Can RTX 4070 Super run Hy-Memory-30B-A3B?
Comfortable: GGUF Q8_0-expsIQ2_XXS needs 9.9 GB at 4K, batch 1; headroom 2.1 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 Q3_K_M | 15.5 | Does not fit | Verified file |
| GGUF Q4_K_M | 19.5 | Does not fit | Verified file |
| GGUF Q6_K | 26.2 | Does not fit | Verified file |
| GGUF Q8_0 | 33.8 | Does not fit | Verified file |
| GGUF Q8_0-expsIQ2_XXS | 9.9 | Comfortable | Verified file |
| GGUF Q8_0-expsIQ1_M | 8.8 | Comfortable | Verified file |
Context capacity · GGUF Q8_0-expsIQ2_XXS
| Context | Total GB | Verdict |
|---|---|---|
| 4K | 9.9 | Comfortable |
| 16K | 11.2 | Tight |
| 32K | 12.8 | Does not fit |
| 128K | 22.7 | Does not fit |
Generation speed
GGUF Q8_0-expsIQ2_XXS: 582 tok/s · Bandwidth ceiling (not measured)