Nom du GPU | GA102 | GP100 |
---|---|---|
GPU variant | GA102-220-A1 | GP100-893-A1 |
Architecture | Ampere | Pascal |
Fonderie | Samsung | TSMC |
Taille du processus | 8 nm | 16 nm |
Transistors | 28,300 million | 15,300 million |
Taille des matrices | 628 mm² | 610 mm² |
Date de publication | Jan 11th, 2022 | Jun 20th, 2016 |
---|---|---|
gpu.details.availability | Jan, 2022 | — |
Génération | GeForce 30 | Tesla |
Prédécesseur | GeForce 20 | — |
Production | Active | End-of-life |
Prix de lancement | 799 USD | 5,699 USD |
Bus interface | PCIe 4.0 x16 | PCIe 3.0 x16 |
Base clock | 1260 MHz | 1190 MHz |
---|---|---|
Boost clock | 1710 MHz | 1329 MHz |
Mémoire clock | 1188 MHz 19 Gbps effective | 715 MHz 1430 Mbps effective |
Taille de la mémoire | 12 GB | 16 GB |
---|---|---|
Type de mémoire | GDDR6X | HBM2 |
Bus mémoire | 384 bit | 4096 bit |
Bande passante | 912.4 GB/s | 732.2 GB/s |
Shading units | 8960 | 3584 |
---|---|---|
TMUs | 280 | 224 |
ROPs | 112 | 96 |
Comptage SM | 70 | 56 |
Tensor cores | 280 | — |
RT cores | 70 | — |
Cache L1 | 128 KB (per SM) | 24 KB (per SM) |
Cache L2 | 5 MB | 4 MB |
Taux de pixel | 191.5 GPixel/s | 127.6 GPixel/s |
---|---|---|
Taux de texture | 478.8 GTexel/s | 297.7 GTexel/s |
FP16 (half) performance | 30.64 TFLOPS (1:1) | 19.05 TFLOPS (2:1) |
FP32 (float) performance | 30.64 TFLOPS | 9.526 TFLOPS |
FP64 (double) performance | 478.8 GFLOPS (1:64) | 4.763 TFLOPS (1:2) |
Taille Slot | Dual-slot | Dual-slot |
---|---|---|
Longueur | 285 mm 11.2 inches | 267 mm 10.5 inches |
Taille | 112 mm 4.4 inches | — |
Taille | 40 mm 1.6 inches | — |
TDP | 350 W | 250 W |
Suggestions pour le PSU | 750 W | 600 W |
Sorties | 1x HDMI3x DisplayPort | No outputs |
Connecteurs de puissance | 1x 12-pin | 1x 8-pin |
DirectX | 12 Ultimate (12_2) | 12 (12_1) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 3.0 | 3.0 |
Vulkan | 1.2 | 1.2 |
CUDA | 8.6 | 6.0 |
Shader model | 6.6 | 6.4 |