🔌
New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
// RUN — FITS
Can the NVIDIA L40S 48GB run P-Llama2?
Yes — here's how fast and up to what context, with real measured numbers.
P-Llama2 on the NVIDIA L40S 48GB
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 219k | 45.1-53.7 t/s | ~2.1-2.5 (slow) |
| Q8 | yes | 272k | 90.1-107.4 t/s | ~4.2-5 (slow) |
| Q4 | yes | 296k | 160.9-191.8 t/s | ~7.4-8.9 (slow) |
Estimate (memory-bound). Beyond "max context" the KV cache spills to system RAM offload (depends on your system RAM) and speed drops to the "beyond" figure.
FAQ
Can the NVIDIA L40S 48GB run P-Llama2?▶
Yes. At Q4 it generates ~160.9-191.8 tok/s and fits up to 296k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~7.4-8.9 tok/s.
How much context fits?▶
At Q4, up to about 296k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.