Nom du GPU | Juniper | GA104 |
---|---|---|
GPU variant | Juniper LE (215-0754009) | GA104-200-A1 |
Architecture | TeraScale 2 | Ampere |
Fonderie | TSMC | Samsung |
Taille du processus | 40 nm | 8 nm |
Transistors | 1,040 million | 17,400 million |
Taille des matrices | 166 mm² | 392 mm² |
Date de publication | Jan 21st, 2011 | Dec 1st, 2020 |
---|---|---|
Génération | Northern Islands | GeForce 30 |
Production | End-of-life | Active |
Bus interface | PCIe 2.0 x16 | PCIe 4.0 x16 |
Critiques | 4 in our database | 64 in our database |
Prédécesseur | Evergreen | GeForce 20 |
Successeur | Southern Islands | — |
gpu.details.availability | — | Dec 2nd, 2020 |
Prix de lancement | — | 399 USD |
GPU clock | 700 MHz | — |
---|---|---|
Mémoire clock | 1150 MHz 4.6 Gbps effective | 1750 MHz 14 Gbps effective |
Base clock | — | 1410 MHz |
Boost clock | — | 1665 MHz |
Taille de la mémoire | 1024 MB | 8 GB |
---|---|---|
Type de mémoire | GDDR5 | GDDR6 |
Bus mémoire | 128 bit | 256 bit |
Bande passante | 73.60 GB/s | 448.0 GB/s |
Shading units | 720 | 4864 |
---|---|---|
TMUs | 36 | 152 |
ROPs | 16 | 80 |
Compute units | 9 | — |
Cache L1 | 8 KB (per CU) | 128 KB (per SM) |
Cache L2 | 256 KB | 4 MB |
Comptage SM | — | 38 |
Tensor cores | — | 152 |
RT cores | — | 38 |
Taux de pixel | 11.20 GPixel/s | 133.2 GPixel/s |
---|---|---|
Taux de texture | 25.20 GTexel/s | 253.1 GTexel/s |
FP32 (float) performance | 1,008 GFLOPS | 16.20 TFLOPS |
FP16 (half) performance | — | 16.20 TFLOPS (1:1) |
FP64 (double) performance | — | 253.1 GFLOPS (1:64) |
Taille Slot | Dual-slot | Dual-slot |
---|---|---|
Longueur | 170 mm 6.7 inches | 242 mm 9.5 inches |
TDP | 86 W | 200 W |
Suggestions pour le PSU | 250 W | 550 W |
Sorties | 2x DVI1x HDMI2x mini-DisplayPort | 1x HDMI3x DisplayPort |
Connecteurs de puissance | 1x 6-pin | 1x 12-pin |
Board number | C012 | PG142 SKU 20 |
Taille | — | 112 mm 4.4 inches |
DirectX | 11.2 (11_0) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.4 | 4.6 |
OpenCL | 1.2 | 3.0 |
Vulkan | — | 1.2 |
Shader model | 5.0 | 6.6 |
CUDA | — | 8.6 |