← AI Hashrate

RTX 4090D 24GB (China) vs RTX 5080 16GB for Local LLMs

The RTX 4090D 24GB (China) has about 5% more memory bandwidth (1,008 vs 960 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 5% faster on the RTX 4090D 24GB (China). The RTX 4090D 24GB (China) fits more of the catalog at Q4/8K — 33 vs 25 of 61 models (24 vs 16 GB VRAM). Largest model that fits the RTX 4090D 24GB (China) but not the RTX 5080 16GB: GLM-4.7-Flash (31.2B). MSRP is $1,599 for the RTX 4090D 24GB (China) vs $999 for the RTX 5080 16GB. 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) (datacenter-class; cloud rental is the realistic way to use it). In the Q4/8K head-to-head below (44 shared models), the RTX 4090D 24GB (China) posts the higher tok/s in 33 rows vs 11. Methodology.

Spec comparison

RTX 4090D 24GB (China)RTX 5080 16GB
VendorNVIDIANVIDIA
VRAM24 GB16 GB
Memory typeGDDR6XGDDR7
Memory bandwidth1,008 GB/s960 GB/s
TDP425 W360 W
MSRP (list)$1,599$999
Buy—NVIDIA GeForce RTX 5080 Founders Edition (affiliate) · More listings

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 44 shared models

ModelParams (B)RTX 4090D 24GB (China) tok/sRTX 4090D 24GB (China) fitsRTX 5080 16GB tok/sRTX 5080 16GB fitsWinner
Qwen3-0.6B0.8801.8 est.Yes763.6 est.YesRTX 4090D 24GB (China)
gemma-3-1b-it1.0641.5 est.Yes610.9 est.YesRTX 4090D 24GB (China)
Llama-3.2-1B-Instruct1.2534.5 est.Yes509.1 est.YesRTX 4090D 24GB (China)
Qwen3-30B-A3B30.5213.8 est.Yes440.9 est.NoRTX 5080 16GB
Qwen3-Coder-30B-A3B-Instruct30.5213.8 est.Yes440.9 est.NoRTX 5080 16GB
Qwen3-30B-A3B-Instruct-250730.5213.8 est.Yes440.9 est.NoRTX 5080 16GB
GLM-4.7-Flash31.2213.8 est.Yes423.2 est.NoRTX 5080 16GB
Qwen3-1.7B2.0320.7 est.Yes305.5 est.YesRTX 4090D 24GB (China)
gpt-oss-20b21.5178.2 est.Yes204.9 measuredYesRTX 5080 16GB
Llama-3.2-3B-Instruct3.2200.5 est.Yes190.9 est.YesRTX 4090D 24GB (China)
Qwen3-Coder-Next79.7170.4 est.No72.1 est.NoRTX 4090D 24GB (China)
Phi-4-mini-instruct3.8168.8 est.Yes160.8 est.YesRTX 4090D 24GB (China)
Phi-4-mini-reasoning3.8168.8 est.Yes160.8 est.YesRTX 4090D 24GB (China)
gemma-3-4b-it4.3164.5 est.Yes156.6 est.YesRTX 4090D 24GB (China)
Qwen3-Next-80B-A3B-Instruct81.3164.0 est.No69.4 est.NoRTX 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)RTX 5080 16GB
Catalog models that fit33 of 6125 of 61
Largest exclusive fitGLM-4.7-Flash (31.2B)—

RTX 4090D 24GB (China) details → · RTX 5080 16GB details →