GPU model | Lockhart | TU106 |
---|---|---|
Architecture | RDNA 2.0 | Turing |
Foundry | TSMC | TSMC |
Manufacturing process | 7 nm | 12 nm |
Number of transistors | 8,000 million | 10,800 million |
Die size | 197 mm² | 445 mm² |
GPU variant | — | TU106-200A-KA-A1 |
Release date | Nov 10th, 2020 | Jan 7th, 2019 |
---|---|---|
Generation | Console GPU | GeForce 20 |
Production | Active | Active |
Launch price | 299 USD | 349 USD |
Predecessor | — | GeForce 10 |
Interface | — | PCIe 3.0 x16 |
Reviews | — | 85 in our database |
Successor | — | GeForce 30 |
GPU clock | 1565 MHz | — |
---|---|---|
Memory clock | 1750 MHz 14 Gbps effective | 1750 MHz 14 Gbps effective |
Base clock | — | 1365 MHz |
Boost clock | — | 1680 MHz |
Memory size | 8 GB | 6 GB |
---|---|---|
Memory type | GDDR6 | GDDR6 |
Memory bus width | 128 bit | 192 bit |
Bandwidth | 224.0 GB/s | 336.0 GB/s |
Shading units | 1280 | 1920 |
---|---|---|
TMUs | 80 | 120 |
ROPs | 32 | 48 |
Compute units | 20 | — |
Cache L2 | 4 MB | 3 MB |
SM count | — | 30 |
Tensor cores | — | 240 |
RT cores | — | 30 |
Cache L1 | — | 64 KB (per SM) |
Pixel rate | 50.08 GPixel/s | 80.64 GPixel/s |
---|---|---|
Texture rate | 125.2 GTexel/s | 201.6 GTexel/s |
FP16 (half) performance | 8.013 TFLOPS (2:1) | 12.90 TFLOPS (2:1) |
FP32 (float) performance | 4.006 TFLOPS | 6.451 TFLOPS |
FP64 (double) performance | 250.4 GFLOPS (1:16) | 201.6 GFLOPS (1:32) |
Length | 274 mm 10.8 inches | 229 mm 9 inches |
---|---|---|
Width | 151 mm 5.9 inches | 113 mm 4.4 inches |
Height | 64 mm 2.5 inches | 35 mm 1.4 inches |
Weight | 1.93 kg (4.25 lbs) | — |
TDP | 100 W | 160 W |
Outputs | No outputs | 1x DVI1x HDMI2x DisplayPort1x USB Type-C |
Slot width | — | Dual-slot |
Suggested PSU | — | 450 W |
Power connectors | — | 1x 8-pin |
Board number | — | PG160 SKU 42 |
DirectX | 12 Ultimate (12_2) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 1.2 | 3.0 |
Vulkan | 1.2 | 1.2 |
Shader model | 6.5 | 6.6 |
CUDA | — | 7.5 |