Rendered at 16:32:22 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
ladyanita22 10 hours ago [-]
This is something I've been fantasizing about for long.
Let's say we took Rust, a language that makes parallelization easier than others (as it helps you avoid some common footguns). How difficult would it be to have a massively parallel computer system made out of many tiny, simple microcontroller-like chips? Let's say we picked many little Risc-V's. Surely this would be an interesting experiment (though I'm not sure whether it'd make economic sense or not...)
sigmoid10 10 hours ago [-]
It would certainly not make any economic sense, and I guess that's also why noone is seriously looking into stuff like volunteer/enthusiast clusters of home computers to do inference in the same way that e.g. LHC@home works. The main bottleneck for LLMs is still memory bandwidth. Any memory bus not directly soldered on your GPU is terribly slow. That's why one big GPU with twice the VRAM will always perform significantly better than two GPUs with half the VRAM each. And it's also not like you can just solder more memory onto a chip. At modern speeds, the speed of light is a hard limit. For current GDDR7, signals may only travel like 10mm per cycle.
If you spread such a system out over dozens or hundreds of tiny chips, you'll be wasting most of its resources and lose hard to anyone who built a single chip setup.
tyingq 3 hours ago [-]
Not exactly what you're describing, and not shipping yet, but a cluster in a box. 8 cores/node, 8 nodes.
Look at Xmos, founded by a transputers-dad. Small uCs, that can be connected together into a massive cluster, while being first-citizen of their xC language.
godojo 3 hours ago [-]
Man I miss Slashdot's Beowolf culsters
akavel 7 hours ago [-]
See GreenArrays' 144-core Forth chips by Chuck Moore.
ur-whale 8 hours ago [-]
> How difficult would it be to have a massively parallel computer system made out of many tiny, simple microcontroller-like chips?
It's scaling the communication that becomes hard.
In this project they daisy-chain SPI. I don't believe that would scale very far.
tdhz77 14 hours ago [-]
Soon ai in every lightbulb running Kubernetes
_joel 5 hours ago [-]
How many rollingupdate pods does it take to change a lightbulb?
jagged-chisel 4 hours ago [-]
Eventually.
oneZergArmy 10 hours ago [-]
Praise the Omnissiah.
Tade0 10 hours ago [-]
With the proliferation of Abominable Intelligence? Quite the contrary!
tombert 12 hours ago [-]
You know, I've always liked Futurama but I always kind of thought it was silly that literally everything has an AI and a personality.
But, you know, I actually think that there might be a logic to it. Economies of scale might mean that almost-literally every computer you buy in the year 3000 has some kind of AI-assistance chip in there, and sure maybe it will have full AI with a personality spitting out one-liners.
abroadwin 11 hours ago [-]
Kind of like how disposable vape pens often have a 24 MHz Cortex-M0+ with 3 kB SRAM and 24 kB flash, which would have seemed ludicrous a while back.
KeplerBoy 10 hours ago [-]
Change the year 3000 to the 2030s and it might be just as accurate.
cameron_b 16 hours ago [-]
It is a bit of a bummer to see that the degree of 'compression' makes it a fancy llm noise-maker. It is still charming.
NDlurker 13 hours ago [-]
I'm curious how this would handle grammar checking on a basic word processor. Or maybe generate worlds for small text based games. I have no idea what the capabilities are of a cluster like this.
librasteve 10 hours ago [-]
haha … this is precisely the kind of project that https://bil-lang.org is aimed at: Go for parallel (ie in this case pipeline processing).
don’t get too excited until we get the TinyGo backend built though ;-)
matthewfcarlson 13 hours ago [-]
I’m actually working on a small project that’s exactly this! Less quant so it’s only 150M parameters but this is amazing.
AmazingTurtle 4 hours ago [-]
"Your scientists were so preoccupied with whether they could, they didn't stop to think if they should"
Seriously though, what are the low cost chips that can usefully run LLMs? Is a Mac Mini the lowest we can go? Are there iGPUs on mini-itx that can do it, or are there dedicated AI chips that one could turn into a pi HAT?
chorylee 3 hours ago [-]
An Orange Pi 5 Max does this job for real. It's an RK3588 board — $75 for 4GB, $95 for 8GB on AliExpress. A community test got Qwen2.5-0.5B at about 12 tok/s on the CPU via llama.cpp. The chip also has a 6 INT8 TOPS NPU if you'd rather go the RKNN route. Won't beat a Mac Mini, but it's an actual computer for under a hundred bucks.
cameron_b 1 hours ago [-]
I regret to inform you that prices have long departed the lower atmosphere. Even on AliExpress, you're looking at a few hundred bucks for those. Still less than a Mac Mini, but less less.
Risse 8 hours ago [-]
Depends on what you consider to "usefully run LLMs".
Earlier this year, I bought a mini pc from Aliexpress, specs are roughly Ryzen H255, 24GB LPDDR5, 1TB SSD. This was around 350€ including VAT, customs, shipping etc. I would personally consider this somewhat of a lowest class of useful LLM box. It can run 8B models well, up to somewhere around 24B. I currently run Gemma 4 26B A4B Q5 on it, with MTP, and it is quite slow, but smaller models would run okay on it.
bahmboo 5 hours ago [-]
Let's teleport back 10 years and what you have is a magic box that could make you billions.
aidiscoverywire 11 hours ago [-]
[flagged]
nonasking_ 12 hours ago [-]
Thanks for sharing. It's fascinating to see a 0.5B LLM being split across seven ESP32s like this.
Let's say we took Rust, a language that makes parallelization easier than others (as it helps you avoid some common footguns). How difficult would it be to have a massively parallel computer system made out of many tiny, simple microcontroller-like chips? Let's say we picked many little Risc-V's. Surely this would be an interesting experiment (though I'm not sure whether it'd make economic sense or not...)
If you spread such a system out over dozens or hundreds of tiny chips, you'll be wasting most of its resources and lose hard to anyone who built a single chip setup.
https://milkv.io/cluster-08
It's scaling the communication that becomes hard.
In this project they daisy-chain SPI. I don't believe that would scale very far.
But, you know, I actually think that there might be a logic to it. Economies of scale might mean that almost-literally every computer you buy in the year 3000 has some kind of AI-assistance chip in there, and sure maybe it will have full AI with a personality spitting out one-liners.
don’t get too excited until we get the TinyGo backend built though ;-)
Earlier this year, I bought a mini pc from Aliexpress, specs are roughly Ryzen H255, 24GB LPDDR5, 1TB SSD. This was around 350€ including VAT, customs, shipping etc. I would personally consider this somewhat of a lowest class of useful LLM box. It can run 8B models well, up to somewhere around 24B. I currently run Gemma 4 26B A4B Q5 on it, with MTP, and it is quite slow, but smaller models would run okay on it.