← Back to comparison
AMD
AMD Radeon RX 7900 XTX
RDNA 3 · RX 7000 Series · ACTIVE
We earn commissions from purchases made through links on this site. This doesn't affect our rankings or recommendations.
Prices updated 5 days ago
$/GB VRAM (best)
$26/GB
MSRP: $999 (90% of MSRP)
Specifications
| VRAM | 24 GB |
| Memory type | GDDR6 |
| Bus width | 384-bit |
| Memory bandwidth | 960 GB/s |
| Stream processors | 6,144 |
| FP16 | 122.8 TFLOPS |
| TDP | 355W |
| Power connector | 2×8-pin |
| Card length | 287 mm |
| Slot width | 2.5 slots |
| PCIe | Gen 4 x16 |
| Max model (Q4) | ~44B parameters |
Inference Benchmarks (Q4_K_M)
Llama 3.3 8B
80.0 tok/s
Qwen 3 32B
28.0 tok/s
Llama 3.3 70B
—
llama.cpp, batch_size=1, ctx=4096, single GPU.
What Can You Run?
| Model | Q4_K_M | Q8_0 | FP16 |
|---|---|---|---|
| Llama 3.3 8B8B | Excellent~80 tok/s | Usable | Usable |
| Llama 3.3 70B70.6B | Won't fit | Won't fit | Won't fit |
| Qwen 3 8B8.2B | Usable | Usable | Usable |
| Qwen 3 32B32.8B | Good~28 tok/s | Won't fit | Won't fit |
| DeepSeek R1 70B70.6B | Won't fit | Won't fit | Won't fit |
| Mistral Nemo 12B12.2B | Usable | Usable | Won't fit |
| Phi-4 14B14B | Usable | Usable | Won't fit |
| Gemma 3 27B27.4B | Usable | Won't fit | Won't fit |
| Codestral 25B25.3B | Usable | Won't fit | Won't fit |
| Command R 35B35B | Usable | Won't fit | Won't fit |
Notes
Best AMD option for local inference. ROCm support is improving but still behind CUDA.