GPU comparison for local inference

GeForce RTX 5090 D vs Titan V

Titan V holds more — 32 GB against 32 GB, which is what decides whether a model runs at all. GeForce RTX 5090 D has 2.06× the memory bandwidth, which is what decides how fast it generates.

Spec sheet· specsFrom the file· fitPredicted· speed

Specifications

	GeForce RTX 5090 D	Titan V
Memory○	32 GB	32 GB
Usable to GPU◐	32 GB	32 GB
Bandwidth○	1792 GB/s	868 GB/s
Memory type○	GDDR7	HBM2
Bus width○	512-bit	4096-bit
Tensor FP16 (dense)○	419 TFLOPS	119 TFLOPS
TDP○	575 W	250 W
MSRP at launch○	$2299	$5999
Models that fit◐	1954	1954

○Spec sheet◐Predictedwhat these mean

Same model, both cards

modeled tokens per second at 32K context

Model	Quant	Params○	GeForce RTX 5090 D◐	Titan V◐	Difference◐
Qwen3-Coder-30B-A3B-InstructMoE	Q6_K	30.5B	119	60	+97%
Qwen3.6-27B	Q6_K	27.8B	54	27	+102%
Qwen3.6-35B-A3BMoE	UD-Q6_K	36.0B	204	107	+90%
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP	IQ3_M	27.8B	44	22	+103%
Qwen3.8-27B	Q6_K	27.8B	54	27	+102%
Qwen3.5-9B	BF16	9.7B	69	34	+101%
gemma-4-26B-A4B-itMoE	Q8_0	26.5B	46	23	+103%
gemma-4-12B-it	BF16	12.0B	51	25	+102%
nemotron-3.5-asr-streaming-0.6b	F32	638M	393	224	+75%
Qwen3.5-4B	BF16	4.7B	131	67	+96%
gemma-4-12B-it-qat-q4_0-unquantized	Q4_0	12.0B	132	68	+96%
Qwythos-9B-Claude-Mythos-5-1M	Q8_0	9.4B	66	33	+101%

Token rates are modeled from memory bandwidth, so on models both cards can hold the ratio tracks bandwidth closely. That is the honest shape of the answer: for inference, capacity decides what you can run and bandwidth decides how fast. Neither is teraflops.

How to read this

We earn nothing from either of these cards. If both hold the models you care about, the faster one wins on bandwidth alone. If one holds a model the other cannot, that difference usually matters far more than any percentage of token rate — a model that does not fit runs 5–20× slower once it starts spilling to system memory, not slightly slower.