~/gpu/b200 vs mi300x

B200 manufacturerB200vsMI300X manufacturerMI300X

Head-to-head for local LLM inference. The honest comparison: VRAM, bandwidth, compute, and which of the 30 catalog models actually fit on each.

The specs.

$ diff specs b200 mi300x
Stat
b200
mi300x
Δ
VRAM
192 GB
192 GB
0%
Memory bandwidth
8,000 GB/s
5,300 GB/s
-34%
FP16 compute
2250 TFLOPS
1300 TFLOPS
-42%
Weights budget at 8K ctx
154 GB
154 GB
0%

Model fit difference.

$ models that change with the card
Fits on both
27of 30
Only on b200
0
Only on mi300x
0

// showing 12 of 30 models; differing fits first

Model
b200
mi300x
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
overQ4_K_M
overQ4_K_M
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16
fitsFP16/BF16

Which one wins for…

$ ./recommend --by-workload
More VRAM headroom

Tied at 192 GB. Choose on bandwidth or price.

Faster decode (bandwidth)

B200 by +51%.

Faster prefill (compute)

B200 by +73% TFLOPS.

Catalog models that fit

Tied: 27 of 30 fit on each.

Drill into either card.

$ ./vrambudget --gpu

Discussion.

$ gh discussion list

// sign in with github to leave a comment. threads live in the repo's discussions tab.