Strix Halo Monitor
GPU / UMA memory
Shared GPU-addressable GTT/UMA. Useful Strix Halo model headroom.
GPU
CPU
Qwen / llama.cpp · endpoint A
| Build | |
| KV cache | |
| Flash Attention | |
| Vision projector | |
| Speculative / MTP |
Qwen / llama.cpp · endpoint B
| Build | |
| KV cache | |
| Flash Attention | |
| Vision projector | |
| Speculative / MTP |
Storage
System memory