I bought one when they were pretty new and I had issues with rocm (iirc I was getting kernel oopses due to GPU OOMs) when running LLMs. It worked mostly fine with ComfyUI unless I tried to do especially esoteric stuff. From what I've heard lately though, it should work just fine.
"Like the Pi 4, I think this system is the first RISC-V desktop environment that isn't painful to use, just inconvenient. Actions still have delays, but the delays are more reasonable, and don't make me constantly question if the computer's frozen."
also some really odd choices by Eswin for the eic7702x, which is essentially 2 p550 chips glued together.
If you're a bit handy, you can assemble a line filter using a part like this https://enerdoor.com/products/fin27/ for a heck of a lot cheaper than you can buy a filtered power strip.
Even if there's very little audio-frequency attenuation, it's possible for higher frequencies to produce audio-frequency intermodulation distortion, and filtering could reduce this. This is one reason "high definition" (ultrasound sampling rate) audio is a bad idea as a listening format.
In 2013 I bought out 2 Radio Shacks worth of ferrite beads when I was hunting down signal noise in my senior design project (CNC mill rebuild and update.)
All else fails, add more beads.
Also, I learned that you can make your own shielded flat cables with aluminum duct tape.
Who knew that they had a really good reason for using 48V signaling in the original machine controls from 1986?
Maybe you’re right. My experience is with radios, where it’s possible that high frequency noise is conducted into the RF section rather than into the audio amplifier. I know that in one case, both my transmitted signal and received audio output were absolute garbage (edit: because it was picking up noise from the vehicle ignition) until I added a choke to the power input wiring.
I did that for years and I recommend it as well. Pure linux desktop + console for games is a nice combo and a good separation of functions.
Of course... at this point I am back to having a PC with a beastly GPU and I boot Windows for games and CAD. It is hard to resist high framerate 4k gaming once it becomes a possibility, so now I need to figure out the secure boot problem for the occasional game that requires it.
You can do it slowly with ik_llama.cpp, lots of RAM, and one good GPU. Also regular llama.cpp, but the ik fork has some enhancements that make this sort of thing more tolerable.
That is literally the thing the parent poster wants to avoid by running open models.
[edit] I was a little unfair -- lack of access to training data is a bit of an issue (perhaps moreso for analysis than for for actual use, considering what it takes to train these models). I'm thankful that some of them are also distributed as base models, which should be relatively unbiased compared to what happens later during finetuning.
You'd have to be doing something where the unified memory is specifically necessary, and it's okay that it's slow. If all you want is to run large LLMs slowly, you can do that with split CPU/GPU inference using a normal desktop and a 3090, with the added benefit that a smaller model that fits in the 3090 is going to be blazing fast compared to the same model on the spark.
reply