Tesla P100 vs Tesla P40 for local AI
Today the Tesla P100 is cheaper: $127.63 against $273 for the Tesla P40. Per gigabyte of memory the Tesla P100 wins, $7 against $11. Only the Tesla P40 fits Qwen3.8 27B · Q4_K_M with 8K of context. The Tesla P40 holds more, 24 GB against 16 GB, and fits 11 of our 28 reference models at Q4_K_M where the Tesla P100 fits 3.
| NVIDIA Tesla P100 16GB | NVIDIA Tesla P40 | |
|---|---|---|
| Lowest price today, delivered | $127.63 on eBay | $273 on eBay |
| Price per GB | $7/GB | $11/GB |
| Memory | 16 GB HBM2 | 24 GB GDDR5 |
| Memory bandwidth | 732 GB/s | 346 GB/s |
| FP16 compute, dense | 18.7 TFLOPS | n/a |
| Reference models that fit at Q4_K_M | 3 of 28 | 11 of 28 |
| Power | 250 W | 250 W |
| Form factor | Dual slot, passive, needs a fan | Dual slot, passive, needs a fan |
| Interface | PCIe 3 | PCIe 3 |
| Software platform | cuda | cuda |
| Launched | 2016 | 2016 |
| Launch price | n/a | n/a |
Qwen3.8 27B · Q4_K_M by quantization
| Quant | Tesla P100 | Tesla P40 |
|---|---|---|
| FP16 | does not fit | does not fit |
| Q8_0 | does not fit | does not fit |
| Q6_K | does not fit | ~10 tok/s, 5K ctx |
| Q5_K_M | does not fit | ~12 tok/s |
| Q4_K_M | does not fit | ~14 tok/s |
| Q3_K_M | ~36 tok/s | ~17 tok/s |
| IQ2_M | ~52 tok/s | ~24 tok/s |
Fit means the weights plus an 8K context cache fit in memory; speed is the estimated decode rate. Prices are the cheapest live eBay listing that passed our filters, item plus shipping. How we estimate.