RTX 3090 vs RTX 5090

Good budget option

Editorial opinionOur verdict

The RTX 5090 is about twice as fast (1.8 vs 0.94 TB/s) and has 32 GB instead of 24, but in 2026 it costs roughly three times as much as a used 3090. Two RTX 3090s give you 48 GB for less than one 5090, enough for 70B models. Buy the 5090 for speed in one card; the 3090 for value.

Factual data	RTX 3090	RTX 5090
Memory (largest)	24 GB	32 GBMore
Usable for models	24 GB	32 GBMore
Bandwidth	936 GB/s	1.79 TB/sMore
Largest popular model at 4-bit	Qwen3.6 27B	Qwen3.6 35B A3B
Price from	$1,340Advantage	$4,329
Price per GB	$56Advantage	$135
Power	350 W	575 W
Compute stack	CUDA	CUDA

Which models each one runs

Popular models at Q4_K_M with a 32K-token context. Memory needed from real model files; speed is the theoretical ceiling.

Model	RTX 3090	RTX 5090
Qwen3 1.7B	Runs≤ 843 tok/s	Runs≤ 1,614 tok/s
Qwen3.5 9B	Runs≤ 165 tok/s	Runs≤ 315 tok/s
Gemma 4 12B	Runs≤ 131 tok/s	Runs≤ 252 tok/s
gpt-oss-20b	Runs≤ 468 tok/s	Runs≤ 895 tok/s
Qwen3.6 27B	Runs≤ 56 tok/s	Runs≤ 107 tok/s
Qwen3.6 35B A3B	Partly in system RAM	Runs≤ 972 tok/s
Llama 3.3 70B	Not enough memory	Not enough memory
gpt-oss-120b	Partly in system RAM	Partly in system RAM
Qwen3.5 122B A10B	Partly in system RAM	Partly in system RAM
Qwen3.5 397B A17B	Partly in system RAM	Partly in system RAM

Speed ceilings are theoretical; see each page for measured results.