CMP 170HX vs RTX 3090 x2

Best sourced output tok/s per model (single-stream generation) — engine and version, quantization, context and a source link on every row. Hover a value for the quant.

CMP 170HX16 sourced runs · best 255.9 tok/s

RTX 3090 x28 verified runs · best 240.8 tok/s

Specs and evidence

	CMP 170HX	RTX 3090 x2
VRAM	64 GB	24 GB
Sourced runs on the board	16	8
Models measured	11	7
Best output tok/s	255.9 tok/s	240.8 tok/s
Highest measured peak VRAM	58.7 GB	42.9 GB
Street price (latest snapshot)	—	—

Sourced rows only. Peak VRAM is recorded by the person who ran the benchmark, so it is a floor for what the card can hold, not a specification. Price snapshots.

Head-to-head

On the 3 models measured on both cards, CMP 170HX leads on 2, RTX 3090 x2 on 1. The widest gap is on NVIDIA-Nemotron-3.5-Lightning-30B-A3B: CMP 170HX 165.1 tok/s vs RTX 3090 x2 197.0 tok/s (RTX 3090 x2 19% faster).

The measured matrix

Model	CMP 170HX	RTX 3090 x2
Qwen3.8-27B	255.9 W4A16	219.8 AutoRound INT4 + fp8 KV
Qwen3.6-27B	—	240.8 AutoRound INT4
NVIDIA-Nemotron-3.5-Lightning-30B-A3B	165.1 Q8_0	197.0 NVFP4
MiniCPM-V 4.7-35B-A3B	—	172.2 int4
Qwen3.6-35B-A3B	87.2 UD-Q8_K_XL	—
Ornith-1.5-35B-A3B	87.1 Q8_0	—
Nanbeige4.3-33B	—	66.2 W4A16 (INT4 AutoRound, GPTQ, gro
Ornith-1.5-9B	61.5 BF16	—
Qwen3.8-Flash-Next-REAP-256-duo	—	53.5 Unsloth-Dynamic-Q3_K_XL
Qwen3.8-27B-MTP	46.4 INT8-W8A16	—
Qwen3.8-Flash-Next	46.1 UD-IQ1_S	40.1 Unsloth-Dynamic-IQ3_XXS
Llama-4-Scout-17B-16E-Instruct	44.5 UD-Q3_K_XL	—
gemma-4-31B-it	28.3 NVFP4	—
granite-4.2-30b	23.7 Q8_0	—
Llama-3.3-70B-Instruct	20.1 NVFP4	—

Head-to-head (models tested on both): CMP 170HX 2 · RTX 3090 x2 1 · tied 0 · data as of Oct 30, 2026. Full rules: methodology.

Related reading: RTX 5090 vs 3090 · quantization · Value index · Full board