RTX 3090 vs RTX 5060 Ti 16GB for local AI
Today the RTX 5060 Ti 16GB is cheaper: $617.99 against $1,497 for the RTX 3090. Per gigabyte of memory the RTX 5060 Ti 16GB wins, $38 against $62. Only the RTX 3090 fits Qwen3.8 27B · Q4_K_M with 8K of context. The RTX 3090 holds more, 24 GB against 16 GB, and fits 11 of our 28 reference models at Q4_K_M where the RTX 5060 Ti 16GB fits 3.
| NVIDIA GeForce RTX 3090 | NVIDIA GeForce RTX 5060 Ti 16GB | |
|---|---|---|
| Lowest price today, delivered | $1,497 on eBay | $617.99 on eBay |
| Price per GB | $62/GB | $38/GB |
| Memory | 24 GB GDDR6X | 16 GB GDDR7 |
| Memory bandwidth | 936 GB/s | 448 GB/s |
| FP16 compute, dense | 71 TFLOPS | 95 TFLOPS |
| Reference models that fit at Q4_K_M | 11 of 28 | 3 of 28 |
| Power | 350 W | 180 W |
| Form factor | Triple slot, full height | Dual slot, full height |
| Interface | PCIe 4 | PCIe 5 |
| Software platform | cuda | cuda |
| Launched | 2020 | 2025 |
| Launch price | $1,499 | $429 |
Qwen3.8 27B · Q4_K_M by quantization
| Quant | RTX 3090 | RTX 5060 Ti 16GB |
|---|---|---|
| FP16 | does not fit | does not fit |
| Q8_0 | does not fit | does not fit |
| Q6_K | ~27 tok/s, 5K ctx | does not fit |
| Q5_K_M | ~32 tok/s | does not fit |
| Q4_K_M | ~37 tok/s | does not fit |
| Q3_K_M | ~46 tok/s | ~22 tok/s |
| IQ2_M | ~66 tok/s | ~32 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.