← AI Hashrate 中文

MI300X 192GB vs V100 16GB PCIe for Local LLMs

The MI300X 192GB has about 489% more memory bandwidth (5,300 vs 900 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 489% faster on the MI300X 192GB. The MI300X 192GB fits more of the catalog at Q4/8K — 61 vs 25 of 73 models (192 vs 16 GB VRAM). Largest model that fits the MI300X 192GB but not the V100 16GB PCIe: MiMo-V2.5 (310.0B). MSRP is $15,000 for the MI300X 192GB vs $8,000 for the V100 16GB PCIe. MSRP is a launch list price, not live retail — Amazon prices move; click through for the current price. Not retail-shoppable: MI300X 192GB, V100 16GB PCIe (datacenter-class; cloud rental is the realistic way to use it). In the Q4/8K head-to-head below (49 shared models), the MI300X 192GB posts the higher tok/s in 49 rows vs 0. Methodology.

Spec comparison

MI300X 192GBV100 16GB PCIe
VendorAMDNVIDIA
VRAM192 GB16 GB
Memory typeHBM3HBM2
Memory bandwidth5,300 GB/s900 GB/s
TDP750 W250 W
MSRP (list)$15,000$8,000
Buy——

MSRP is the launch list price, not live retail. Buy links are Amazon affiliate links — the price after you click is the live price.

Q4 / 8K head-to-head — top 15 of 49 shared models

ModelParams (B)MI300X 192GB tok/sMI300X 192GB fitsV100 16GB PCIe tok/sV100 16GB PCIe fitsWinner
MiniCPM5-1B1.085000 est.Yes895.6 est.YesMI300X 192GB
Qwen3.5-2B2.02720.3 est.Yes483.6 est.YesMI300X 192GB
DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx2.32365.5 est.Yes420.5 est.YesMI300X 192GB
Qwen3.6-35B-A3B36.01813.6 est.Yes316.9 est.NoMI300X 192GB
Qwen3.5-35B-A3B34.71813.6 est.Yes339.8 est.NoMI300X 192GB
GLM-4.7-Flash31.21813.6 est.Yes313.0 est.NoMI300X 192GB
Qwen3-30B-A3B-Instruct-250730.51813.6 est.Yes405.4 est.NoMI300X 192GB
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF1630.01813.6 est.Yes434.0 est.NoMI300X 192GB
Qwen3-Coder-Next79.71813.6 est.Yes67.0 est.NoMI300X 192GB
Ornith-1.0-35B35.01813.6 est.Yes334.2 est.NoMI300X 192GB
Qwen3-Next-80B-A3B-Instruct81.31813.6 est.Yes64.5 est.NoMI300X 192GB
qwen35b-a3b-fable-sft-abliterated34.71813.6 est.Yes325.2 est.NoMI300X 192GB
Llama-3.2-3B-Instruct3.21700.2 est.Yes302.3 est.YesMI300X 192GB
gpt-oss-20b21.51511.3 est.Yes268.7 est.YesMI300X 192GB
gemma-4-26B-A4B-it25.21431.8 est.Yes254.5 est.YesMI300X 192GB

est. = bandwidth-based estimate; measured = curated real benchmark that overrides the estimate. Fits = weights + KV(8K) + 1 GB overhead ≤ 95% of VRAM.

Fit coverage at Q4 / 8K

MI300X 192GBV100 16GB PCIe
Catalog models that fit61 of 7325 of 73
Largest exclusive fitMiMo-V2.5 (310.0B)—

MI300X 192GB details → · V100 16GB PCIe details →