Product Compare
NVIDIA GeForce RTX 3090 vs NVIDIA Rubin GPU
| GPU | GPU | |
|---|---|---|
| Graphics Card | ||
| Release Date | Sep 2020 | Jan 2026 |
| Generation | GeForce 30 | Server Rubin |
| Type | Desktop | Desktop |
| Bus Interface | PCIe 4.0 x16 | PCIe 6.0 x16 |
| Clock Speeds | ||
| Base Clock | 1395 MHz | 700 MHz |
| Boost Clock | 1695 MHz | 2267 MHz |
| Memory Clock | 1219 MHz | 2695 MHz |
| Memory | ||
| Memory Size | 24GB | 288GB |
| Memory Type | GDDR6X | HBM4 |
| Memory Bus | 384bit | 16384bit |
| Bandwidth | 936.2GB/s | 22.1TB/s |
| Render Config | ||
| SM Count | 82 | 224 |
| Shading Units | 10496 | 28672 |
| TMUs | 328 | 896 |
| ROPs | 112 | 24 |
| Tensor Cores | 328 | 896 |
| RT Cores | 82 | |
| L1 Cache | 128 KB (per SM) | 256 KB (per SM) |
| L2 Cache | 6 MB | 128 MB |
| Theoretical Performance | ||
| Pixel Rate | 189.8 GPixel/s | 54.41 GPixel/s |
| Texture Rate | 556.0 GTexel/s | 2031.2 GTexel/s |
| FP16 (half) | 35.58 TFLOPS | 260.0 TFLOPS |
| FP32 (float) | 35.58 TFLOPS | 130.0 TFLOPS |
| FP64 (double) | 556.0 GFLOPS | 32.50 TFLOPS |
| Graphics Processor | ||
| GPU Name | GA102 | GR100 |
| GPU Variant | GA102-300-A1 | |
| Architecture | Ampere | Rubin |
| Foundry | Samsung | TSMC |
| Process Size | 8 nm | 3 nm |
| Transistors | 28.3 billion | 336 billion |
| Die Size | 628 mm² | 1456 mm² |
| Board Design | ||
| TDP | 350W | 2300W |
| Suggested PSU | 750 W | 2700 W |
| Outputs | 1x HDMI 2.13x DisplayPort 1.4a | No outputs |
| Power Connectors | 1x 12-pin | |
| Graphics Features | ||
| DirectX | 12 Ultimate (12_2) | N/A |
| OpenGL | 4.6 | N/A |
| OpenCL | 3.0 | 3.0 |
| Vulkan | 1.3 | N/A |
| CUDA | 8.6 | 10.7 |
| Shader Model | 6.6 | N/A |