Can A40 48G run Xing4.0 29B-A4B?
Comfortable: GGUF Q8_0 needs 34.9 GB at 4K, batch 1; headroom 13.1 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 Q8_0 | 34.9 | Comfortable | Verified file |
| GGUF Q6_K | 28.6 | Comfortable | Verified file |
| GGUF Q5_K_M | 25.3 | Comfortable | Verified file |
| GGUF Q4_K_M | 21.2 | Comfortable | Verified file |
| GGUF IQ4_XS | 19.0 | Comfortable | Verified file |
| GGUF Q3_K_M | 17.0 | Comfortable | Verified file |
| GGUF IQ3_XXS | 15.3 | Comfortable | Verified file |
| GGUF Q2_K | 13.6 | Comfortable | Verified file |
| GGUF IQ2_XXS | 11.0 | Comfortable | Verified file |
Context capacity · GGUF Q8_0
| Context | Total GB | Verdict |
|---|---|---|
| 4K | 34.9 | Comfortable |
| 16K | 35.5 | Comfortable |
| 32K | 36.3 | Comfortable |
| 128K | 40.9 | Comfortable |
Generation speed
GGUF Q8_0: 172 tok/s · Bandwidth ceiling (not measured)