GPU Name | GA102 | GM20B |
---|---|---|
GPU Variante | GA102-890-A1 | TM670D-A1 |
Architektur | Ampere | Maxwell 2.0 |
Hersteller | Samsung | TSMC |
Prozessgröße | 8 nm | 20 nm |
Transistoren | 28,300 million | 2,000 million |
Chipgröße | 628 mm² | 118 mm² |
Veröffentlichungsdatum | Apr 12th, 2021 | — |
---|---|---|
gpu.details.availability | 2021 | — |
Generation | Tesla | — |
Produktion | Active | — |
Businterface | PCIe 4.0 x16 | — |
Basis-Taktfrequenz | 885 MHz | — |
---|---|---|
Boost-Takt | 1695 MHz | — |
Speicher-Taktung | 1563 MHz 12.5 Gbps effective | System Shared |
GPU-Taktung | — | 1000 MHz |
Speichergröße | 24 GB | System Shared |
---|---|---|
Speichertyp | GDDR6 | System Shared |
Speicherbus | 384 bit | System Shared |
Bandbreite | 600.2 GB/s | System Dependent |
Shading-Einheiten | 9216 | 256 |
---|---|---|
TMUs | 288 | 16 |
ROPs | 96 | 16 |
Anzahl SM | 72 | — |
Tensor Cores | 288 | — |
RT-Kerne | 72 | — |
Cache L1 | 128 KB (per SM) | 48 KB (per SMM) |
Cache L2 | 6 MB | 256 KB |
Anzahl SMM | — | 2 |
Pixelrate | 162.7 GPixel/s | 16.00 GPixel/s |
---|---|---|
Texturrate | 488.2 GTexel/s | 16.00 GTexel/s |
FP16 (half) Leistung | 31.24 TFLOPS (1:1) | 1,024 GFLOPS (2:1) |
FP32 (float) Leistung | 31.24 TFLOPS | 512.0 GFLOPS |
FP64 (double) Leistung | 976.3 GFLOPS (1:32) | 16.00 GFLOPS (1:32) |
Steckplatzbreite | Single-slot | IGP |
---|---|---|
Länge | 267 mm 10.5 inches | — |
Breite | 112 mm 4.4 inches | — |
TDP | 150 W | 15 W |
Vorgeschlagenes Netzteil | 450 W | — |
Ausgabe | No outputs | No outputs |
Stromverbindung | 8-pin EPS | — |
Board-Nummer | PG133 SKU 215 | — |
DirectX | 12 Ultimate (12_2) | 12 (12_1) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 3.0 | 1.2 |
Vulkan | 1.2 | 1.1 |
CUDA | 8.6 | 5.3 |
Shader-Modell | 6.6 | 6.4 |
Veröffentlichungsdatum | — | Jan 4th, 2015 |
---|---|---|
Generation | — | Tegra |
Produktion | — | End-of-life |
Businterface | — | IGP |