NVIDIA RTX Pro 6000 Blackwell vs AMD Strix Halo 128GB for local LLMs
58 reports on the NVIDIA RTX Pro 6000 Blackwell and 111 on the AMD Strix Halo 128GB
Which is faster for local LLMs?
On Qwen3.8 125B · 6B active at 4-bit, the NVIDIA RTX Pro 6000 Blackwell runs at 101.9 tokens per second (5 runs) and the AMD Strix Halo 128GB at 22.0 (4 runs). Those are community reports whose engines and settings differ, so the gap is a rough sign, not a controlled measure of the two cards. Memory bandwidth, which caps how fast a card can write, is 7.0x higher on the NVIDIA RTX Pro 6000 Blackwell (1,792 GB/s on the NVIDIA RTX Pro 6000 Blackwell, 256 on the AMD Strix Halo 128GB). Across the 2 models both have been measured on, the NVIDIA RTX Pro 6000 Blackwell's median is higher on 2 and the AMD Strix Halo 128GB's on 0.
- Memory
- 96GB
- Memory bandwidth
- 1,792 GB/s
- FP16 compute
- 503 TFLOPS
- Reports
- 58
- Memory
- 128GB unified
- Memory bandwidth
- 256 GB/s
- FP16 compute
- 59 TFLOPS
- Reports
- 111
Measured on both
Median tokens per second from plain runs (one device, one request, no speculative decoding, the whole model in the device's memory) of the same model size at the same bit class. Each side shows the engines and reported context behind its median. Context can mean a configured window or prompt depth, so these are community comparisons with differing settings. How to read these
| Model | NVIDIA RTX Pro 6000 Blackwell | AMD Strix Halo 128GB |
|---|---|---|
| Qwen3.8 125B · 6B active4-bit | 101.95 runsllama.cpp, ik_llama.cpp +1, 1 not stated512 to 240K context, 1 not stated | 22.04 runsGufo, llama.cpp128K to 160K context |
| Qwen3.8 27B4-bit | 67.0one runengine not statedcontext not stated | 14.02 runsllama.cpp32K context, 1 not stated |
Estimated by the calculator
The quant the calculator recommends for each card and its estimated speed at a 32,768-token context with 32 GB of system RAM, from memory bandwidth and the model's shape. Where a model was also measured above, the measurement wins.
| Model | NVIDIA RTX Pro 6000 Blackwell | AMD Strix Halo 128GB |
|---|---|---|
| Qwen3.8 125B · 6B active | 83.1 t/sQ5_K_S | 12.6 t/sQ6_K |
| Qwen3.8 27B | 18.3 t/sFP16 | 3.5 t/sFP16 |
| DeepSeek V4 Flash 284B · 13B active | 130.8 t/sQ2_K | 33.5 t/sQ2_K |
| Qwen3.6 35B · 3B active | 57.5 t/sFP16 | 12.6 t/sFP16 |
| Qwen3.6 27B | 18.3 t/sFP16 | 3.5 t/sFP16 |
| GLM-5.3 320B · 18B active | 6.8 t/sQ2_K, experts in RAM | 11.1 t/sQ2_K |
| Gemma 4 26B · 4B active | 49.3 t/sFP16 | 9.5 t/sFP16 |
| DeepSeek V4.1 Flash 552B · 16B active | doesn't fit at these settings | doesn't fit at these settings |
Change the context, RAM or model in the calculator: NVIDIA RTX Pro 6000 Blackwell · AMD Strix Halo 128GB
Every report on each card, with its source: NVIDIA RTX Pro 6000 Blackwell · AMD Strix Halo 128GB