GPU model | Loveland | GA102 |
---|---|---|
Architecture | TeraScale 2 | Ampere |
Foundry | TSMC | TSMC |
Manufacturing process | 40 nm | 7 nm |
Number of transistors | 450 million | 40,000 million |
Die size | 75 mm² | 627 mm² |
Release date | Jun 6th, 2012 | — |
---|---|---|
Generation | Wrestler (HD 7000 Mobile) | — |
Production | End-of-life | — |
Interface | IGP | — |
Predecessor | Palm | — |
Successor | Trinity | — |
GPU clock | 523 MHz | — |
---|---|---|
Memory clock | System Shared | 1750 MHz 14000 MHz effective |
Base clock | — | 1110 MHz |
Boost clock | — | 1500 MHz |
Memory size | System Shared | 48 GB |
---|---|---|
Memory type | System Shared | GDDR6 |
Memory bus width | System Shared | 384 bit |
Bandwidth | System Dependent | 672.0 GB/s |
Shading units | 80 | 7552 |
---|---|---|
TMUs | 8 | 472 |
ROPs | 4 | 96 |
Compute units | 2 | — |
SM count | — | 118 |
Tensor cores | — | 472 |
RT cores | — | 118 |
Cache L1 | — | 64 KB (per SM) |
Cache L2 | — | 6 MB |
Pixel rate | 2.092 GPixel/s | 144.0 GPixel/s |
---|---|---|
Texture rate | 4.184 GTexel/s | 708.0 GTexel/s |
FP32 (float) performance | 83.68 GFLOPS | 22.66 TFLOPS |
FP16 (half) performance | — | 45.31 TFLOPS (2:1) |
FP64 (double) performance | — | 708.0 GFLOPS (1:32) |
Slot width | IGP | Dual-slot |
---|---|---|
TDP | 18 W | 260 W |
Outputs | No outputs | 4x DisplayPort1x USB Type-C |
Length | — | 267 mm 10.5 inches |
Suggested PSU | — | 600 W |
Power connectors | — | 1x 6-pin + 1x 8-pin |
DirectX | 11.2 (11_0) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.4 | 4.6 |
OpenCL | 1.2 | 2.0 |
Vulkan | — | 1.2.140 |
Shader model | 5.0 | 6.5 |
CUDA | — | 8.5 |
Release date | — | Unknown |
---|---|---|
Generation | — | Quadro RTX |
Production | — | Active |
Interface | — | PCIe 4.0 x16 |