GPU Name | Lexa | TU102 |
---|---|---|
GPU Variante | Lexa PRO GL | TU102-875-A1 |
Architektur | GCN 4.0 | Turing |
Hersteller | GlobalFoundries | TSMC |
Prozessgröße | 14 nm | 12 nm |
Transistoren | 2,200 million | 18,600 million |
Chipgröße | 103 mm² | 754 mm² |
Veröffentlichungsdatum | Jun 4th, 2017 | Unknown |
---|---|---|
Generation | Radeon Pro | GRID |
Produktion | End-of-life | Active |
Originalpreis | 149 USD | — |
Businterface | PCIe 3.0 x8 | PCIe 3.0 x16 |
Basis-Taktfrequenz | 925 MHz | 1065 MHz |
---|---|---|
Boost-Takt | 1219 MHz | 1395 MHz |
Speicher-Taktung | 1500 MHz 6 Gbps effective | 1575 MHz 12.6 Gbps effective |
Speichergröße | 2 GB | 16 GB |
---|---|---|
Speichertyp | GDDR5 | GDDR6 |
Speicherbus | 64 bit | 384 bit |
Bandbreite | 48.00 GB/s | 604.8 GB/s |
Shading-Einheiten | 512 | 4608 |
---|---|---|
TMUs | 32 | 288 |
ROPs | 16 | 96 |
Compute Units | 8 | — |
Cache L1 | 16 KB (per CU) | 64 KB (per SM) |
Cache L2 | 256 KB | 6 MB |
Anzahl SM | — | 72 |
Tensor Cores | — | 576 |
RT-Kerne | — | 72 |
Pixelrate | 19.50 GPixel/s | 133.9 GPixel/s |
---|---|---|
Texturrate | 39.01 GTexel/s | 401.8 GTexel/s |
FP16 (half) Leistung | 1,248 GFLOPS (1:1) | 25.71 TFLOPS (2:1) |
FP32 (float) Leistung | 1,248 GFLOPS | 12.86 TFLOPS |
FP64 (double) Leistung | 78.02 GFLOPS (1:16) | 401.8 GFLOPS (1:32) |
Steckplatzbreite | Single-slot | Dual-slot |
---|---|---|
Länge | 168 mm 6.6 inches | 267 mm 10.5 inches |
Breite | 69 mm 2.7 inches | — |
TDP | 35 W | 260 W |
Vorgeschlagenes Netzteil | 200 W | 600 W |
Ausgabe | 1x DisplayPort2x mini-DisplayPort | No outputs |
Stromverbindung | None | 1x 6-pin + 1x 8-pin |
Board-Nummer | D091 | PG150 SKU 215 |
DirectX | 12 (12_0) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 2.1 | 3.0 |
Vulkan | 1.2 | 1.2 |
Shader-Modell | 6.4 | 6.6 |
CUDA | — | 7.5 |