NVIDIA GeForce RTX 5090
GeForce RTX 50
The fastest gaming card NVIDIA makes, and the only GeForce with 32 GB of memory, which is what makes it the local-AI card: it runs models a 16 GB card cannot hold. In Tom's Hardware's testing it leads the RTX 5080 by about 44% in Cyberpunk 2077 at 1440p, but it draws up to 575 W, needs a 1000 W power supply and has rarely sold near its $1,999 launch price.
Against the line-up
Cyberpunk 2077
Cyberpunk 2077
Cyberpunk 2077
Cyberpunk 2077
Alan Wake 2
Black Myth: Wukong
Counter-Strike 2
Fortnite
Local LLM (llama.cpp / Ollama)
Local LLM (llama.cpp / Ollama)
Blender Benchmark
The whole Graphics cards line-up side by side, with prices
Every benchmark
Quoted from the publications that ran them; we don't test hardware ourselves. Figures from different publications were run on different systems and scenes, so compare within a row's source.
Cyberpunk 2077
| Setting | Result | Source |
|---|---|---|
| 1080p Ultra | 157fps | Tom's Hardware, Jun 24, 2026 Ryzen 7 9800X3D, 32GB DDR5-6000 CL30, Asus TUF X670E-Plus (Tom's Hardware 2026 GPU testbed) Tom's Hardware GPU benchmarks hierarchy, per-game chart (average fps, native, no upscaling) |
| 237fps | KitGuru, Jan 29, 2025 Ryzen 7 9800X3D (KitGuru 2025 GPU test system) Card tested: RTX 5090 Founders Edition. Average fps. |
|
| 1440p Ultra | 135fps | Tom's Hardware, Jun 24, 2026 Ryzen 7 9800X3D, 32GB DDR5-6000 CL30, Asus TUF X670E-Plus (Tom's Hardware 2026 GPU testbed) Tom's Hardware GPU benchmarks hierarchy, per-game chart (average fps, native, no upscaling) |
| 190fps | KitGuru, Jan 29, 2025 Ryzen 7 9800X3D (KitGuru 2025 GPU test system) Card tested: RTX 5090 Founders Edition. Average fps. |
|
| 1440p Ultra + Ultra RT |
92.1fps | Tom's Hardware, Jun 24, 2026 Ryzen 7 9800X3D, 32GB DDR5-6000 CL30, Asus TUF X670E-Plus (Tom's Hardware 2026 GPU testbed) Tom's Hardware GPU hierarchy, ray tracing chart (average fps, native, no upscaling) |
| 4K Ultra | 61.5fps | Tom's Hardware, Jun 24, 2026 Ryzen 7 9800X3D, 32GB DDR5-6000 CL30, Asus TUF X670E-Plus (Tom's Hardware 2026 GPU testbed) Tom's Hardware GPU benchmarks hierarchy, per-game chart (average fps, native, no upscaling) |
| 101fps | KitGuru, Jan 29, 2025 Ryzen 7 9800X3D (KitGuru 2025 GPU test system) Card tested: RTX 5090 Founders Edition. Average fps. |
|
| 4K Ultra + Path Tracing, DLSS Performance |
84.4fps | KitGuru, Jan 23, 2025 Ryzen 7 9800X3D (KitGuru 2025 GPU test system) Card tested: RTX 5090 Founders Edition. KitGuru: Ultra preset, path tracing, Ray Reconstruction on, DLSS Super Resolution Performance, frame generation off. |
Alan Wake 2
| Setting | Result | Source |
|---|---|---|
| 1440p Ultra |
107fps | Tom's Hardware, Jun 24, 2026 Ryzen 7 9800X3D, 32GB DDR5-6000 CL30, Asus TUF X670E-Plus (Tom's Hardware 2026 GPU testbed) Tom's Hardware GPU benchmarks hierarchy, per-game chart (average fps, native, no upscaling) |
Black Myth: Wukong
| Setting | Result | Source |
|---|---|---|
| 1440p Very High |
99.4fps | Tom's Hardware, Jun 24, 2026 Ryzen 7 9800X3D, 32GB DDR5-6000 CL30, Asus TUF X670E-Plus (Tom's Hardware 2026 GPU testbed) Tom's Hardware GPU benchmarks hierarchy, per-game chart (average fps, native, no upscaling) |
Counter-Strike 2
| Setting | Result | Source |
|---|---|---|
| 1440p High |
623fps | Tom's Hardware, Jun 24, 2026 Ryzen 7 9800X3D, 32GB DDR5-6000 CL30, Asus TUF X670E-Plus (Tom's Hardware 2026 GPU testbed) Tom's Hardware GPU benchmarks hierarchy, per-game chart (average fps, native, no upscaling) |
Fortnite
| Setting | Result | Source |
|---|---|---|
| 1440p High (Lumen + Nanite High) |
141fps | Tom's Hardware, Jun 24, 2026 Ryzen 7 9800X3D, 32GB DDR5-6000 CL30, Asus TUF X670E-Plus (Tom's Hardware 2026 GPU testbed) Tom's Hardware GPU benchmarks hierarchy, per-game chart (average fps, native, no upscaling) |
Local LLM (llama.cpp / Ollama)
| Setting | Result | Source |
|---|---|---|
| pp512 Llama 2 7B Q4_0, Prompt processing |
14,073tok/s | llama.cpp community benchmarks (GitHub), Oct 19, 2026 llama-bench on CUDA, flash attention off; scoreboard entry by @totaldev, build 8cf6b42 |
| tg128 Llama 2 7B Q4_0, Generation |
290tok/s | llama.cpp community benchmarks (GitHub), Oct 19, 2026 llama-bench on CUDA, flash attention off; scoreboard entry by @totaldev, build 8cf6b42 |
UL Procyon AI Text Generation
| Setting | Result | Source |
|---|---|---|
| Llama 3.1 |
6,648points | KitGuru, Jan 29, 2025 Ryzen 7 9800X3D (KitGuru 2025 GPU test system) Card tested: RTX 5090 Founders Edition. UL Procyon AI Text Generation, ONNX Runtime with DirectML, Llama 3.1 overall score. |
Blender Benchmark
| Setting | Result | Source |
|---|---|---|
| Standard run |
14,966samples/min | Blender Open Data, Oct 19, 2026 Blender 4.5.0, OptiX: median score of the 1,555 runs submitted for "NVIDIA GeForce RTX 5090". |
Sources
- Blender Open Data (results database)
- KitGuru (review)
- Tom's Hardware (review)
- llama.cpp community benchmarks (GitHub) (results database)
As an Amazon Associate we earn from qualifying purchases, and we may earn from other shop links. Amazon's price is on Amazon; Best Buy's is read from Best Buy and shown with the time it was checked.