AMD Radeon AI PRO R9700 vs NVIDIA GeForce RTX 4090
For running models locally, priced in Canada. At 64k context the first runs 27 models fully in memory, the second 22.
Specifications
| Spec | AMD Radeon AI PRO R9700 | NVIDIA GeForce RTX 4090 |
|---|---|---|
| GPU | Radeon AI PRO R9700 | GeForce RTX 4090 |
| Memory | 32 GB GDDR6 | 24 GB GDDR6X |
| Memory bus | 256-bit | 384-bit |
| Memory bandwidth | 640 GB/s | 1,008 GB/s |
| FP16 compute | 191 TFLOPS | 165 TFLOPS |
| Software | ROCm | CUDA |
| Board power | 300 W | 450 W |
| PCIe generation | Gen 5 | Gen 4 |
| PCIe lanes | x16 | x16 |
What each runs
Estimated decode speed with the most precise quant that fits each one est.
| Model | AMD Radeon AI PRO R9700 | NVIDIA GeForce RTX 4090 |
|---|---|---|
| Qwen 3.8 27B | 13tok/sUD-Q6_K | 30tok/sUD-Q4_K_M |
| Qwen 3.6 27B | 13tok/sQ6_K | 26tok/sQ5_K_M |
| Gemma 4 31B | 11tok/sQ6_K | offload |
| Qwen 3.6 35B-A3B | 55tok/sUD-Q6_K | 115tok/sUD-Q4_K_M |
| Qwen 3.5 27B | 13tok/sQ6_K | 30tok/sQ4_K_M |
| Gemma 4 26B-A4B | 37tok/sQ8_0 | 78tok/sUD-Q5_K_M |
| Qwen 3.5 35B-A3B | 55tok/sQ6_K | 116tok/sQ4_K_M |
| GLM 4.7 Flash | 28tok/sQ6_K | 57tok/sQ4_K_M |
| Devstral Small 2 24B | 12tok/sQ6_K | offload |
| Kimi Linear 48B-A3B | 88tok/sQ4_K_M | offload |
| gpt-oss 20B | 52tok/sMXFP4 | 93tok/sMXFP4 |
| Qwen3 Coder 30B-A3B | 24tok/sQ6_K | offload |
| Gemma 4 12B | 25tok/sQ8_0 | 44tok/sQ8_0 |
| Qwen 3.5 9B | 30tok/sQ8_0 | 53tok/sQ8_0 |
| Gemma 3 27B | 13tok/sQ6_K | 28tok/sQ4_K_M |
| Mistral Small 3.2 24B | 12tok/sQ6_K | offload |
| Ministral 3 14B | 14tok/sQ8_0 | 29tok/sQ6_K |
| Qwen3 14B | 13tok/sQ8_0 | 27tok/sQ6_K |
| DeepSeek R1 0528 Qwen3 8B | 19tok/sQ8_0 | 34tok/sQ8_0 |
| Qwen3 8B | 19tok/sQ8_0 | 34tok/sQ8_0 |
| Phi-4 14B | 12tok/sQ8_0 | 26tok/sQ5_K_M |
| Qwen 3.5 4B | 53tok/sQ8_0 | 94tok/sQ8_0 |
| Ministral 3 8B | 19tok/sQ8_0 | 34tok/sQ8_0 |
| Granite 4.2 8B | 18tok/sQ8_0 | 31tok/sQ8_0 |
| Gemma 4 E4B | 35tok/sQ8_0 | 62tok/sQ8_0 |
| LFM2.5 8B-A1B | 49tok/sQ8_0 | 88tok/sQ8_0 |
| Llama 3.1 8B | 23tok/sQ6_K | 41tok/sQ6_K |
| Qwen 3.5 122B-A10B | offload | offload |
| Qwen3 Coder Next | offload | offload |
| Nemotron 3 Super 120B-A12B | offload | won't run |
| Mistral Small 4 119B | offload | offload |
| GLM 4.5 Air | offload | won't run |
| Qwen3 Next 80B-A3B | offload | offload |
| Nemotron 3.5 Lightning 30B-A3B | offload | offload |
| Qwen3 32B | offload | offload |
| DeepSeek R1 Distill Qwen 32B | offload | offload |
| DeepSeek V4 Flash | won't run | won't run |
| GLM 5.3 Flash | won't run | won't run |
| Qwen 3.8 Flash Next | won't run | won't run |
| MiniMax M2.7 | won't run | won't run |
| gpt-oss 120B | won't run | won't run |
| Llama 3.3 70B | won't run | won't run |