NVIDIA GeForce RTX 5090
- Graphics Processor: GB202
- TMUs: 680
- Memory Type: GDDR7
- Bus Width: 512 bit
- ROPs: 192
NVIDIA GeForce RTX 5090 Benchmarks
The NVIDIA GeForce RTX 5090 is powered by a single 16-pin power connection, which has a maximum power demand rating of 500 W. Among the display outputs are three DisplayPort 1.4a and one HDMI 2.1. An interface of PCI-Express 5.0 x16 connects the GeForce RTX 5090 to the rest of the system. With a triple-slot cooling system, the card measures 304 mm by 137 mm by 61 mm.
NVIDIA is planning to release the GeForce RTX 5090 graphics card in 2025. Based on the GB202 graphics processor and built using an unidentified process, the GB202-400-A1 model of the card supports DirectX 12 Ultimate. This guarantees that GeForce RTX 5090 will run all contemporary games.
There are 170 raytracing acceleration cores on the GPU as well. Using a 512-bit memory interface, NVIDIA has linked 32 GB of GDDR7 memory with the GeForce RTX 5090. Memory is running at 1750 MHz (22.2 Gbps effective), while the GPU is functioning at 2235 MHz, which may be increased to 2520 MHz, being a card with three slots.
Specifications
BASIC GPU INFO
GPU name | GB202 |
Manufacturer | NVIDIA |
Released date | 2025 |
Launch price | N/A |
Purpose | gaming |
Architecture | Blackwell 2.0 |
Generation | GeForce 50 |
Successor | N/A |
Predecessor | GeForce 40 |
BOARD DESIGN
TDP | 500 W |
Slot width | Triple-slot |
Power connectors | 1x 16-pin |
Length | 304 mm 12 inches |
Height | 61 mm 2.4 inches |
Width | 137 mm 5.4 inches |
Outputs | 1x HDMI 2.1 3x DisplayPort 1.4a |
Board number | PG144 SKU 30 |
Weight | N/A |
Bus interface | PCIe 5.0 x16 |
GRAPHICS PROCESSING UNIT
Base clock | 2235 MHz |
Boost clock | 2520 MHz |
Foundry | TSMC |
Transistors | N/A |
Process type | N/A |
MCD Die density | N/A |
GPU MEMORY
Memory type | GDDR7 |
Memory size | 32 GB |
Memory bus | 512 bit |
Bandwidth | 1.42 TB/s |
Memory clock | 1750 MHz 22.2 Gbps |
Memory speed | N/A |
GRAPHICS FEATURES (API)
Vulkan | 1.3 |
DirectX | 12 Ultimate (12_2) |
OpenGL | 4.6 |
OpenCL | 3.0 |
Shader model | 6.8 |
CUDA | 10.1 |
CONFIGURATION
ROPs | 192 |
Shading units | 21760 |
TMUs | 680 |
RT Cores | 170 |
Compute units | N/A |
L1 cache | 128 KB (per SM) |
L2 cache | 88 MB |
GPU PERFORMANCE
Vertex rate | N/A |
Pixel rate | 483.8 GPixel/s |
Texture rate | 1,714 GTexel/s |
FP16 (half) | 109.7 TFLOPS (1:1) |
FP32 (float) | 109.7 TFLOPS |
FP64 (double) | 1.714 TFLOPS (1:64) |
Share your thoughts
Disclaimer Note
We can not guarantee that the information on this page is 100% correct.