Head to head
RTX 5090 vs RTX 4060 Ti 16GB
The RTX 5090 holds more — 32 GB against 16 GB — which decides what you can load at all. The RTX 5090 has more bandwidth at 1792 GB/s, which decides how fast tokens come out once a model fits.
| Model (Q4_K_M, 8K) | RTX 5090 | tok/s | RTX 4060 Ti 16GB | tok/s |
|---|---|---|---|---|
| Llama 3.1 8B | Runs comfortably | 222 | Runs comfortably | 33.6 |
| Llama 3.1 70B | Won't fit | 2.65 | Won't fit | 1.17 |
| Llama 3.2 3B | Runs comfortably | 478 | Runs comfortably | 74.8 |
| Llama 3.2 1B | Runs comfortably | 1235 | Runs comfortably | 200 |
| Llama 4 Scout 109B-A17B | Won't fit | 6.75 | Won't fit | 4.28 |
| Qwen3 8B | Runs comfortably | 214 | Runs comfortably | 32.5 |
| Qwen3 14B | Runs comfortably | 126 | Runs comfortably | 18.9 |
| Qwen3 32B | Runs comfortably | 59.0 | Won't fit | 3.71 |
| Qwen3 4B | Runs comfortably | 380 | Runs comfortably | 59.6 |
| Qwen3 30B-A3B | Runs comfortably | 159 | Won't fit | 32.0 |
| Qwen3 235B-A22B | Won't fit | 3.83 | Won't fit | 3.01 |
| Qwen2.5-Coder 32B | Runs comfortably | 59.0 | Won't fit | 3.71 |
| Qwen2.5 7B | Runs comfortably | 247 | Runs comfortably | 37.3 |
| Qwen2.5 72B | Won't fit | 2.50 | Won't fit | 1.13 |
| Gemma 3 4B | Runs comfortably | 405 | Runs comfortably | 63.6 |
| Gemma 3 12B | Runs comfortably | 151 | Runs comfortably | 22.8 |
Specifications
| RTX 5090 | RTX 4060 Ti 16GB | |
|---|---|---|
| Memory | 32 GB | 16 GB |
| Bandwidth | 1792 GB/s | 288 GB/s |
| FP16 compute | 209 TFLOPS | 44 TFLOPS |
| Launch price | $1,999 | $499 |
| Architecture | Blackwell | Ada |