Microsoft released a freely available developer preview of its forthcoming Windows 8 operating system last week. Our own Seth Rosenblatt has a video below offering an overview of the new OS. What we haven't seen yet is an overview of Windows 8's impact on general computing performance.
A first, rough look at Windows 8
Microsoft has promised improved efficiency in Windows 8, due at least in part to the new OS relying on fewer background operations sucking up memory and processing power. A look at the process tab in Windows 8's new task manager bears this out. With Windows 8, we see 21 operating system processes running. With Windows 7 we count 31.
That slimming down alone should help overall performance. If Microsoft has done more to improve Windows 8's operating speed in this developer build, we haven't been told about it. We'll rely on our benchmark tests to bear that out.
Caveat time. Know that the version of Windows 8 we tested is a 64-bit developer's preview build. Any test results below could very well look different from results when the final version ships. Consider our performance comparison, like the operating system preview itself, a look at a work in progress.
Our test bed is a hand-built Intel Core i3-based desktop. Specs include a 3.1GHz Intel Core i3 2105; 8GB of 1,333MHz DDR3 SDRAM; an Intel DZ68D8 motherboard; a 500GB, 7,200rpm Western Digital SATA II hard drive; and a 2GB Nvidia GeForce GTX 560 Ti graphics card (for 3D gaming tests).
Application tests
We wanted to test the Windows 8 preview not just for its performance but also to check its compatibility with our current suite of benchmark applications, and unfortunately both 32-bit and 64-bit Photoshop CS5 crashed every time we tried to load an image file. That's unfortunate, since that test can be a good indicator of memory performance.
Instead, our iTunes/QuickTime multimedia multitasking test was the only case where we saw Windows 8 make a noticeable impact on speed. We welcome any speed improvement, and it shows that at least in this build, Microsoft appears to be on the right track for providing some noticeable application performance benefits, at least where memory utilization plays a role. We'll have a better idea once (or if) we can test out 64-bit Photoshop CS5.
Gaming tests
In addition to our CS5 difficulties, we were also unable to run 3DMark 11, and we had difficulty with Far Cry 2. The latter wouldn't run our lower-resolution test, and the scores between the operating systems on the higher-resolution test are wildly divergent. Our problems with both programs could be because of the operating system, the individual applications, the graphics card drivers, or a combination.
For the games that did work, we were most impressed by Windows 8's handling of the original Crysis. That game is notoriously hard for PCs, in part because it's a known system-memory hog. We'd expect a Core i3-based desktop with 8GB of RAM and a decent midrange graphics card to handle Crysis reasonably well, but even at high resolution, the Windows 8-based test bed showed impressive frame rate gains over the Windows 7-based system. Crysis is not the most relevant title anymore, but the fact that this build of Windows 8 makes easy work of this memory-chewing game suggests again that Microsoft's new OS does indeed bring with it some noticeable benefits to memory efficiency.
For newer games, don't take the Far Cry 2 scores seriously. We show them only to demonstrate the disconnect between operating systems with that game. The Metro 2033 results are more encouraging for gamers. We ran that test in DirectX 11 mode with full detail and at the highest resolutions we could muster on a single display. We didn't expect to see much of a performance boost between operating systems, but at least for this game, and with this OS build, Windows 8 doesn't seem to have lost a step on more-demanding 3D workloads. It's not surprising that we had some problems with certain programs in Windows 8, and it's reasonable to expect that most issues will be resolved as Microsoft and application developers update their code. What we did not necessarily expect is the noticeable performance boost between operating systems to applications and tasks that depend on system memory.
Most of the talk about Windows 8 thus far has centered around its new Metro touch interface and its potential as a cross-platform operating system. We're also interested to see how the mouse-driven interface develops over build iterations, because right now it feels too heavily touch-centered. But even if Windows is undergoing an interface identity crisis, based on what we've seen in our testing, computing traditionalists should look forward to the streamlining Microsoft has achieved in Windows 8's memory utilization. Let's hope Microsoft can hang on to those performance gains when it ships final code.
A first, rough look at Windows 8
Microsoft has promised improved efficiency in Windows 8, due at least in part to the new OS relying on fewer background operations sucking up memory and processing power. A look at the process tab in Windows 8's new task manager bears this out. With Windows 8, we see 21 operating system processes running. With Windows 7 we count 31.
That slimming down alone should help overall performance. If Microsoft has done more to improve Windows 8's operating speed in this developer build, we haven't been told about it. We'll rely on our benchmark tests to bear that out.
Caveat time. Know that the version of Windows 8 we tested is a 64-bit developer's preview build. Any test results below could very well look different from results when the final version ships. Consider our performance comparison, like the operating system preview itself, a look at a work in progress.
Our test bed is a hand-built Intel Core i3-based desktop. Specs include a 3.1GHz Intel Core i3 2105; 8GB of 1,333MHz DDR3 SDRAM; an Intel DZ68D8 motherboard; a 500GB, 7,200rpm Western Digital SATA II hard drive; and a 2GB Nvidia GeForce GTX 560 Ti graphics card (for 3D gaming tests).
Application tests
Apple iTunes encoding test (in seconds)
(Shorter bars indicate better performance)
(Shorter bars indicate better performance)
Windows 8 Developer Preview Build 8102 (64-bit)
110
Windows 7 Home Premium SP1 (64-bit)
111
Adobe Photoshop CS3 image-processing test (in seconds)
(Shorter bars indicate better performance)
(Shorter bars indicate better performance)
Windows 7 Home Premium SP1 (64-bit)
83
Windows 8 Developer Preview Build 8102 (64-bit)
87
Multimedia multitasking (in seconds)
(Shorter bars indicate better performance)
(Shorter bars indicate better performance)
Windows 8 Developer Preview Build 8102 (64-bit)
380
Windows 7 Home Premium SP1 (64-bit)
403
Cinebench 11.5
(Longer bars indicate better performance)
(Longer bars indicate better performance)
 | Rendering multiple CPUs | | Rendering single CPU |
Windows 7 Home Premium SP1 (64-bit)
3.01
1.24
Windows 8 Developer Preview Build 8102 (64-bit)
3
1.24
Instead, our iTunes/QuickTime multimedia multitasking test was the only case where we saw Windows 8 make a noticeable impact on speed. We welcome any speed improvement, and it shows that at least in this build, Microsoft appears to be on the right track for providing some noticeable application performance benefits, at least where memory utilization plays a role. We'll have a better idea once (or if) we can test out 64-bit Photoshop CS5.
Gaming tests
Metro 2033 (in fps)
(Longer bars indicate better performance)
(Longer bars indicate better performance)
| 2,560x1,600 (very high) | | 1,920x1,00 (very high) |
Windows 8 Developer Preview Build 8102 (64-bit)
10
18
Windows 7 Home Premium SP1 (64-bit)
10
18
Far Cry 2 (in fps)
(Longer bars indicate better performance)
(Longer bars indicate better performance)
| 1,920x1,200 (very high) |
Windows 7 Home Premium SP1 (64-bit)
98
Windows 8 Developer Preview Build 8102 (64-bit)
26
Crysis (in fps)
(Longer bars indicate better performance)
(Longer bars indicate better performance)
| 1,600x1,200 | | 1,280x1,024 |
Windows 8 Developer Preview Build 8102 (64-bit)
51
58
Windows 7 Home Premium SP1 (64-bit)
42
44
In addition to our CS5 difficulties, we were also unable to run 3DMark 11, and we had difficulty with Far Cry 2. The latter wouldn't run our lower-resolution test, and the scores between the operating systems on the higher-resolution test are wildly divergent. Our problems with both programs could be because of the operating system, the individual applications, the graphics card drivers, or a combination.
For the games that did work, we were most impressed by Windows 8's handling of the original Crysis. That game is notoriously hard for PCs, in part because it's a known system-memory hog. We'd expect a Core i3-based desktop with 8GB of RAM and a decent midrange graphics card to handle Crysis reasonably well, but even at high resolution, the Windows 8-based test bed showed impressive frame rate gains over the Windows 7-based system. Crysis is not the most relevant title anymore, but the fact that this build of Windows 8 makes easy work of this memory-chewing game suggests again that Microsoft's new OS does indeed bring with it some noticeable benefits to memory efficiency.
For newer games, don't take the Far Cry 2 scores seriously. We show them only to demonstrate the disconnect between operating systems with that game. The Metro 2033 results are more encouraging for gamers. We ran that test in DirectX 11 mode with full detail and at the highest resolutions we could muster on a single display. We didn't expect to see much of a performance boost between operating systems, but at least for this game, and with this OS build, Windows 8 doesn't seem to have lost a step on more-demanding 3D workloads. It's not surprising that we had some problems with certain programs in Windows 8, and it's reasonable to expect that most issues will be resolved as Microsoft and application developers update their code. What we did not necessarily expect is the noticeable performance boost between operating systems to applications and tasks that depend on system memory.
Take an early tour of Windows 8 (photos)
0 comments