Nom du GPU | GM20B | GV100 |
---|---|---|
GPU variant | TM670D-A1 | — |
Architecture | Maxwell 2.0 | Volta |
Fonderie | TSMC | TSMC |
Taille du processus | 20 nm | 12 nm |
Transistors | 2,000 million | 21,100 million |
Taille des matrices | 118 mm² | 815 mm² |
Date de publication | Jan 4th, 2015 | — |
---|---|---|
Génération | Tegra | — |
Production | End-of-life | — |
Bus interface | IGP | — |
GPU clock | 1000 MHz | — |
---|---|---|
Mémoire clock | System Shared | 876 MHz 1752 Mbps effective |
Base clock | — | 1230 MHz |
Boost clock | — | 1380 MHz |
Taille de la mémoire | System Shared | 32 GB |
---|---|---|
Type de mémoire | System Shared | HBM2 |
Bus mémoire | System Shared | 4096 bit |
Bande passante | System Dependent | 897.0 GB/s |
Shading units | 256 | 5120 |
---|---|---|
TMUs | 16 | 320 |
ROPs | 16 | 128 |
Comptage SMM | 2 | — |
Cache L1 | 48 KB (per SMM) | 128 KB (per SM) |
Cache L2 | 256 KB | 6 MB |
Comptage SM | — | 80 |
Tensor cores | — | 640 |
Taux de pixel | 16.00 GPixel/s | 176.6 GPixel/s |
---|---|---|
Taux de texture | 16.00 GTexel/s | 441.6 GTexel/s |
FP16 (half) performance | 1,024 GFLOPS (2:1) | 28.26 TFLOPS (2:1) |
FP32 (float) performance | 512.0 GFLOPS | 14.13 TFLOPS |
FP64 (double) performance | 16.00 GFLOPS (1:32) | 7.066 TFLOPS (1:2) |
Taille Slot | IGP | Dual-slot |
---|---|---|
TDP | 15 W | 250 W |
Sorties | No outputs | No outputs |
Suggestions pour le PSU | — | 600 W |
Connecteurs de puissance | — | 2x 8-pin |
Board number | — | PG500 SKU 200 |
DirectX | 12 (12_1) | 12 (12_1) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 1.2 | 3.0 |
Vulkan | 1.1 | 1.2 |
CUDA | 5.3 | 7.0 |
Shader model | 6.4 | 6.6 |
Date de publication | — | Mar 27th, 2018 |
---|---|---|
Génération | — | Tesla |
Production | — | End-of-life |
Bus interface | — | PCIe 3.0 x16 |