pkroll

And it's literally the first day. Someone will make a distilled 4-8 step LoRA and we're off to the races.

72.06 t/s. That's the full Qwen 3.6 27B model BF16, using MTP, running on Ollama. Yes I know I should bite the bullet and get vllm running on that box.

That was, also, at a 570 watt limit: I normally run a little less, but when I first tried this I actually forgot I had set the limit to 300 (it's a hot day, I figured why fight the A/C?), and at 300 watts the same question came back at 69.38 t/s. (The extra power matters more for compute bound things, the difference in generating LTX2.3 videos is considerably higher... but still not linear.)

I'd bet money when TVs are advertised at 120 FPS, they're really 119.88 FPS, so no judder showing 23.976 FPS and the other NTSC-off display rates.

I get 136 results, from that query.

But really all the various video models really want an 80+ gig vram card, to run comfortably. The contortions the ComfyUI community goes through to get things running at a reasonable speed on the current, dinky-sized vram consumer cards, are impressive.

Slightly frustrating. But good to know.

"Our study suggests that LLMs have achieved superhuman performance on general medical diagnostic and management reasoning"

128G of unified memory. $3K. Throw ollama and ComfyUI on that sucker and things could get interesting. The question is how much slower than a 5090, is this gonna be? The memory bandwidth isn't going to match a 512 bit bus.

None of which takes away from it being terrible that we've lost Terri Garr.

Well. Not GIVE. It's $20K.

But are you asking in good faith? Or are you asking because you want to get to a point where you use a service that costs money to maintain and yet you don't spend money or view ads but you still feel ethical?