Nom du GPU | GA102 | GA102 |
---|---|---|
GPU variant | GA102-300-A1 | GA102-350-A1 |
Architecture | Ampere | Ampere |
Fonderie | TSMC | Samsung |
Taille du processus | 7 nm | 8 nm |
Transistors | 40,000 million | 28,300 million |
Taille des matrices | 627 mm² | 628 mm² |
Date de publication | 2020 | Unknown |
---|---|---|
Génération | GeForce 30 | GeForce 30 |
Production | Unreleased | Unreleased |
Prix de lancement | 1,399 USD | 1,699 USD |
Bus interface | PCIe 4.0 x16 | PCIe 4.0 x16 |
gpu.details.availability | — | 2022 |
Prédécesseur | — | GeForce 20 |
Critiques | — | 55 in our database |
Base clock | 1410 MHz | 1560 MHz |
---|---|---|
Boost clock | 1740 MHz | 1860 MHz |
Mémoire clock | 2375 MHz 19000 MHz effective | 1325 MHz 21.2 Gbps effective |
Taille de la mémoire | 12 GB | 24 GB |
---|---|---|
Type de mémoire | GDDR6X | GDDR6X |
Bus mémoire | 384 bit | 384 bit |
Bande passante | 912.0 GB/s | 1,018 GB/s |
Shading units | 5120 | 10752 |
---|---|---|
TMUs | 320 | 336 |
ROPs | 96 | 112 |
Comptage SM | 80 | 84 |
Tensor cores | 640 | 336 |
RT cores | 160 | 84 |
Cache L1 | 192 KB (per SM) | 128 KB (per SM) |
Cache L2 | 6 MB | 6 MB |
Taux de pixel | 167.0 GPixel/s | 208.3 GPixel/s |
---|---|---|
Taux de texture | 556.8 GTexel/s | 625.0 GTexel/s |
FP16 (half) performance | 35.64 TFLOPS (2:1) | 40.00 TFLOPS (1:1) |
FP32 (float) performance | 17.82 TFLOPS | 40.00 TFLOPS |
FP64 (double) performance | 556.8 GFLOPS (1:32) | 625.0 GFLOPS (1:64) |
Taille Slot | Dual-slot | Triple-slot |
---|---|---|
Longueur | 267 mm 10.5 inches | 336 mm 13.2 inches |
TDP | 320 W | 450 W |
Suggestions pour le PSU | 700 W | 850 W |
Sorties | 1x HDMI3x DisplayPort1x USB Type-C | 1x HDMI3x DisplayPort |
Connecteurs de puissance | 1x 12-pin | 1x 12-pin |
Board number | PG132 SKU 20 | PG132 SKU |
Taille | — | 140 mm 5.5 inches |
Taille | — | 61 mm 2.4 inches |
DirectX | 12 Ultimate (12_2) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 2.0 | 3.0 |
Vulkan | 1.2.140 | 1.2 |
CUDA | 8.5 | 8.6 |
Shader model | 6.5 | 6.6 |