Side by side datacenter GPU specifications from the Flopper.io database.
| NVIDIA GeForce RTX 5050 8GB | NVIDIA GeForce RTX 5090 32GB | |
|---|---|---|
| FP32 TFLOPS | 13.2 | 104.8 |
| VRAM | 8 GB | 32 GB |
| Memory Bandwidth | 320 GB/s | 1.8 TB/s |
| TDP | 130 W | 575 W |
| Precision | NVIDIA GeForce RTX 5050 8GB peak TFLOPS (dense) | NVIDIA GeForce RTX 5090 32GB peak TFLOPS (dense) |
|---|---|---|
| FP64 | 0.2 | 1.6 |
| FP32 | 13.2 | 104.8 |
| TF32 | 13.2 | 104.8 |
| BF16 | 26.3 | 209.5 |
| FP16 | 52.6 | 419.1 |
| FP8 | 105.3 | 838.2 |
| FP6 | — | — |
| FP4 | 210.5 | 1676.0 |
| INT8 | 105.3 | 838.2 |
| NVIDIA GeForce RTX 5050 8GB | NVIDIA GeForce RTX 5090 32GB | |
|---|---|---|
| Architecture | Blackwell | Blackwell |
| Process Node | 4nm | 4nm |
| Launch Year | 2025 | 2025 |
| Form Factor | PCIe | PCIe |
| Memory Type | GDDR6 | GDDR7 |
| Interface Width | 128-bit | 512-bit |
| Interconnect | — | — |
| TDP | 130 W | 575 W |
| FP32 Efficiency | 0.10 TFLOPS/W | 0.18 TFLOPS/W |