Can the Tesla P40 run Qwen3.8-27B-DSpine-b16?
Qwen3.8-27B-DSpine-b16·2B·Tesla P40·24 GB VRAM
Yes. The Tesla P40 runs Qwen3.8-27B-DSpine-b16 fully in VRAM at FP16 / BF16 (about 4.6 GB of 24 GB), at roughly 47 tokens/s.
- best quant
- FP16 / BF16
- needs
- 4.6 GB
- usable
- 24 GB
- est. speed
- ~47 tok/s
Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the Qwen3.8-27B-DSpine-b16 page →
Every quantisation of Qwen3.8-27B-DSpine-b16 on the Tesla P40
| Quant | Memory | Fit | Speed |
|---|---|---|---|
| Q2_K | 1.3 GB | fits | ~287 tok/s |
| Q3_K_M | 1.6 GB | fits | ~193 tok/s |
| Q4_K_M | 1.9 GB | fits | ~155 tok/s |
| Q5_K_M | 2.1 GB | fits | ~132 tok/s |
| Q6_K | 2.3 GB | fits | ~114 tok/s |
| Q8_0 | 2.8 GB | fits | ~89 tok/s |
| FP16 / BF16 pick | 4.6 GB | fits | ~47 tok/s |
Speed is a bandwidth-based estimate for generating text (347 GB/s); prompt processing and long contexts are slower.
Measured speed on the Tesla P40
No community reports for this pair yet.
Report your own speed →Chat models that fit the Tesla P40 comfortably
All that fit →| Model | Params | Best quant | Memory | Speed |
|---|---|---|---|---|
| 1.7B | Q8_0 | ≈2.7 GBFits in memory | ~104 tok/s | |
| 2B | Q8_0 | ≈3.2 GBFits in memory | ~88 tok/s | |
| 3.2B | Q8_0 | ≈4.7 GBFits in memory | ~56 tok/s | |
| 1.5B | Q8_0 | ≈2.6 GBFits in memory | ~116 tok/s | |
| 2.5B | Q8_0 | ≈3.5 GBFits in memory | ~71 tok/s | |
| 3.1B | Q8_0 | ≈4.3 GBFits in memory | ~58 tok/s |
Other GPUs that run Qwen3.8-27B-DSpine-b16
Frequently asked questions
- Can the Tesla P40 run Qwen3.8-27B-DSpine-b16?
- Yes. The Tesla P40 runs Qwen3.8-27B-DSpine-b16 fully in VRAM at FP16 / BF16 (about 4.6 GB of 24 GB), at roughly 47 tokens/s.
- How fast does Qwen3.8-27B-DSpine-b16 run on the Tesla P40?
- An estimate from memory bandwidth gives about 47 tokens/s at FP16 / BF16.
- Which quantisation of Qwen3.8-27B-DSpine-b16 should I use on the Tesla P40?
- FP16 / BF16: the largest one that fits. Expected quality: lossless compared with the full-precision weights.
Memory and speed are estimates from the model's size and the card's memory bandwidth; real numbers depend on the runtime, drivers and context length. Community figures are medians of user reports.