| GPU Name: | GB202-300-A1 |
| Architecture: | Blackwell 2.0 |
| Process Size: | 5 nm (TSMC 4N FinFET) |
| Transistors: | 92,200 million |
| Die Size: | 750 mm² |
| Shading Units: | 21760 |
| TMUs: | 680 |
| ROPs: | 176 |
| Tensor Cores: | 680 |
| RT Cores: | 170 |
| Base Clock: | 2017 MHz |
| Boost Clock: | 2437 MHz (2407 MHz standard boost) |
| L1 Cache: | 128 KB (per SM) |
| L2 Cache: | 96 MB |
| Memory Size: | 32 GB |
| Memory Type: | GDDR7 |
| Memory Clock: | 1750 MHz (28 Gbps effective) |
| Memory Bus: | 512-bit |
| Bandwidth: | 1.79 TB/s |
| Bus Interface: | PCIe 5.0 x16 |
| Slot Width: | Triple-slot (physical width of 67mm) |
| Length: | 325 mm (12.8 inches) |
| Height: | 139 mm (5.5 inches) |
| Width: | 67 mm (2.6 inches) |
| Weight: | 1875 g |
| TDP (Thermal Design Power): | 575 W |
| Suggested PSU: | 950 W (NVIDIA recommends 1000W for the RTX 5090) |
| Power Connectors: | 1x 16-pin (12V-2x6) |
| Display Outputs: | 1x HDMI 2.1b, 3x DisplayPort 2.1b |
| Multi-Monitor Support: | 4 |
| Max Resolution: | 7680 x 4320 |
| DirectX: | 12 Ultimate (12_2) |
| OpenGL: | 4.6 |
| OpenCL: | 3.0 |
| Vulkan: | 1.4 |
| CUDA: | 12.0 |
| Shader Model: | 6.8 |
| HDCP Ready: | Yes |
