Performance results: Synthetic benchmarks
Let's first compare the results of the Futuremark benchmarks.
PCMark 7 is a general-purpose benchmark, with a minor, DirectX 9-based game component. The general PCMark score that PCMark 7 generated under Windows 8 was about 5 percent faster than the result under Windows 7: 5501 for Windows 8 versus 5248 under Windows 7.
The DirectX 10 3DMark Vantage performance test posted a score of 31,183 on Windows 8, versus 30,874 on Windows 7. Don't be fooled: That's less than a 0.1 percent difference, so it's a statistical dead heat.
3DMark 2011 uses DirectX 11, including hardware tessellation and DirectCompute for computing physics. Running Windows 7, the system posted a 3DMark 2011 performance score of 9299; it hit a score of 9361 on Windows 8. Again, that difference is so minor as to be essentially identical.
Unigine Heaven can really hammer on the GPU's tessellation engine. With the test running at 1920 by 1200 with 4x antialiasing enabled, Windows 8 and Windows 7 each posted a score of 51 frames per second.
Ultimately, in the synthetic benchmark suites, we saw a minor improvement in general performance under Windows 8, but the results were pretty much a dead heat on 3D rendering. What about games?
Crysis 2: At first blush, Crysis 2 ran substantially more slowly on Windows 8, operating at 61 frames per second, while hitting frame rates of about 69 fps on Windows 7. The issue turned out to be vsync (vertical synchronization), which synchronizes frames generated in games with the refresh rate of the monitor (60Hz). Turning off vsync in Crysis 2 or even in the Nvidia driver control panel had no effect. Some users have reported that turning vsync on, and then turning it back off, fixes such problems. But our benchmark script can't do that, so for the time being our Crysis 2 test is pointless, since vsync is essentially locked to the monitor refresh rate.
Shogun 2: Total War: Shogun 2 allows you to adjust more GPU knobs and levers than just about any other game on the market. When I cranked up everything to maximum settings, Shogun 2 under Windows 8 generated substantially higher frame rates than it did on Windows 7. In Windows 8, Shogun 2 at 1920 by 1200 with 4x antialiasing enabled hit 56 fps; under Windows 7 it managed only 35 fps. That's a hefty difference. You probably won't see that big a difference in actual gameplay, but it's still worth noting.
Dirt 3: Windows 7 posted a marginally better score in the Dirt 3 test, achieving a rate of 117 fps versus 113 fps for Windows 8. That's about a 2.5 percent difference, and probably nothing to get worked up over.
Metro 2033: This first-person shooter is an incredible system hog. Our results—33 fps for Windows 7, and 34 fps for Windows 8—were a dead heat.
In addition to running the above games, I played Civilization V, Mass Effect 3 (mostly multiplayer), and Bioshock 2 for extended periods of time. Inside each game, I saw no real performance issues, nor did I notice any image-quality changes between Windows 7 and Windows 8. I set all games to maximum graphics settings during my playing sessions.
I also played a few levels of the Crysis 2 single-player campaign. Despite the frame-rate cap, the game ran smoothly, with no significant issues. Finally, I fired up Borderlands 2 for a little multiplayer action, which was both smooth and adrenaline inducing, as you'd expect.
Built-in cloud saves on the game-download services seemed to work well, too. I was able to open both Civilization V and Mass Effect 3 single-player saved games, which had been synced with the Steam and EA Origin services.