Tesla P100 vs Instinct MI50 16GB for local AI
Today the Tesla P100 is cheaper: $127.63 against $228.78 for the Instinct MI50 16GB. Per gigabyte of memory the Tesla P100 wins, $7 against $14. Neither fits Qwen3.8 27B · Q4_K_M with 8K of context.
| NVIDIA Tesla P100 16GB | AMD Radeon Instinct MI50 16GB | |
|---|---|---|
| Lowest price today, delivered | $127.63 on eBay | $228.78 on eBay |
| Price per GB | $7/GB | $14/GB |
| Memory | 16 GB HBM2 | 16 GB HBM2 |
| Memory bandwidth | 732 GB/s | 1,024 GB/s |
| FP16 compute, dense | 18.7 TFLOPS | 26.5 TFLOPS |
| Reference models that fit at Q4_K_M | 3 of 28 | 3 of 28 |
| Power | 250 W | 300 W |
| Form factor | Dual slot, passive, needs a fan | Dual slot, passive, needs a fan |
| Interface | PCIe 3 | PCIe 4 |
| Software platform | cuda | rocm |
| Launched | 2016 | 2018 |
| Launch price | n/a | n/a |
Qwen3.8 27B · Q4_K_M by quantization
| Quant | Tesla P100 | Instinct MI50 16GB |
|---|---|---|
| FP16 | does not fit | does not fit |
| Q8_0 | does not fit | does not fit |
| Q6_K | does not fit | does not fit |
| Q5_K_M | does not fit | does not fit |
| Q4_K_M | does not fit | does not fit |
| Q3_K_M | ~36 tok/s | ~39 tok/s |
| IQ2_M | ~52 tok/s | ~56 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.