NVIDIA A100 40GB vs AMD Strix Halo 128GB for local LLMs
2 reports on the NVIDIA A100 40GB and 320 on the AMD Strix Halo 128GB
Which is faster for local LLMs?
Nobody has reported the same model at the same quant on both cards yet, so there's no measured head-to-head. Memory bandwidth, which caps how fast a card can write, is 6.1x higher on the NVIDIA A100 40GB (1,555 GB/s on the NVIDIA A100 40GB, 256 on the AMD Strix Halo 128GB). The estimates below compare them on the models people run most.
- Memory
- 40GB
- Memory bandwidth
- 1,555 GB/s
- FP16 compute
- 312 TFLOPS
- Reports
- 2
- Memory
- 128GB unified
- Memory bandwidth
- 256 GB/s
- FP16 compute
- 59 TFLOPS
- Reports
- 320
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.
| Model | NVIDIA A100 40GB | AMD Strix Halo 128GB |
|---|---|---|
| Qwen3.8 27B | 31.7 t/sQ8_0 | 3.5 t/sFP16 |
| Qwen3.8 125B · 6B active | 16.8 t/sQ3_K_S, experts in RAM | 12.6 t/sQ6_K |
| Qwen3.6 35B · 3B active | 125.3 t/sQ6_K | 12.6 t/sFP16 |
| DeepSeek V4 Flash 284B · 13B active | doesn't fit at these settings | 33.5 t/sQ2_K |
| Qwen3.6 27B | 31.7 t/sQ8_0 | 3.5 t/sFP16 |
| Gemma 4 26B · 4B active | 81.5 t/sQ8_0 | 9.5 t/sFP16 |
Change the context, RAM or model in the calculator: NVIDIA A100 40GB · AMD Strix Halo 128GB
Every report on each card, with its source: NVIDIA A100 40GB · AMD Strix Halo 128GB