GPUs & Macs › AMD Radeon Pro W5700
AMD Radeon Pro W5700: local LLM speed
Llama 2 7B (Q4_0) generates 71.3 tokens/s on the AMD Radeon Pro W5700 (Vulkan), and processes prompts at 447 tokens/s.
Measured results
| Model | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| Llama 2 7B | 71.3 | 447 | Q4_0 | Vulkan FA on | tg128 | llama.cpp #10879 details |
| Llama 2 7B | 68.5 | 450 | Q4_0 | Vulkan FA off | tg128 | llama.cpp #10879 details |
What fits on the AMD Radeon Pro W5700
Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 71.3 tok/s (Vulkan, source). Fit assumes the whole model in GPU memory with ~8K context.
Memory size unknown for this device.
Similar hardware
AMD Radeon Pro W6800X Duo · AMD Radeon Pro W6800X · AMD Radeon Pro V540 · AMD FirePro S9300 x2 · AMD Radeon Pro V340 · AMD Radeon RX 470 · AMD Radeon Pro W5500 · AMD Radeon RX 480