← AI Hashrate 中文

RTX 4090D 24GB (China) vs V100 32GB SXM2 for Local LLMs

The RTX 4090D 24GB (China) has about 12% more memory bandwidth (1,008 vs 900 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 12% faster on the RTX 4090D 24GB (China). The V100 32GB SXM2 fits more of the catalog at Q4/8K — 43 vs 41 of 73 models (32 vs 24 GB VRAM). Largest model that fits the V100 32GB SXM2 but not the RTX 4090D 24GB (China): Qwen3.6-35B-A3B-FP8 (36.0B). MSRP is $1,599 for the RTX 4090D 24GB (China) vs $10,000 for the V100 32GB SXM2. MSRP is a launch list price, not live retail — Amazon prices move; click through for the current price. Not retail-shoppable: RTX 4090D 24GB (China), V100 32GB SXM2 (datacenter-class; cloud rental is the realistic way to use it). In the Q4/8K head-to-head below (56 shared models), the RTX 4090D 24GB (China) posts the higher tok/s in 43 rows vs 13. Methodology.

Spec comparison

RTX 4090D 24GB (China)V100 32GB SXM2
VendorNVIDIANVIDIA
VRAM24 GB32 GB
Memory typeGDDR6XHBM2
Memory bandwidth1,008 GB/s900 GB/s
TDP425 W300 W
MSRP (list)$1,599$10,000
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 56 shared models

ModelParams (B)RTX 4090D 24GB (China) tok/sRTX 4090D 24GB (China) fitsV100 32GB SXM2 tok/sV100 32GB SXM2 fitsWinner
MiniCPM5-1B1.081003.1 est.Yes895.6 est.YesRTX 4090D 24GB (China)
Qwen3.5-2B2.0541.7 est.Yes483.6 est.YesRTX 4090D 24GB (China)
DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx2.3471.0 est.Yes420.5 est.YesRTX 4090D 24GB (China)
Qwen3.6-35B-A3B36.0361.1 est.Yes322.4 est.YesRTX 4090D 24GB (China)
Qwen3.5-35B-A3B34.7361.1 est.Yes322.4 est.YesRTX 4090D 24GB (China)
GLM-4.7-Flash31.2361.1 est.Yes322.4 est.YesRTX 4090D 24GB (China)
Qwen3-30B-A3B-Instruct-250730.5361.1 est.Yes322.4 est.YesRTX 4090D 24GB (China)
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF1630.0361.1 est.Yes322.4 est.YesRTX 4090D 24GB (China)
Ornith-1.0-35B35.0361.1 est.Yes322.4 est.YesRTX 4090D 24GB (China)
qwen35b-a3b-fable-sft-abliterated34.7361.1 est.Yes322.4 est.YesRTX 4090D 24GB (China)
Llama-3.2-3B-Instruct3.2338.5 est.Yes302.3 est.YesRTX 4090D 24GB (China)
gpt-oss-20b21.5300.9 est.Yes268.7 est.YesRTX 4090D 24GB (China)
gemma-4-26B-A4B-it25.2285.1 est.Yes254.5 est.YesRTX 4090D 24GB (China)
Qwen3-4B-Instruct-25074.0270.8 est.Yes241.8 est.YesRTX 4090D 24GB (China)
NVIDIA-Nemotron-3-Nano-4B-BF164.0270.8 est.Yes241.8 est.YesRTX 4090D 24GB (China)

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

RTX 4090D 24GB (China)V100 32GB SXM2
Catalog models that fit41 of 7343 of 73
Largest exclusive fit—Qwen3.6-35B-A3B-FP8 (36.0B)

RTX 4090D 24GB (China) details → · V100 32GB SXM2 details →