Author Topic: Benchmarking Windows 7: Harder, Better, Faster, Stronger?  (Read 879 times)

Offline javajolt

  • Administrator
  • Hero Member
  • *****
  • Posts: 36193
  • Gender: Male
  • I Do Windows
    • windows10newsinfo.com
Benchmarking Windows 7: Harder, Better, Faster, Stronger?
« on: October 28, 2009, 09:54:53 PM »

While most Tom’s Hardware readers initially resisted the switch from Windows XP, responses to our own articles since last year suggested that many have given up the fight and moved to Vista.

Perhaps Microsoft’s end-of-life announcement for that venerable operating system had something to do with the rise in popularity of Windows Vista. Or maybe high-end users simply wanted to enjoy the newer features, such as its Aero visual enhancement, as well as Media Center, available to Home Premium and Ultimate buyers. However, a few of Vista’s nagging problems have persisted, including slow loading times, a ravenous appetite for RAM, and the need for even more RAM to minimize those slow loading times via SuperFetch.

Meanwhile, radically reduced prices for high-speed, high-capacity memory is most likely the reason behind the increased popularity of Vista among enthusiasts. New applications and games that required additional RAM arrived just in time to match the increase in mid-priced DRAM density, but a 64-bit operating system would be required to properly support capacities of 4GB or more. Vista was the first Windows OS to mainstream its 64-bit versions, and spotty driver support under 64-bit Windows XP has made Vista x64 the easier choice.

But for many of us, the compromise of using an OS that both supports and requires additional memory seems wasteful, especially when that added memory doesn’t give us back 100% of the responsiveness previously felt in the OS it replaced. Microsoft is finally ready to handle that concern with Windows 7, which is really an improved version of Vista.


Tom’s Hardware has already published articles explaining some of the more complex changes Microsoft incorporated into its latest operating system. However, hardware editors are more concerned with the simple matters of program performance and responsiveness. Thus, while we’ve heard a great many details of how Windows 7 improves such things as security, features, and “user experience,” we probably define “experience” differently than the software guys. When narrowing our focus to basic performance, how much better is Windows 7 than the OS it replaces? It's time to quantify the differences!

Test Settings


Modern hardware and software deserve each other, so we used some of our latest parts to gauge the performance difference of each operating system.


We used Asus’ P7P55D motherboard. Featuring Intel’s latest LGA 1156 socket and P55 Express PCH, the board took top honors, alongside one from Gigabyte, in our recent mainstream roundup. Better compatibility with this tester’s ancient drive-imaging software made Asus the default choice.


The reason for using so modern a motherboard is to support Intel’s latest LGA 1156 processors. Turbo Boost allows its Core i7-870 to increase CPU multipliers up to 27x (3.6 GHz), 26x (3.46 GHz), and 24x (3.2 GHz) for single-, dual-, and multi-threaded applications.


One significant fact about LGA 1156 processors is that they require high C-states (sleep states) for three cores in order to allow the highest multiplier to affect a single core. Rumors of better C-state function under Windows 7, if true, could give the new OS a significant performance advantage in single-threaded applications.


Thermalright’s new MUX-120 cooler brings Cogage True Spirit performance to LGA 1156 processors.



Benchmark Results: Crysis And Far Cry 2


Windows 7 shows a moderate improvement in our highest Crysis test, but results that don’t yield smooth gameplay have little significance.



There’s no noticeable performance difference between the two operating systems in Far Cry 2.

Benchmark Results: Clear Sky And World In Conflict




World in Conflict gives us the strangest results yet, scaling worse under Windows 7 but with better performance at the lowest setting.

Benchmark Results: Audio And Video Encoding


Audio Encoding takes less time under Windows 7 in both iTunes and Lame.


TMPGEnc appears ill-suited for Windows 7, so we’ll keep our eye out for an update. A difference of this size normally indicates some type of incompatibility.


MainConcept runs marginally faster under Vista.

Benchmark Results: Productivity


Adobe Photoshop CS4 favors Windows Vista slightly, though differences this small are occasionally seen between consecutive runs. That is to say, it could be by chance.


In 3ds Max, the difference between Vista and Windows 7 is once again only one second, even though this shorter benchmark makes it appear larger than it had in Photoshop. With number rounding, a difference of less than a second could still be by chance.



AVG appears to favor Windows 7, but once again a difference of less than a second (after rounding) could also be by chance.


In a more definitive benchmark, WinRAR favors Vista. Meanwhile, WinZip does not.

Real-World Benchmark Analysis


Before moving on to synthetics, let's take a closer look at what our real-world application benchmarks have revealed.


Games appear to have a negligible preference for Windows 7, while productivity applications prefer Windows Vista. The average for encoding applications is thrown off—way off—by a problem in TMPGEnc, leading to a combined performance difference of nearly 4% favoring Vista.

What would these results look like if we eliminated the “partially-broken” TMPGEnc test?


With TMPGEnc out of the equation, the encoding difference drops to 0.41% with a combined average difference of 0.57% favoring Vista.

But none of these applications reveal anything about load times. For that we have two synthetics, PCMark and SYSmark.

Benchmark Results: Synthetics


Sandra is relatively OS-agnostic, so a quick run of a few benchmarks should show hardware performance unchanged.



Now that we’ve determined nothing has gone awry in hardware, let’s see how the rest of our synthetics respond to the change in software.


3DMark performance changes certainly aren’t noteworthy.


PCMark tests several aspects of system performance, including startup time. Its Productivity results indicate that Windows 7 will “feel” around 10% faster in daily use.


SYSmark 2007 Preview shows a huge 25% gain for Windows 7 in its Video Creation suite, although the average improvement is a more realistic 7%. Getting a 7% improvement with a hardware change would typically cost hundreds of dollars. Unlike most other synthetics, SYSmark uses real productivity suites such as Adobe Creative Suite and Microsoft Office to calculate its performance scores, so users should be able to take that 7% performance gain as a solid indicator of how much faster the system will feel in daily use.

Conclusion

From the benchmarker’s standpoint, the change from Windows Vista to Windows 7 is simply a matter of documentation. The one benchmark that showed a big difference was actually slower in Windows 7, likely due to a compatibility issue with one of the program’s components.

Yet, by being designed to show only the performance differences of various hardware, typical benchmarks do not represent the total performance experience of end users. A game certainly appears smoother at 40 frames per second (FPS) than at 30 FPS, but the benchmark doesn’t indicate how long it takes to load a map. Rendering a 3D animation frame in 24 seconds certainly saves time over rendering it in 25 seconds, but again it doesn’t indicate how much time the user wastes starting the program or changing between menus.

To better gauge the user’s performance experience or how fast the system feels, load times must be captured. While we could certainly have used a stopwatch, an electronic timer is far more accurate, and benchmarks such as PCMark and SYSmark use them. These programs indicate that Windows 7 feels 7% to 10% faster than Windows Vista, and that’s enough to make us give the new OS the nod in spite of its lack of differentiation in most of our other test.

source:tom'shardware