Can Radeon RX 9070 run Gemma 3 27B IT?
Tight: GGUF Q3_K_M needs 15.1 GB at 4K, batch 1; headroom 0.9 GB.
Text input, FP16 KV cache, one card. Unified-memory capacity uses the GPU guide’s usable limit; negative headroom may require host offload. Speed source settings may differ from this 4K estimate.
Measured runs
No measured run for this pair. Sizing is an estimate, not a runtime test.
Quant memory at 4K
| Quant | Total GB | Verdict | Weight sizing |
|---|---|---|---|
| GGUF Q4_K_M | 18.3 | Does not fit | Verified file |
| GGUF Q3_K_M | 15.1 | Tight | Verified file |
| AWQ INT4 | 15.1 | Tight | Estimated weights |
Context capacity · GGUF Q3_K_M
| Context | Total GB | Verdict |
|---|---|---|
| 4K | 15.1 | Tight |
| 16K | 16.1 | Tight |
| 32K | 17.5 | Does not fit |
| 128K | 25.7 | Does not fit |
Generation speed
GGUF Q3_K_M: 32–43 tok/s · Derived range