GPU model | GA102 | TU104 |
---|---|---|
GPU variant | GA102-300-A1 | TU104-895-A1 |
Architecture | Ampere | Turing |
Foundry | TSMC | TSMC |
Manufacturing process | 7 nm | 12 nm |
Number of transistors | 40,000 million | 13,600 million |
Die size | 627 mm² | 545 mm² |
Release date | 2020 | Sep 13th, 2018 |
---|---|---|
Generation | GeForce 30 | Tesla |
Production | Unreleased | Active |
Launch price | 1,399 USD | — |
Interface | PCIe 4.0 x16 | PCIe 3.0 x16 |
Base clock | 1410 MHz | 585 MHz |
---|---|---|
Boost clock | 1740 MHz | 1590 MHz |
Memory clock | 2375 MHz 19000 MHz effective | 1250 MHz 10 Gbps effective |
Memory size | 12 GB | 16 GB |
---|---|---|
Memory type | GDDR6X | GDDR6 |
Memory bus width | 384 bit | 256 bit |
Bandwidth | 912.0 GB/s | 320.0 GB/s |
Shading units | 5120 | 2560 |
---|---|---|
TMUs | 320 | 160 |
ROPs | 96 | 64 |
SM count | 80 | 40 |
Tensor cores | 640 | 320 |
RT cores | 160 | 40 |
Cache L1 | 192 KB (per SM) | 64 KB (per SM) |
Cache L2 | 6 MB | 4 MB |
Pixel rate | 167.0 GPixel/s | 101.8 GPixel/s |
---|---|---|
Texture rate | 556.8 GTexel/s | 254.4 GTexel/s |
FP16 (half) performance | 35.64 TFLOPS (2:1) | 65.13 TFLOPS (8:1) |
FP32 (float) performance | 17.82 TFLOPS | 8.141 TFLOPS |
FP64 (double) performance | 556.8 GFLOPS (1:32) | 254.4 GFLOPS (1:32) |
Slot width | Dual-slot | Single-slot |
---|---|---|
Length | 267 mm 10.5 inches | 168 mm 6.6 inches |
TDP | 320 W | 70 W |
Suggested PSU | 700 W | 250 W |
Outputs | 1x HDMI3x DisplayPort1x USB Type-C | No outputs |
Power connectors | 1x 12-pin | None |
Board number | PG132 SKU 20 | PG183 SKU 200 |
DirectX | 12 Ultimate (12_2) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 2.0 | 3.0 |
Vulkan | 1.2.140 | 1.2 |
CUDA | 8.5 | 7.5 |
Shader model | 6.5 | 6.6 |