GPU model | TU104 | GA104 |
---|---|---|
GPU variant | TU104-450-A1 | GA104-150-A1 |
Architecture | Turing | Ampere |
Foundry | TSMC | Samsung |
Manufacturing process | 12 nm | 8 nm |
Number of transistors | 13,600 million | 17,400 million |
Die size | 545 mm² | 392 mm² |
Release date | Jul 23rd, 2019 | Sep 1st, 2021 |
---|---|---|
Generation | GeForce 20 | GeForce 30 |
Predecessor | GeForce 10 | GeForce 20 |
Production | Active | Active |
Launch price | 699 USD | 329 USD |
Current price | Amazon / Newegg | — |
Interface | PCIe 3.0 x16 | PCIe 4.0 x16 |
Reviews | 51 in our database | 60 in our database |
Successor | GeForce 30 | — |
Base clock | 1650 MHz | 1320 MHz |
---|---|---|
Boost clock | 1815 MHz | 1777 MHz |
Memory clock | 1937 MHz 15.5 Gbps effective | 1875 MHz 15 Gbps effective |
Memory size | 8 GB | 12 GB |
---|---|---|
Memory type | GDDR6 | GDDR6 |
Memory bus width | 256 bit | 192 bit |
Bandwidth | 495.9 GB/s | 360.0 GB/s |
Shading units | 3072 | 3584 |
---|---|---|
TMUs | 192 | 112 |
ROPs | 64 | 48 |
SM count | 48 | 28 |
Tensor cores | 384 | 112 |
RT cores | 48 | 28 |
Cache L1 | 64 KB (per SM) | 128 KB (per SM) |
Cache L2 | 4 MB | 3 MB |
Pixel rate | 116.2 GPixel/s | 85.30 GPixel/s |
---|---|---|
Texture rate | 348.5 GTexel/s | 199.0 GTexel/s |
FP16 (half) performance | 22.30 TFLOPS (2:1) | 12.74 TFLOPS (1:1) |
FP32 (float) performance | 11.15 TFLOPS | 12.74 TFLOPS |
FP64 (double) performance | 348.5 GFLOPS (1:32) | 199.0 GFLOPS (1:64) |
Slot width | Dual-slot | Dual-slot |
---|---|---|
Length | 267 mm 10.5 inches | 242 mm 9.5 inches |
Width | 116 mm 4.6 inches | 112 mm 4.4 inches |
Height | 35 mm 1.4 inches | — |
TDP | 250 W | 170 W |
Suggested PSU | 600 W | 450 W |
Outputs | 1x HDMI3x DisplayPort1x USB Type-C | 1x HDMI3x DisplayPort |
Power connectors | 1x 6-pin + 1x 8-pin | 1x 12-pin |
Board number | PG181 | — |
DirectX | 12 Ultimate (12_2) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 3.0 | 3.0 |
Vulkan | 1.2 | 1.2 |
CUDA | 7.5 | 8.6 |
Shader model | 6.6 | 6.6 |