CMP 170HX x4 vs RTX 3090 x8

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.

Sourced data on this pair is thin — most of the board is still partial. Submit your own run and help fill the matrix: one prompt, three steps, no downloads.

CMP 170HX x42 sourced runs · best 264.6 tok/s

RTX 3090 x84 verified runs · best 91.2 tok/s

Specs and evidence

	CMP 170HX x4	RTX 3090 x8
VRAM	64 GB	24 GB
Sourced runs on the board	2	4
Models measured	2	3
Best output tok/s	264.6 tok/s	91.2 tok/s
Highest measured peak VRAM	—	184.3 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 2 models measured on both cards, CMP 170HX x4 leads on 2, RTX 3090 x8 on 0. The widest gap is on GLM-5.3-Flash: CMP 170HX x4 264.6 tok/s vs RTX 3090 x8 75.2 tok/s (CMP 170HX x4 252% faster).

The measured matrix

Model	CMP 170HX x4	RTX 3090 x8
GLM-5.3-Flash	264.6 W4A16	75.2 AWQ-4bit
DeepSeek-V4-Flash-0731	92.1 FP8	91.2 FP8_E4M3
keys-DeepSeekV4Flash-Vision-EXP-ablit	—	66.8 FP8

Head-to-head (models tested on both): CMP 170HX x4 2 · RTX 3090 x8 0 · tied 0 · data as of Oct 30, 2026. Full rules: methodology.

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