RTX 5090 vs RTX 5090 D for Blender
Compare RTX 5090 and RTX 5090 D for Blender rendering. See benchmark scores, VRAM, memory bandwidth, power draw, and estimated OptiX/CUDA render times side by side on Renderjuice. RTX 5090 leads the Blender benchmark score by 0.6%.
Fastest in Blender
RTX 5090 (15,016.75 score)
| Spec | ||
|---|---|---|
| Performance | ||
| Blender benchmark scoreBlender benchmark scoreThis score is derived from Blender Open Data benchmark samples. It is useful for comparing GPUs against each other, but it is still a benchmark summary rather than a guarantee for every real project. | 15,016.75 | 14,931.27 -1% |
| CUDA coresCUDA coresMore CUDA cores generally means the GPU can process more rendering work in parallel. However, core count alone does not determine performance — architecture, clock speed, and memory bandwidth all play a role. | 21,760 | 21,760 |
| Boost clockBoost clockBoost clock is one of the signals of peak GPU speed, but it does not define Blender performance on its own. Architecture and memory behavior matter too. | 2407 MHz | 2407 MHz |
| RT coresRay tracingHardware ray tracing support can speed up tasks that rely on realistic light transport, reflections, shadows, and path tracing. | RT cores: 170 | RT cores: 170 |
| Tensor coresTensor coresIn Blender, tensor cores primarily accelerate the OptiX AI denoiser, which can clean up noisy renders much faster than traditional denoising. They also help with AI-powered features like DLSS in the viewport. | Tensor cores: 680 | Tensor cores: 680 |
| Memory | ||
| VRAMVRAMIf a scene does not fit into VRAM, rendering can slow down sharply or fail depending on the renderer and setup. More VRAM gives you more room for large scenes and high-resolution assets. | 32 GB | 32 GB |
| Memory bandwidthMemory bandwidthBandwidth is usually shown in GB/s. Higher bandwidth helps when the renderer needs to move lots of texture, geometry, or shading data quickly. | 1,790 GB/s | 1,790 GB/s |
| Memory typeMemory typeYou will often see names like GDDR6 or GDDR6X. In practice, memory type matters less on its own than overall memory bandwidth and total VRAM. | GDDR7 | GDDR7 |
| Memory busMemory busA wider bus can move more data at once, but real performance also depends on memory speed and architecture. Bus width alone does not tell the full story. | 512-bit | 512-bit |
| Platform | ||
| ArchitectureArchitectureNames like Ampere or Ada Lovelace refer to different GPU designs. Newer architectures often bring better ray tracing, denoising, efficiency, and rendering throughput. | Blackwell 2.0 | Blackwell 2.0 |
| Render supportRender supportNVIDIA GPUs typically support CUDA (the general compute path) and OptiX (a ray-tracing-optimized path that is usually faster on RTX cards). Some older GPUs may only support one of these. The best backend depends on the GPU generation. | OptiX, CUDA | OptiX |
| TDPTDPHigher TDP usually means the GPU can sustain more performance, but it also means more heat and stronger PSU and cooling requirements. | 575 W | 575 W |
| Release year | 2025 | 2025 |
Benchmark comparison
Estimated seconds to render one frame of each standard Blender benchmark scene. Lower is faster.
These timings are derived from Blender Open Data benchmark medians and should be treated as comparative estimates, not guaranteed real-project render times.
Quick take on RTX 5090 vs RTX 5090 D
RTX 5090 leads the Blender benchmark score by 0.6%.
Fastest in Blender: RTX 5090 15,016.75 score.
Both GPUs ship with 32 GB of VRAM, so the tradeoff is more about speed, architecture, and efficiency than memory capacity.
If you are deciding between these cards for Blender, focus first on Blender benchmark score, VRAM capacity, memory bandwidth, and whether your scenes are likely to benefit more from raw speed or extra memory headroom. The comparison table above keeps those tradeoffs in one place.