NVIDIA Tesla T4 vs NVIDIA Tesla V100 SXM3 32 GB
Processeur graphique
| Nom du GPU | TU104 | GV100 |
|---|---|---|
| GPU variant | TU104-895-A1 | — |
| Architecture | Turing | Volta |
| Fonderie | TSMC | TSMC |
| Taille du processus | 12 nm | 12 nm |
| Transistors | 13,600 million | 21,100 million |
| Taille des matrices | 545 mm² | 815 mm² |
Carte graphique
| Date de publication | Sep 13th, 2018 | Mar 27th, 2018 |
|---|---|---|
| Génération | Tesla | Tesla |
| Production | Active | End-of-life |
| Bus interface | PCIe 3.0 x16 | PCIe 3.0 x16 |
Vitesse clock
| Base clock | 585 MHz | 1290 MHz |
|---|---|---|
| Boost clock | 1590 MHz | 1530 MHz |
| Mémoire clock | 1250 MHz 10 Gbps effective | 876 MHz 1752 Mbps effective |
Mémoire
| Taille de la mémoire | 16 GB | 32 GB |
|---|---|---|
| Type de mémoire | GDDR6 | HBM2 |
| Bus mémoire | 256 bit | 4096 bit |
| Bande passante | 320.0 GB/s | 897.0 GB/s |
Render Config
| Shading units | 2560 | 5120 |
|---|---|---|
| TMUs | 160 | 320 |
| ROPs | 64 | 128 |
| Comptage SM | 40 | 80 |
| Tensor cores | 320 | 640 |
| RT cores | 40 | — |
| Cache L1 | 64 KB (per SM) | 128 KB (per SM) |
| Cache L2 | 4 MB | 6 MB |
Performance théorique
| Taux de pixel | 101.8 GPixel/s | 195.8 GPixel/s |
|---|---|---|
| Taux de texture | 254.4 GTexel/s | 489.6 GTexel/s |
| FP16 (half) performance | 65.13 TFLOPS (8:1) | 31.33 TFLOPS (2:1) |
| FP32 (float) performance | 8.141 TFLOPS | 15.67 TFLOPS |
| FP64 (double) performance | 254.4 GFLOPS (1:32) | 7.834 TFLOPS (1:2) |
Board Design
| Taille Slot | Single-slot | Dual-slot |
|---|---|---|
| Longueur | 168 mm 6.6 inches | — |
| TDP | 70 W | 250 W |
| Suggestions pour le PSU | 250 W | 600 W |
| Sorties | No outputs | No outputs |
| Connecteurs de puissance | None | None |
| Board number | PG183 SKU 200 | — |
Caractéristiques graphiques
| DirectX | 12 Ultimate (12_2) | 12 (12_1) |
|---|---|---|
| OpenGL | 4.6 | 4.6 |
| OpenCL | 3.0 | 3.0 |
| Vulkan | 1.2 | 1.2 |
| CUDA | 7.5 | 7.0 |
| Shader model | 6.6 | 6.6 |
