) from the LLM training and inference processes. >How was MatMul removed from a neural network while maintaining the same performance and accuracy? The researchers combined two methods. First, they converted the numeric system to a "ternary" system using -1, 0, and 1. This makes computation possible with summing rather than multiplying numbers. They then introduced time-based computation to the equation, giving the network an effective "memory" to allow it to perform even faster with fewer operations being run. >The mainstream model that the researchers used as a reference point is Meta's LLaMA LLM. The endeavor was inspired by a Microsoft paper on using ternary numbers in neural networks, though Microsoft did not go as far as removing matrix multiplication or open-sourcing their model like the UC Santa Cruz researchers did. >It boils down to an optimization problem. Rui-Jie Zhu, one of the graduate students working on the paper, says, "We replaced the expensive operation with cheaper operations." Whether the approach can be universally applied to AI and LLM solutions remains to be seen, but if viable it has the potential to radically alter the AI landscape. >We've witnessed a seemingly insatiable desire for power from leading AI companies over the past year. This research suggests that much of this has been a race to be first while using inefficient processing methods. We've heard comments from reputable figures like Arm's CEO warning that AI power demands continuing to increase at current rates would consume one-third of the United States' power by 2030. Cutting power use down to 1/50 of the current amount would represent a massive improvement. >Here's hoping Meta, OpenAI, Google, Nvidia, and all the other major players will find ways to leverage this open-source breakthrough. Faster and far more efficient processing of AI workloads would bring us closer to human brain level functioning—a brain gets by with approximately 0.3 kWh of power per day by some estimates, or 1/56 of what an Nvidia H100 requires. Of course, many LLMs require tens of thousands of such GPUs and months of training; hence, our gray matter isn't exactly outdated yet. It seems to be mostly coming from some software, but the FPGA will likely still be necessary. Not sure if this should rather go into >>250, but it's very much related to the hardware we would need. - https://www.tomshardware.com/tech-industry/artificial-intelligence/ai-researchers-found-a-way-to-run-llms-at-a-lightbulb-esque-13-watts-with-no-loss-in-performance - https://arxiv.org/abs/2406.02528 - https://news.ucsc.edu/2024/06/matmul-free-llm.html - https://github.com/ridgerchu/matmulfreellm -

This is some great information content recently ITT, Anons. Thanks! >=== -parse-edit post into R&D thread; per : ( >>32925 )

>>32927 Really good links. Very important stuff.

I talked about an idea for pressure touch senors here(in the wrong thread) >>32918 and here >>32919 and much earlier >>24233 >>30980

I did not mean to post the above here. Sigh, but it does reference the previous post so...maybe keep it??

>>32957 >>32960 Eheh, it's fine Grommet. Since this is a) great stuff, and b) you've helped everyone by 'consolidating' the access to your related-posts, and c) it's almost impossible with this forum format to stay on-topic (w/o rather Draconian moderation -- no one wants that) * ... then I say let's keep it! :^) --- * However I do think that (especially given the highly-technical & expansive nature of what we're attempting here) we do a better job at this than most any other IB I've ever seen. Give yourselves a pat on the back, Anons! :D >=== -add footnote

An idea for for waifu computers. I expect the sweet spot price wise is standard PC mini ATX motherboards. Another good choice would be the Mini-ITX is a 170 mm × 170 mm (6.7 in × 6.7 in) motherboard. https://en.wikipedia.org/wiki/Mini-ITX These usually are fan less but most would be lacking in computing power. Micro ATX 244x244mm (9.6×9.6 inches) likely would be the best for performance while staying manageable in size. You would have to have a full size waifu to fit it though. The problem is waifus are, or need to be, able to go in water to wash themselves and they are going to get wet. So I was thinking a good way to to waterproof them would be to use standard food packaging material. Vacuum packing. On the other side of this you could make a plastic box enclosing all the electronics and run water over the whole board through this box for cooling. The box would make a firm mount for the motherboard and contain the water cooling. The plastic film would stay cool and keep the water off the electronics while providing excellent cooling. This could be circulated through the body to simulate body heat. If it would not be too expensive you could use optical outputs to carry signals in and out of the board.

>>33785 > On the other side of this you could make a plastic box enclosing all the electronics and run water over the whole board through this box for cooling. The box would make a firm mount for the motherboard and contain the water cooling. The plastic film would stay cool and keep the water off the electronics while providing excellent cooling. Yep. I call this little jewel the robowaifu's Breadbox, and if push-comes-to-shove, thermal-engineering-wise, then we'll actually need to duct cooling tubes into the thing to keep the electronics cool. I expect just regular distilled water should be sufficient for this, thermal-density-wise. Good thinking, Anon!

>>33785 >Mini ITX The only mobo size that actually makes sense for a waifu. Placement in her frame would be tricky, I'd suggest tilting it in the chest. >Lacking in compute I'm unsure of what you mean. ITX boards can run almost any desktop processor and GPU. As for what I'd actually do with an ITX PC, I'd just stick it in her movement cart. That's a battery behind the ITX board.

>>33809 >Maidcom Mini's gettin' some roller skates! Based. :^)

>>33809 >>Lacking in compute >I'm unsure of what you mean Fanless as a general rule means, less power. And yes there are exceptions but "generally". For us, likely water cooled so no problem. I did not think about that at the time and was talking...generally. >=== -fmt patch