can it run
11.1 GB vs 10 GB usable — partial CPU offload, expect large speed loss With a q8 (quantized) KV cache it comes to the edge (9.6GB) — possible with modest context, at some quality cost.
10GB · RDNA2 · model file 8.0GB · 13B
computed roofline: 320 GB/s × 0.45–0.72 efficiency window / 8.0GB (Q4_K_M) — bands, never points.
→ Tesla P100 (16GB, used) — cheapest catalog machine that runs llava 13B fully ($135)
→ rx-6700-xt — the next step up from AMD Radeon RX 6700 10GB (non-XT)
→ every model the AMD Radeon RX 6700 10GB (non-XT) can run — full list
→explore the full catalog84 machines indexed · live prices · what each one can run