Windows 7: Microsoft Listens To Intel, Finally
Over the course of its life, Windows Vista has taken a load of abuse—much of which is deserved.
One area we’ve seen both Intel and AMD affected is power management. In AMD’s case, enabling Cool’n’Quiet technology on its original Phenom processors caused a sizable performance hit as Vista’s scheduler moved threads from active to idle cores running at half-speed in a process called migration.
Why did it do this? In order to maintain the symmetry of a system under full load, you don’t want I/O to become dependent on just one core. If you keep threads rotating between cores running at their maximum performance (this whole concept goes out the window when you start talking about spinning cores down), you get better responsiveness.
This was an implementation decision made during Microsoft’s Windows NT kernel design, and based on our experiences with both processor vendors' hardware, it wasn't considered a "feature" to either company. Of course, it affected Intel in a much different way than AMD. The problem Intel had in Vista was one of power consumption. For every migration, you have to write-combine the Nehalem architecture’s L3 cache, which costs power.
This changes with Windows 7 and a feature called ideal core. If a task’s needs are being addressed by one core, the operating system will let you stay there. This means two things to Intel: first, you don’t use power on the migration, and second, idle cores are able to remain in a C6 state. Purportedly, this migration fix alone will yield an extra 10 to 15 minutes of battery life on Nehalem-based notebooks, though this won’t become a major issue until the mobile dual-core Arrandale launches later this year. Perhaps more interesting, though, is that processors without C6 will not realize this gain (including AMD’s CPUs).
Core parking is a second optimization, based on the observation in previous operating systems that you might have four cores running background processes at 10% utilization each. The idea is to load all of those tasks onto one core and let the others idle if operating load levels allow for it. Now, you can see how these two features working together might have a significant impact on power, as ideal core prevents rabid thread migration, while core parking optimizes loading. Taken together, the pair intelligently maximizes the number of idle cores, and then keeps them from being spun up unnecessarily, yielding the theoretical power gains.
If you want to know more about the changes incorporated into Windows 7, check out this interview with Mark Russinovich, a Technical fellow for Microsoft.
We’re Making The Switch
Based on reader feedback to Windows Vista, access to the final Windows 7 code, and great app compatibility with our current benchmark suite, we updated as many software versions as possible and made the leap to Windows 7 for our review here (along with the gaming analysis of CrossFire and SLI graphics configurations, published separately).
But before we did, we wanted to quantify these power-saving claims from Intel for ourselves. So, we logged runs of PCMark Vantage on clean drives with an install of Windows Vista updated to Service Pack 2 and the Windows 7 RTM code, both x64 builds.
The results were actually counter to what we expected. The Windows 7-based build averaged six watts higher over the course of its run, but finished the test three minutes faster than the Vista machine. Also noteworthy, though, is that when the Windows 7 machine has a chance to idle (which is where we'd expect to see ideal core and core parking actually having an effect), it does dip down lower than the Windows Vista box.
We checked these results over with Intel, and came away with the following interpretation: the Windows 7 P-state promotion policies are more aggressive than Vista's, meaning a Windows 7 system ramps to Turbo Boost faster, resulting in the better performance and higher power consumption. At idle, the previously-discussed features enable the Windows 7 config to dip below the idle power draw of the Vista machine.
Overall, though, Windows 7 actually averages higher power consumption in this experiment, even if it simply idles for the three minutes while the Vista box finishes its run. We are fairly certain of why, exactly, this is, but will hold off on comment until we're able to present power data substantiating the claim. However, that doesn't change the fact that, in this case, Windows 7 won't be cutting your power bill. In order to show Windows 7 cutting consumption, we'd have to spend a lot more time at idle (admittedly more representative of how most PCs are typically used), replace a certain component, or disable certain settings in the OS.
