NVIDIA GeForce RTX 5090 Founders Edition (32GB)
NVIDIA’s flagship Blackwell gaming/creator GPU: 32GB GDDR7, 21,760 CUDA cores.
-Max AI performance with FP4 and NVIDIA DLSS
The NVIDIA GeForce RTX 5090 Founders Edition is built for gamers and creators who refuse to compromise. NVIDIA’s flagship Blackwell-generation consumer GPU pairs the largest GeForce die the company has ever shipped with a Founders Edition cooler engineered to fit it into just two slots. Whether you’re pursuing native 4K gaming at maximum settings, editing 8K footage, or running diffusion models and LLMs locally, the RTX 5090 is designed to deliver performance that lasts for years.
Key Features
-
21,760 CUDA Cores — NVIDIA’s largest-ever GeForce GPU configuration, built on a custom TSMC 4NP process for a substantial generational leap over the RTX 4090.
-
32GB GDDR7 on a 512-bit bus — extensive memory capacity and bandwidth for 4K/8K textures, expansive open-world environments, and local AI models that exceed the limits of a 24GB card.
-
98MB L2 cache — an expanded cache hierarchy that keeps compute cores supplied with data while reducing trips to memory.
-
680 fifth-generation Tensor Cores — up to 3,352 AI TOPS, powering DLSS 4.5 with Multi Frame Generation for dramatically higher frame rates and native-quality image fidelity.
-
170 fourth-generation RT Cores — full-scene path tracing in supported titles without the frame-rate penalties associated with older RT hardware.
-
104.8 TFLOPS FP32 shader compute — a substantial generational increase in raw rasterization and compute throughput over Ada Lovelace.
-
2.41GHz boost clock — high clock speeds despite the enormous core count, enabled by the refined 4NP process node.
-
Compact 2-slot Founders Edition design — flagship performance in a space-efficient footprint suited to SFF and standard ATX builds alike, compared with the triple-slot coolers common on similarly powerful partner cards.
Flagship Gaming, Creation, and Local AI
At native 4K with ray tracing maxed out, the RTX 5090 is currently the fastest consumer GPU on the market by a considerable margin. Its increased CUDA core count over the RTX 4090 — 21,760 versus 16,384 — translates directly into higher sustained frame rates in demanding modern titles, even before DLSS 4.5 and Multi Frame Generation enter the equation. For competitive players using high-refresh 1440p or 4K monitors, the RTX 5090 provides the headroom to keep frame rates close to the display’s ceiling with settings maxed out, offering a performance reserve that less powerful cards cannot guarantee as games become more demanding.
The case is equally compelling for creators. With 32GB of GDDR7, 8K video timelines, multilayer compositions in DaVinci Resolve or Premiere, and large 3D scenes in Blender or Unreal Engine fit comfortably in VRAM. NVIDIA’s NVENC/NVDEC hardware, backed by AV1 support, also significantly reduces export and streaming times compared with CPU-only workflows.
For local AI work — including Stable Diffusion, fine-tuning smaller LLMs, and inference on quantized 30B-class models — 32GB can make the difference between keeping a model in VRAM and repeatedly spilling into system RAM. It cannot match a dedicated 96GB workstation card for the largest models, but for local generative AI experimentation alongside gaming and creative work, no other consumer GPU combines this much memory and compute in a single card.
The Founders Edition pairs NVIDIA’s vapor-chamber cooling with a flow-through design, tested and validated alongside the card’s power delivery. The result is a distinct thermal solution that is typically quieter and more consistent than many third-party coolers handling a 575W flagship. With total graphics power rated at 575W, a capable 1000W PSU and a case with strong airflow are essential. In return, you get the most capable GPU NVIDIA has ever offered under the GeForce name.
Specifications
- Architecture: NVIDIA Blackwell, TSMC 4NP
- CUDA Cores: 21,760
- Tensor Cores: 680 (5th-generation)
- RT Cores: 170 (4th-generation)
- Memory: 32GB GDDR7, 512-bit bus, 98MB L2 cache
- FP32 Compute: ~104.8 TFLOPS
- AI Performance: Up to 3,352 AI TOPS
- Boost Clock: 2.41GHz
- Total Graphics Power: 575W
- Power Connector: 1x 16-pin (12V-2×6)
- Interface: PCIe 5.0
- Display Outputs: DisplayPort 2.1b, up to 8K @ 165Hz
- Recommended PSU: 1000W
NVIDIA® GeForce RTX™ 50 Series GPUs are game-changing for gamers and creators.