When I was working on this article I was not sure if this section was necessary at this time. How fair would a test session in a beta version of the operating system actually be? I have come across some comparisons where Windows 7 was faster than Windows Vista as well as Windows XP, so I was really curious to see if that was in fact the case. However, while we can and should test in Windows Vista and XP, the results obtained in Windows 7 are all preliminary and more of a preview type. This OS is still far from being finalized: some services are non-operational, some functions will still be added, some will be deleted or modified. By the time the operating system is ready to launch, its actual performance may be faster or slower than what we will obtain today. Nevertheless, the obtained results turned out very interesting, so I decided to share them with you, though I encourage you to remember that they are far from being final.
Here is our test platform:
- Gigabyte GA-EP45T-Extreme mainboard, rev. 1.0, BIOS F4;
- Intel Core 2 Quad Q9300 CPU (2.5 GHz, 333 MHz FSB, 6MB L2 cache, Yorkfield, rev. M1, 1.175 V Vcore);
- 2 x 1024 MB DDR3 OCZ PC3-14400 Platinum Series,OCZ3P18002GK, (1800 MHz, 8-8-8-24, 1.95 V voltage);
- ATI Radeon HD 4870 512 MB;
- Samsung SP2504C HDD (250 GB, SATA II, 7200 RPM, 8 MB, rev.A);
- Zalman CNPS9500 LED cooler;
- Noctua thermal interface;
- Antec NeoPower HE 550 PSU (550W);
- Antec Skeleton case.
The system worked in nominal mode. The CPU worked at nominal 333MHz bus frequency with x7.5 multiplier. The memory worked at 1067MHz frequency with 7-7-7-20-2T timings. The graphics card also worked at its nominal speeds: 750MHz chip and 3600MHz GDDR5 memory frequency. We used ATI Catalyst 8.12 driver. We compared the results obtained in 32-bit versions of Microsoft Windows Vista Ultimate SP1 and Microsoft Windows 7 Beta.
The utilized test applications included not only synthetic utilities, but also practical tasks from different fields. I have to say that after two weeks with Windows 7 Beta I didn’t have any problems with this OS, however, I had to give up a few benchmarks because of compatibility or other issues that surfaced during the test session. Among these benchmarks are BAPCo SYSmark 2007, Custom PC Benchmarks Suite 2007, PCMark Vantage and Crysis Warhead (the game runs just fine, but the performance measuring utility doesn’t). As a result, here is the complete list of tests that we performed:
- 3DMark Vantage v1.0.1, default settings;
- Crysis v1.2.1, DirectX 10, 1280х1024, Medium Quality;
- FarCry 2, DirectX 10, 1280х1024, High Quality;
- Cinebench 10 rendering;
- Fritz Chess Benchmark;
- TOC F@H Bench v0.4.7.0, DGROMACS 1 (+);
- Paint.NET Bench 3.36;
- X264 Benchmark.
Moreover, we used performance tests built into Microsoft Windows Vista and Microsoft Windows 7 Beta. Of course, we didn’t use the relative scores in points, but the actual numbers. For example, no matter how fast the RAM is in Windows 7, it will never score more than 5.5 points if its capacity is less than 3GB. At the same time, when we test the memory subsystem performance, we measure its real bandwidth in MB/s, which we are going to use for further analysis. Windows 7 Beta has a completely new set of performance benchmarks for the graphics subsystem, while all other results should be pretty comparable. Here are the tests we used:
- CPU LZW Compression;
- CPU AES256 Encryption;
- CPU Vista Compression;
- CPU SHA1 Hash;
- Memory Performance;
- Disk Performance;
- Media Foundation Decode Time.
We ran each test three times, discarded the peaks, calculated the average mean and then marked the best score with bold font in the table below:
Although Windows 7 Beta loses to its opponent in most of the tests, it is not the time to be upset. We have already named the main reason: the performance score of the final version may still be dramatically different from that of the current beta version. Moreover, the performance difference is often very small and falls within the measuring error and a few more serious discrepancies may still be written off as the results obtained on a beta version. And how could it be otherwise if we compared two identical systems that differed only by the OS versions, which have a lot in common. However, there are a lot of examples, when much more significant differences do not affect the performance results. We can change memory frequency and timings, however, it will remain unnoticed if the test is not sensitive to these parameters. We can overclock the CPU, but the gaming performance will not change if we have a weak graphics card. And graphics card overclocking will not speed up processor-dependent calculations. So, the performance difference in these tests simply has to be small, considering that these systems are identical.
Moreover, I won’t be upset even if the performance of the final Windows 7 version will be lower than that of Windows Vista. It is quite possible if the new OS acquires a lot of new functions and options. So what? Tomorrow I will upgrade my RAM, replace the CPU with a faster one, overclock the system and in the end make up for the “lost” performance. However, I will have a much more user friendly, responsive, secure and reliable operating system.