NVIDIA Tesla P4 vs NVIDIA Tesla T4
Processeur graphique
| Nom du GPU | GP104 | TU104 |
|---|---|---|
| GPU variant | GP104-895-A1 | TU104-895-A1 |
| Architecture | Pascal | Turing |
| Fonderie | TSMC | TSMC |
| Taille du processus | 16 nm | 12 nm |
| Transistors | 7,200 million | 13,600 million |
| Taille des matrices | 314 mm² | 545 mm² |
Carte graphique
| Date de publication | Sep 13th, 2016 | Sep 13th, 2018 |
|---|---|---|
| Génération | Tesla | Tesla |
| Production | End-of-life | Active |
| Bus interface | PCIe 3.0 x16 | PCIe 3.0 x16 |
Vitesse clock
| Base clock | 886 MHz | 585 MHz |
|---|---|---|
| Boost clock | 1114 MHz | 1590 MHz |
| Mémoire clock | 1502 MHz 6 Gbps effective | 1250 MHz 10 Gbps effective |
Mémoire
| Taille de la mémoire | 8 GB | 16 GB |
|---|---|---|
| Type de mémoire | GDDR5 | GDDR6 |
| Bus mémoire | 256 bit | 256 bit |
| Bande passante | 192.3 GB/s | 320.0 GB/s |
Render Config
| Shading units | 2560 | 2560 |
|---|---|---|
| TMUs | 160 | 160 |
| ROPs | 64 | 64 |
| Comptage SM | 20 | 40 |
| Cache L1 | 48 KB (per SM) | 64 KB (per SM) |
| Cache L2 | 2 MB | 4 MB |
| Tensor cores | — | 320 |
| RT cores | — | 40 |
Performance théorique
| Taux de pixel | 71.30 GPixel/s | 101.8 GPixel/s |
|---|---|---|
| Taux de texture | 178.2 GTexel/s | 254.4 GTexel/s |
| FP16 (half) performance | 89.12 GFLOPS (1:64) | 65.13 TFLOPS (8:1) |
| FP32 (float) performance | 5.704 TFLOPS | 8.141 TFLOPS |
| FP64 (double) performance | 178.2 GFLOPS (1:32) | 254.4 GFLOPS (1:32) |
Board Design
| Taille Slot | Single-slot | Single-slot |
|---|---|---|
| Longueur | 168 mm 6.6 inches | 168 mm 6.6 inches |
| TDP | 75 W | 70 W |
| Suggestions pour le PSU | 250 W | 250 W |
| Sorties | No outputs | No outputs |
| Connecteurs de puissance | None | None |
| Board number | — | PG183 SKU 200 |
Caractéristiques graphiques
| DirectX | 12 (12_1) | 12 Ultimate (12_2) |
|---|---|---|
| OpenGL | 4.6 | 4.6 |
| OpenCL | 3.0 | 3.0 |
| Vulkan | 1.2 | 1.2 |
| CUDA | 6.1 | 7.5 |
| Shader model | 6.4 | 6.6 |
