▸ DEVICE UNDER TEST
NVIDIA CMP 90HX — 10 GB VRAM.
▸ CMP 90HX SPEC
- BRAND
- NVIDIA
- VRAM
- 10 GB GDDR6X
- BANDWIDTH
- 760 GB/s
- FP16 COMPUTE
- 21.9 TFLOPS
- FP32 COMPUTE
- 21.9 TFLOPS
- CUDA CORES
- 6,400
- TENSOR CORES
- 200
- TDP
- 320 W
- ARCHITECTURE
- Ampere
▸ AI CAPABILITY
177/ 331 models @ Q4
With 10 GB VRAM and 760 GB/s bandwidth, this GPU handles models up to 13.1B parameters.
Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~87 tok/s.
§ 01TOP MODELS FOR CMP 90HX
177 FIT · SHOWING 20| MODEL | SIZE | VRAM Q4 | TOK/S | AVG |
|---|---|---|---|---|
| LLaVA-1.5 13B | 13.1B | 8.5 GB | 46 | 52.1 |
| Baichuan2 13B | 13B | 8.4 GB | 47 | 23.6 |
| Llama 2 13B | 13B | 8.4 GB | 47 | 19.7 |
| CodeLlama 13B | 13B | 8.4 GB | 47 | 19.7 |
| Vicuna 13B | 13B | 8.4 GB | 47 | 11.8 |
| LLaMA 1 13B | 13B | 8.4 GB | 47 | 32.9 |
| OPT 13B | 13B | 8.4 GB | 47 | 35.8 |
| Orca 2 13B | 13B | 8.4 GB | 47 | 25.4 |
| LLaVA-1.6 Vicuna 13B | 13B | 8.4 GB | 47 | 37.4 |
| WizardCoder Python 13B | 13B | 8.4 GB | 47 | 60.1 |
| WizardLM 13B | 13B | 8.4 GB | 47 | 19.5 |
| Mistral-Nemo 12.2B | 12.2B | 7.9 GB | 50 | 22.4 |
| Dolly v2 12B | 12B | 7.8 GB | 51 | 6.4 |
| gemma-3-12b | 12B | 7.8 GB | 51 | 26.6 |
| TranslateGemma 12B | 12B | 7.8 GB | 51 | 35.1 |
| Pixtral 12B | 12B | 7.8 GB | 51 | 52.1 |
| StableLM 2 12B | 12B | 7.8 GB | 51 | 21.3 |
| Falcon2 11B | 11B | 7.2 GB | 55 | 33.2 |
| Llama-3.2-11B-Vision-Instruct | 11B | 7.2 GB | 55 | 34.4 |
| SOLAR-10.7B | 10.7B | 7.0 GB | 57 | 28.2 |