Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.
We measure real-world power consumption using Powenetics testing hardware and software. We capture in-line GPU power consumption by collecting data while looping Metro Exodus (the original, not the enhanced version) and while running the FurMark stress test. Our test PC remains the same old Core i9-9900K as we've used previously, to keep results consistent.
For the RTX 4090 cards, we've upped the quality setting in Metro Exodus to 4K extreme in order to tax the GPU as much as possible, and we run at 2560x1440 with FurMark. The following charts are intended to represent something close to worst-case power consumption, temps, etc. and so we do check other settings to ensure we're pushing the GPUs as much as reasonably possible.
The MSI 4090 ends up using slightly more power than the other two 4090 cards we've tested when running stock settings, at least in Metro — the Asus card 'wins' in the FurMark competition, though that's mostly thanks to having higher clocks. With overclocking, Asus also uses the most power, but again that's due to the allowed settings. We could set the power limit at 120% on the Asus card but only 110% on the MSI card.
GPU core clocks reflect the power settings. The MSI card at stock runs about 70 MHz faster in FurMark than the Founders Edition, while in Metro Exodus the gap narrows to 50 MHz. With overclocking, all three cards in Metro are within about 20 MHz of each other, indicating once more that there's not a massive difference in overclocking potential on the GPU core. Memory on the other hand is far more variable.
Fan speeds and temperatures are a bit interesting with the MSI card. There are three fans, two 120mm fans on the radiator and the 95mm fan on the card. It's not clear which fan speeds are reported by GPU-Z, but its fans averaged just under 1300 RPM in the Metro test, and increased to 1440 RPM in FurMark. Overclocking with our custom fan curve rocketed RPMs into the 2000–2200 range.
Temperatures were great, both stock and overclocked, so if you were worried the 240mm radiator wouldn't be sufficient for the 450W power draw, these results should assuage your fears. In FurMark, the GPU didn't even break 60C and came in 7C below the Asus card, though again that's also due to the clocks and power the card was willing to allow. With Metro, the Asus and MSI cards are effectively tied.
For noise levels, we use an SPL (sound pressure level) meter, and we position the radiator fans so that they're about 10cm away from the microphone, which is also 10cm from the fan on the card. This is done to capture the full noise profile and also avoid the influence of other potential noise sources. Idle noise measured less than 32 dB(A), which is more or less the limit of our noise floor.
Running a game for 15 minutes or more resulted in fan speeds of 44% on the MSI card and a noise level of 42.6 dB(A). That's about 1 dB(A) louder than the Asus card and 3.5 dB(A) quieter than the Founders Edition model. With our overclocked settings, the fan speed ramped up to 81% and 55.0 dB(A), which was better than the Asus card by 5.5 dB(A).
We also set a static fan speed of 75%, which resulted in 53.3 dB(A) on the MSI card, 57.2 dB(A) for the Founders Edition, and 58.1 dB(A) on the Asus card. Besides the fan noise, there's also slight noise from the pump on the card– not enough to really affect the noise readings, but in a quiet room at idle, you can just make out the sound of the pump circulating the liquid in the card.
| Game | Setting | Avg FPS | Avg Clock | Avg Power | Avg Temp | Avg %Load |
|---|---|---|---|---|---|---|
| Borderlands 3 | 1440p Badass | 212.9 | 2790.0 | 403.5 | 59.0 | 94.3% |
| Borderlands 3 | 4k Badass | 128.9 | 2758.4 | 444.0 | 61.4 | 99.0% |
| Bright Memory Infinite | 1440p Very High | 85.8 | 2773.9 | 422.0 | 60.1 | 99.0% |
| Bright Memory Infinite | 4k Very High | 41.1 | 2775.0 | 447.8 | 57.6 | 99.0% |
| Bright Memory Infinite OC | 4k Very High | 43.5 | 2970.0 | 484.1 | 54.4 | 99.0% |
| Bright Memory Infinite UC | 4k Very High | 33.0 | 1899.8 | 303.3 | 58.6 | 99.0% |
| Control | 1440p High | 136.1 | 2790.0 | 432.4 | 54.4 | 98.9% |
| Control | 4k High | 67.1 | 2751.2 | 444.5 | 58.9 | 99.0% |
| Control OC | 4k High | 72.3 | 2970.0 | 488.7 | 52.1 | 99.0% |
| Control UC | 4k High | 53.9 | 1801.8 | 295.2 | 56.2 | 99.0% |
| Cyberpunk 2077 | 1440p RT-Ultra | 87.5 | 2790.0 | 425.5 | 55.0 | 97.7% |
| Cyberpunk 2077 | 4k RT-Ultra | 44.6 | 2773.2 | 462.9 | 57.6 | 98.5% |
| Cyberpunk 2077 OC | 4k RT-Ultra | 47.0 | 2969.9 | 499.1 | 54.4 | 98.4% |
| Cyberpunk 2077 UC | 4k RT-Ultra | 33.8 | 1655.6 | 297.3 | 57.6 | 98.9% |
| Far Cry 6 | 1440p Ultra | 153.1 | 2790.0 | 238.2 | 53.0 | 65.8% |
| Far Cry 6 | 4k Ultra | 134.4 | 2789.3 | 342.9 | 52.8 | 96.0% |
| Flight Simulator | 1440p Ultra | 77.8 | 2804.7 | 224.4 | 42.3 | 40.2% |
| Flight Simulator | 4k Ultra | 77.5 | 2790.0 | 350.1 | 54.4 | 70.6% |
| Fortnite | 1440p Epic | 100.8 | 2734.0 | 429.7 | 62.6 | 97.2% |
| Fortnite | 4k Epic | 49.1 | 2668.9 | 438.2 | 62.4 | 98.0% |
| Fortnite OC | 4k Epic | 53.8 | 2933.9 | 518.5 | 58.2 | 98.2% |
| Fortnite UC | 4k Epic | 36.7 | 1807.9 | 296.2 | 59.1 | 97.7% |
| Forza Horizon 5 | 1440p Extreme | 173.1 | 2790.0 | 255.8 | 54.9 | 82.2% |
| Forza Horizon 5 | 4k Extreme | 150.7 | 2790.0 | 314.4 | 48.5 | 91.8% |
| Horizon Zero Dawn | 1440p Ultimate | 177.7 | 2782.1 | 263.8 | 56.9 | 67.6% |
| Horizon Zero Dawn | 4k Ultimate | 155.5 | 2790.1 | 381.3 | 51.7 | 95.1% |
| Metro Exodus Enhanced | 1440p Extreme | 115.6 | 2749.7 | 415.9 | 59.7 | 91.8% |
| Metro Exodus Enhanced | 4k Extreme | 75.6 | 2732.0 | 470.1 | 59.3 | 98.6% |
| Metro Exodus Enhanced OC | 4k Extreme | 79.4 | 2943.6 | 514.4 | 55.4 | 98.3% |
| Metro Exodus Enhanced UC | 4k Extreme | 56.5 | 1712.8 | 296.7 | 58.3 | 99.0% |
| Minecraft | 1440p RT 24-Blocks | 107.9 | 2790.0 | 343.3 | 53.9 | 66.1% |
| Minecraft | 4k RT 24-Blocks | 62.7 | 2774.8 | 413.4 | 58.8 | 80.1% |
| Minecraft OC | 4k RT 24-Blocks | 65.4 | 2970.0 | 438.2 | 51.6 | 78.5% |
| Minecraft UC | 4k RT 24-Blocks | 55.6 | 2149.3 | 303.6 | 57.5 | 81.5% |
| Total War Warhammer 3 | 1440p Ultra | 171.0 | 2778.0 | 404.6 | 57.4 | 94.1% |
| Total War Warhammer 3 | 4k Ultra | 107.0 | 2773.8 | 432.7 | 60.3 | 98.6% |
| Watch Dogs Legion | 1440p Ultra | 135.1 | 2789.4 | 285.3 | 55.3 | 74.1% |
| Watch Dogs Legion | 4k Ultra | 114.6 | 2786.8 | 391.9 | 56.5 | 97.6% |
Besides the Powenetics testing, we also collected temperature, power, and clock speed data (via Nvidia FrameView) for most of our benchmarks — "most" because FrameView causes Red Dead Redemption 2 Vulkan to crash to desktop. These are the results from the test runs used in our charts.
You can see that the GPU utilization in a lot of games at 1440p isn't anywhere close to 100%, particularly in the non-DXR tests. A faster CPU like the Core i9-13900K should increase FPS and utilization in such cases, whereas performance in all of the tests with GPU use in the 97%–99% range likely wouldn't change at all with a faster CPU.
Looking at the results with power limited to 300W (the "UC" in the charts and table), this also correlates with GPU utilization. If you limit the power on the RTX 4090 and then test in games where the GPU is already running well below 100% utilization and 450W power, performance won't drop too much. And that's why we confined our third party overclocking results to the DXR games, where we knew utilization was generally much higher.
- MORE: Best Graphics Cards
- MORE: GPU Benchmarks and Hierarchy
- MORE: All Graphics Content
30-year-old Pentium FDIV bug tracked down in the silicon — Ken Shirriff takes the microscope to Intel's first-ever recall
Cyberpunk 2077 update 2.2 claims to improve Arrow Lake performance by up to 33%, theoretically matching the Ryzen 7 7800X3D
Empyrean Technology gives control to CEC after U.S. blacklisting — China’s top developer of chip design systems hands reins to state-owned firm
