RTX 3090 Ti 24GB vs RTX 4090D 24GB (China) for Local LLMs

Both list the same memory bandwidth (1,008 GB/s), so per-token decode speed should be similar whenever a model fits both. Both fit the same number of catalog models at Q4/8K (33 of 61; 24 vs 24 GB VRAM). MSRP is $1,999 for the RTX 3090 Ti 24GB vs $1,599 for the RTX 4090D 24GB (China). 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 (47 shared models), the two cards split the rows 0–0. Methodology.

Spec comparison

	RTX 3090 Ti 24GB	RTX 4090D 24GB (China)
Vendor	NVIDIA	NVIDIA
VRAM	24 GB	24 GB
Memory type	GDDR6X	GDDR6X
Memory bandwidth	1,008 GB/s	1,008 GB/s
TDP	450 W	425 W
MSRP (list)	$1,999	$1,599
Buy	Search Amazon (affiliate)	—

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 47 shared models

Model	Params (B)	RTX 3090 Ti 24GB tok/s	RTX 3090 Ti 24GB fits	RTX 4090D 24GB (China) tok/s	RTX 4090D 24GB (China) fits	Winner
Qwen3-0.6B	0.8	801.8 est.	Yes	801.8 est.	Yes	Tie
gemma-3-1b-it	1.0	641.5 est.	Yes	641.5 est.	Yes	Tie
Llama-3.2-1B-Instruct	1.2	534.5 est.	Yes	534.5 est.	Yes	Tie
Qwen3-1.7B	2.0	320.7 est.	Yes	320.7 est.	Yes	Tie
Qwen3-30B-A3B	30.5	213.8 est.	Yes	213.8 est.	Yes	Tie
Qwen3-Coder-30B-A3B-Instruct	30.5	213.8 est.	Yes	213.8 est.	Yes	Tie
GLM-4.7-Flash	31.2	213.8 est.	Yes	213.8 est.	Yes	Tie
Qwen3-30B-A3B-Instruct-2507	30.5	213.8 est.	Yes	213.8 est.	Yes	Tie
Llama-3.2-3B-Instruct	3.2	200.5 est.	Yes	200.5 est.	Yes	Tie
gpt-oss-20b	21.5	178.2 est.	Yes	178.2 est.	Yes	Tie
Qwen3-Coder-Next	79.7	170.4 est.	No	170.4 est.	No	Tie
Phi-4-mini-instruct	3.8	168.8 est.	Yes	168.8 est.	Yes	Tie
Phi-4-mini-reasoning	3.8	168.8 est.	Yes	168.8 est.	Yes	Tie
gemma-3-4b-it	4.3	164.5 est.	Yes	164.5 est.	Yes	Tie
Qwen3-Next-80B-A3B-Instruct	81.3	164.0 est.	No	164.0 est.	No	Tie

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 3090 Ti 24GB	RTX 4090D 24GB (China)
Catalog models that fit	33 of 61	33 of 61
Largest exclusive fit	—	—

RTX 3090 Ti 24GB details → · RTX 4090D 24GB (China) details →