Side by side datacenter GPU specifications from the Flopper.io database.
| NVIDIA Quadro RTX 8000 48GB | NVIDIA T4 16GB | |
|---|---|---|
| FP32 TFLOPS | 16.3 | 8.1 |
| VRAM | 48 GB | 16 GB |
| Memory Bandwidth | 672 GB/s | 300 GB/s |
| TDP | 295 W | 70 W |
| Precision | NVIDIA Quadro RTX 8000 48GB peak TFLOPS (dense) | NVIDIA T4 16GB peak TFLOPS (dense) |
|---|---|---|
| FP64 | 0.5 | — |
| FP32 | 16.3 | 8.1 |
| TF32 | — | — |
| BF16 | — | — |
| FP16 | 130.5 | 65.0 |
| FP8 | — | — |
| FP6 | — | — |
| FP4 | — | — |
| INT8 | 261.0 | 130.0 |
| INT4 | 522.0 | 260.0 |
| NVIDIA Quadro RTX 8000 48GB | NVIDIA T4 16GB | |
|---|---|---|
| Architecture | Turing | Turing |
| Process Node | 12nm | 12nm |
| Launch Year | 2018 | 2019 |
| Form Factor | PCIe | PCIe |
| Memory Type | GDDR6 | GDDR6 |
| Interface Width | 384-bit | — |
| Interconnect | NVLink | — |
| TDP | 295 W | 70 W |
| FP32 Efficiency | 0.06 TFLOPS/W | 0.12 TFLOPS/W |