CMP 170HX vs Radeon AI Pro R9700

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

Radeon AI Pro R970014 verified runs · best 969.4 tok/s

Specs and evidence

	CMP 170HX	Radeon AI Pro R9700
VRAM	64 GB	32 GB
Sourced runs on the board	16	14
Models measured	11	14
Best output tok/s	255.9 tok/s	969.4 tok/s
Highest measured peak VRAM	58.7 GB	92.2 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, Radeon AI Pro R9700 on 1. The widest gap is on Qwen3.6-35B-A3B: CMP 170HX 87.2 tok/s vs Radeon AI Pro R9700 969.4 tok/s (Radeon AI Pro R9700 1012% faster).

The measured matrix

Model	CMP 170HX	Radeon AI Pro R9700
Qwen3.6-35B-A3B	87.2 UD-Q8_K_XL	969.4 MQ4R
Qwen3.5-0.8B	—	554.5 MQ4
LFM2.5-230M	—	520.0 MQ4
LFM2.5-350M	—	466.0 MQ4
Qwen3.5-27B	—	286.6 MQ4-AWQ
Qwen3.8-27B	255.9 W4A16	64.3 Q4_0
LFM2.5-8B-A1B	—	241.1 MQ4
gemma-4-E2B-it	—	169.7 Q8_0
NVIDIA-Nemotron-3.5-Lightning-30B-A3B	165.1 Q8_0	—
gemma-4-E4B-it	—	137.0 Q4_K_M
Muse-Glimmer-30B	—	134.6 Dynamic-Q4_K_XL + DFlash2-Q4_K_M
Ornith-1.5-35B-A3B	87.1 Q8_0	—
Ornith-1.5-9B	61.5 BF16	—
DeepSeek-V4-Flash-0731	—	59.4 IQ2_XXS
Qwen 3.8 Flash-Next Base	—	57.0 IQ3_S
Qwen3.8-27B-MTP	46.4 INT8-W8A16	—
Qwen3.8-Flash-Next	46.1 UD-IQ1_S	25.6 IQ4_XS
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	—
GLM-5.3-Flash	—	17.4 UD-IQ1_M

Head-to-head (models tested on both): CMP 170HX 2 · Radeon AI Pro R9700 1 · tied 0 · data as of Oct 30, 2026. Full rules: methodology.

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