3d • 💬 General discussion
qwen3.6 on k80
I'm an independent AI researcher. I got qwen3.6-35B-A3B running at 21 tok/s decode on Tesla K80s. Those cards are $60 on eBay and I can run 8 on one crypto mining motherboard.
Nobody writes software for old hardware. So I did. Kernel driver, inference stack in C, multi-GPU orchestration. It works.
All of it self-funded so far. That's the wall I'm at. I need capital, not ideas.
Looking for funding. Angel, grant, contract, whatever shape it takes. If I get it I hire C devs and go faster.
Reach out if that's you or you know someone.
2
0 comments
qwen3.6 on k80
skool.com/cliefnotes
What we give away free beats most paid courses. Build durable AI systems with a Marine vet and Edinburgh researcher. 40+ lessons, growing.
Leaderboard (30-day)
1
+1069
2
+777
3
+612
4
+570
5
+477
Powered by