GPU model | Mach32 | TU102 |
---|---|---|
GPU variant | Mach32-03 (2106880003) | TU102-875-A1 |
Architecture | Mach | Turing |
Manufacturing process | 700 nm | 12 nm |
Number of transistors | unknown | 18,600 million |
Die size | 90 mm² | 754 mm² |
Foundry | — | TSMC |
Release date | 1992 | Unknown |
---|---|---|
Generation | Mach 32 | GRID |
Production | End-of-life | Active |
Launch price | 799 USD | — |
Interface | PCI | PCIe 3.0 x16 |
Predecessor | Mach 8 | — |
Successor | Mach 64 | — |
GPU clock | 10 MHz | — |
---|---|---|
Memory clock | 10 MHz | 1750 MHz 14 Gbps effective |
Base clock | — | 1065 MHz |
Boost clock | — | 1395 MHz |
Memory size | 2 MB | 8 GB |
---|---|---|
Memory type | DRAM | GDDR6 |
Memory bus width | 64 bit | 384 bit |
Bandwidth | 80.00 MB/s | 672.0 GB/s |
Pixel shaders | 1 | — |
---|---|---|
Vertex shaders | 0 | — |
TMUs | 0 | 288 |
ROPs | 1 | 96 |
Shading units | — | 4608 |
SM count | — | 72 |
Tensor cores | — | 576 |
RT cores | — | 72 |
Cache L1 | — | 64 KB (per SM) |
Cache L2 | — | 6 MB |
Pixel rate | 10.00 MPixel/s | 133.9 GPixel/s |
---|---|---|
Texture rate | 0 MTexel/s | 401.8 GTexel/s |
FP16 (half) performance | — | 25.71 TFLOPS (2:1) |
FP32 (float) performance | — | 12.86 TFLOPS |
FP64 (double) performance | — | 401.8 GFLOPS (1:32) |
Slot width | Single-slot | Dual-slot |
---|---|---|
TDP | unknown | 260 W |
Suggested PSU | 200 W | 600 W |
Outputs | 1x VGA | No outputs |
Board number | 193 | PG150 SKU 220 |
Length | — | 267 mm 10.5 inches |
Power connectors | — | 1x 6-pin + 1x 8-pin |
DirectX | — | 12 Ultimate (12_2) |
---|---|---|
OpenGL | — | 4.6 |
OpenCL | — | 3.0 |
Vulkan | — | 1.2 |
Pixel shader | — | — |
Vertex shader | — | — |
CUDA | — | 7.5 |
Shader model | — | 6.6 |