GPU Performance & Specs
Compare graphics cards with public benchmark scores and official specifications.
GPU DATA LAB
GPU performance & specs
Compare public benchmark scores and official hardware specifications in one place.
Public Time Spy Extreme index + vendor specification cross-checkSearch results
GPUs shown
468Average benchmark
4,692Average performance per watt
0.09 TFLOPS/WAverage design power
154 WTerms: FP32 is graphics compute, TDP is the manufacturer power rating, and VRAM is memory reserved for the graphics card.
| # | Model | Time Spy score | Theoretical graphics compute (FP32) | Graphics memory (VRAM) | Design power (TDP) | Performance per watt | |
|---|---|---|---|---|---|---|---|
| 1 | RTX PRO 6000 Blackwell NVIDIABlackwell | 27,891 | 126.0 TFLOPS | 96 GB | 600 W | 0.21 TFLOPS/W | |
| 2 | GeForce RTX 5090 D NVIDIABlackwell | 25,999 | 104.8 TFLOPS | 32 GB | 575 W | 0.18 TFLOPS/W | |
| 3 | GeForce RTX 5090 32 GB NVIDIABlackwell 2.0 | 25,766 | 104.8 TFLOPS | 32 GB | 575 W | 0.18 TFLOPS/W | |
| 4 | GeForce RTX 5090 NVIDIABlackwell | 25,206 | 104.8 TFLOPS | 32 GB | 575 W | 0.18 TFLOPS/W | |
| 5 | GeForce RTX 5090 D V2 NVIDIABlackwell | 25,185 | 104.8 TFLOPS | 24 GB | 575 W | 0.18 TFLOPS/W | |
| 6 | RTX PRO 6000 Blackwell Max-Q NVIDIABlackwell | 19,954 | 110.1 TFLOPS | 96 GB | 300 W | 0.37 TFLOPS/W | |
| 7 | GeForce RTX 4090 24 GB NVIDIAAda Lovelace | 19,460 | 82.6 TFLOPS | 24 GB | 450 W | 0.18 TFLOPS/W | |
| 8 | GeForce RTX 4090 NVIDIAAda Lovelace | 19,453 | 82.6 TFLOPS | 24 GB | 450 W | 0.18 TFLOPS/W | |
| 9 | RTX PRO 5000 Blackwell 72GB NVIDIABlackwell | 17,934 | 73.7 TFLOPS | 72 GB | 300 W | 0.25 TFLOPS/W | |
| 10 | RTX PRO 5000 Blackwell 48 GB NVIDIABlackwell 2.0 | 17,934 | 66.9 TFLOPS | 48 GB | 300 W | 0.22 TFLOPS/W | |
| 11 | GeForce RTX 4090 D NVIDIAAda Lovelace | 17,890 | 73.5 TFLOPS | 24 GB | 425 W | 0.17 TFLOPS/W | |
| 12 | GeForce RTX 5080 16 GB NVIDIABlackwell 2.0 | 16,578 | 56.3 TFLOPS | 16 GB | 360 W | 0.16 TFLOPS/W | |
| 13 | GeForce RTX 5080 NVIDIABlackwell | 16,252 | 56.3 TFLOPS | 16 GB | 360 W | 0.16 TFLOPS/W | |
| 14 | Radeon RX 7900 XTX 24 GB AMDRDNA 3 | 15,278 | 61.4 TFLOPS | 24 GB | 355 W | 0.17 TFLOPS/W | |
| 15 | Radeon RX 9070 XT 16 GB AMDRDNA 4 | 14,633 | 48.7 TFLOPS | 16 GB | 304 W | 0.16 TFLOPS/W | |
| 16 | Radeon RX 9070 XT AMDRDNA 4 | 14,315 | 48.7 TFLOPS | 16 GB | 304 W | 0.16 TFLOPS/W | |
| 17 | GeForce RTX 4080 SUPER NVIDIAAda Lovelace | 14,219 | 52.2 TFLOPS | 16 GB | 320 W | 0.16 TFLOPS/W | |
| 18 | GeForce RTX 4080 NVIDIAAda Lovelace | 14,067 | 48.7 TFLOPS | 16 GB | 320 W | 0.15 TFLOPS/W | |
| 19 | GeForce RTX 4080 SUPER 16 GB NVIDIAAda Lovelace | 14,052 | 52.2 TFLOPS | 16 GB | 320 W | 0.16 TFLOPS/W | |
| 20 | GeForce RTX 4080 16 GB NVIDIAAda Lovelace | 14,002 | 48.7 TFLOPS | 16 GB | 320 W | 0.15 TFLOPS/W | |
| 21 | Radeon AI PRO R9700 AMDRDNA 4 | 13,758 | 47.8 TFLOPS | 32 GB | 300 W | 0.16 TFLOPS/W | |
| 22 | GeForce RTX 5070 Ti NVIDIABlackwell | 13,534 | 43.9 TFLOPS | 16 GB | 300 W | 0.15 TFLOPS/W | |
| 23 | RTX PRO 4500 Blackwell NVIDIABlackwell | 13,443 | 51.0 TFLOPS | 32 GB | 165 W | 0.31 TFLOPS/W | |
| 24 | Radeon RX 7900 XT 20 GB AMDRDNA 3 | 13,130 | 52.4 TFLOPS | 20 GB | 315 W | 0.17 TFLOPS/W | |
| 25 | Radeon RX 9070 16 GB AMDRDNA 4 | 13,072 | 36.1 TFLOPS | 16 GB | 220 W | 0.16 TFLOPS/W | |
| 26 | Radeon RX 7900 XTX AMDRDNA 3 | 12,751 | 61.4 TFLOPS | 24 GB | 355 W | 0.17 TFLOPS/W | |
| 27 | RTX 5000 Ada Generation NVIDIAAda Lovelace | 12,700 | 65.3 TFLOPS | 32 GB | 250 W | 0.26 TFLOPS/W | |
| 28 | Radeon RX 7900 XT AMDRDNA 3 | 12,686 | 52.4 TFLOPS | 20 GB | 315 W | 0.17 TFLOPS/W | |
| 29 | Radeon RX 9070 AMDRDNA 4 | 12,647 | 36.1 TFLOPS | 16 GB | 220 W | 0.16 TFLOPS/W | |
| 30 | RTX 6000 Ada Generation NVIDIAAda Lovelace | 12,629 | 91.1 TFLOPS | 48 GB | 300 W | 0.30 TFLOPS/W |
Method and sources
The ranking uses public 3DMark Time Spy Extreme results. Graphics memory, memory speed, power and core fields prioritize vendor specifications. FP32 means graphics compute and TDP is the design power rating. Performance per watt is only FP32 divided by TDP; it is not game FPS or an electricity bill.
Rankings can change as products, drivers and benchmark conditions change. Game tools can reuse these hardware fields while measured FPS remains separately labeled.
How to use
- Search or filter by GPU, vendor, generation, or use.
- Choose a table, card, or bar view and a sort order.
- Open GPU details to review the benchmark and official specification sources.