Question 1

Can the NVIDIA H100 80GB PCIe run DEEPSEEK-R1 distill 8B?

Accepted Answer

DEEPSEEK-R1 distill 8B needs about 5.4GB working set (weights 4.9GB + KV cache 0.5GB at 4k context). The NVIDIA H100 80GB PCIe offers ~80GB usable — verdict: full. 5.4 GB needed of 80 GB usable — headroom for context

Question 2

How fast is DEEPSEEK-R1 distill 8B on the NVIDIA H100 80GB PCIe?

Accepted Answer

~190–320 tok/s on the NVIDIA H100 80GB PCIe. computed roofline: 2039 GB/s × 0.46–0.76 efficiency window / 4.9GB (Q4_K_M) — bands, never points.

Question 3

What should I run DEEPSEEK-R1 distill 8B on instead?

Accepted Answer

Tesla M60 (16GB, used) runs it fully ($35).

Can the NVIDIA H100 80GB PCIe run DEEPSEEK-R1 distill 8B?

Run DEEPSEEK-R1 distill 8B on instead