Enthusiast Look: Windows XP 32-Bit versus 64-Bit

[PAGETITLE=Introduction] Since Windows XP Pro 64-Bit was released, there has been a lot of speculation of whether it would help current gaming or anything else in general. I put both versions of the OS through a round of benchmarks to see if the 64-Bit does indeed offer any advantage. When Microsoft released the 64-Bit version of Windows XP Professional, there were many people wanting to upgrade to see if it would make any difference to their applications. Primarily though, I was concerned with gaming advantages, especially games like Half-Life 2 and Far Cry that can really push hardware and the CPU to it's limit. However, there are a couple things to first note about the 64-Bit version of XP. On the "Top 5 reasons to get to Windows XP Professional x64 Edition" of Microsoft.com, they give the straight reasons of why YOU should be using Windows 64-Bit. They note that it is ready to support the next generation of 64-Bit computing. Up to now, we have not really seen any dedicated 64-Bit software, but only 32-Bit software patched to take advantage of 64-Bit. Large memory support, up to 128 Gigabytes of memory, and 16 Terabyte of virtual memory. Considering the normal gaming computer has 1/128th of the highest supported RAM, and 1/80th of the supported virtual memory, we can easily see that this is certainly not catered to a gamer of any sort. Looking on though, they also add great support towards developers, and Dual-Core. Unless you are a hardcore multi-media type, gamers currently don't even have a need for Dual-Core under a regular 32-Bit OS. Whether the 64-Bit OS is catered to gamers or not, it has to be tested to see if any gain in performance is made anywhere. Is there even an inch of reason to upgrade? I have run a collection of gaming and non-gaming benchmarks within both versions of Windows, and have outputted the results in graphs. [PAGEBREAK] [PAGETITLE=What would benefit?] Technically speaking though, games shouldn't really have any reason to work better in a 64-Bit enviroment. The capabilities of the 64-Bit are much greater than 32-Bit, but there are many things that have to come into play before it's optimized in 64-Bit. Firstly, the operating system needs to be 64-Bit capable, which XP 64-Bit Professional obviously is. The game in question also needs to be developed to make use of a 64-Bit architecture. None to my knowledge currently have been developed for 64-Bit, but Far Cry recently had a 64-Bit patch released, which I will be using in the benchmarkings. In addition, your Processor also needs to be fully 64-Bit, such as the AMD64/AMD64 FX series. Lastly, your GPU drivers also need to be developed for the 64-Bit platform. Other things may come into play as well, such as Audio drivers and Memory, but they would likely make very minute differences, if any. Every test in this article was performed with the system running at stock. I left overclocking entirely out of the equation, and didn't bother to test it myself. Of course, overclocking would increase scores all around. System is as follows: Processor AMD 64 3200+ S939 "Venice" Core (512k L2 Cache) @ 2.00GHz Motherboard DFI LanParty NF4 UT Ultra-D Power Supply Ultra X-Finity 600W Memory 512 * 2 Crucial Ballistix DDR400 (2-2-2-8) Hard Disks 200GB * 2 Western Digital 8MB Cache160GB Western Digital 8MB Cache Sound Card AC'97 7.1 Built-In Video Card BFG 6800 GT OC 256MB Etcetera Windows XP Professional with SP2Windows XP 64 Professional with SP132-Bit Video Drivers are official 71.8964-Bit Video Drivers are beta 76.80BIOS is March 10, 2005 Revision All the benchmarks and tests performed during the article has been done on a machine with all uneccessary applications and services shut down. For all the gaming related tests, including 3D Mark, the tests were each run twice, and for the hardware benchmarks, they were run once each. [PAGEBREAK] [PAGETITLE=Futuremark] Let's start off with a classic gamers benchmark, 3D Mark 01. When it first came out, back in the day, '01 really pushed our hardware to the limits. I clearly remember my jaw dropping when I first seen the Nature test, which was quite fluid on my GeForce 3 MX. At any rate, '01 introduced Vertex Shading as well as Shader Model 1.4, and is built upon DirectX 8.1. I used '01 to test, just to see if the 64-Bit had deteriotated quality or performence, which is hasn't. Here we also tested the PC Mark '04 complete suite test. As you can see, both tests scored higher in the 32-Bit OS, but not so much that you would actually notice it during regular gameplay. Here we have the two newest 3D Marks to test. 3D Mark '03 is DirectX 9a based, but only the Nature test utilizes it to the full potential. 3D Mark '05 also uses DirectX 9, specifically the 'c' revision, and is much more intense. We can see from the results, that the 64-Bit performed lower in the '03 tests, but higher in the '05. Once again though, the differentials in between the scores are so little, that we may as well consider them equal. [PAGEBREAK] [PAGETITLE=Half-Life 2] I thoroughly benchmarked through three popular FPS's: Half-Life 2, Doom III and Far Cry. Far Cry is the only one out of the three that has an actual 64-Bit patch for use with x64 XP. To benchmark Half-Life 2, I used the Half-Life 2 Benchmarking Tool. The first round of tests, I used two built-in Demos, and then I created two myself and tested them on both platforms. Half-Life 2 is much more CPU intensive than other FPS out there, so I was interested to see if there would be any improvements here. We can see that the Guru3D Demo played equally on both platforms, but the HardwareOC Coast demo averaged an extra 20 frames through both resolutions. This is the first sign of a gaming advantage with the x64 XP. [PAGEBREAK] [PAGETITLE=Doom III] In the next tests, I benchmarked two custom tests, in the C17 and Canals levels. Don't pay attention to comparing the colors here, because they are seperate from each other, as they specify the resolution and nothing else. The biggest differential here is in the Canals level at 1024*768, with an increase of 14FPS. Throughout all the tests though, the 64-Bit version of Windows made the game perform better. For Doom III testing, I just used the built-in demo1 that everyone knows so well. The first run on each resolution was ignored, and the score from the second and third runs were averaged out to get the final FPS. There is virtual no real performance increase here, although it is still better, albeit slightly. [PAGEBREAK] [PAGETITLE=Far Cry] To benchmark Far Cry, I ran through two resolutions, and two different levels. For some levels, they have built-in demos, so I just used those, for Regulator and Volcano. For Windows 32-Bit, I used the 1.31 patch, and for x64, I used the 64-Bit patch. This is the only (I believe) game currently out to have a 64-Bit patch, so I was really expecting to see some differences here. To use the patch, you must be running x64 Windows. I am not going to get into the differences of what benefits there are over the 32-Bit version, however Hard|OCP recently had an article devoted to it, which you can read here, for more information. Even with the 64-Bit patch, the game benchmarked a few FPS lower than the 32-Bit version. However, the levels I tested directly, don't really take advantage of the 64-Bit as they could. There are two bonus levels with the 64-Bit patch, however the levels will only work under x64, so I was unable to compare them to the 32-Bit, naturally. Since this was a patch, and the game wasn't entirely built upon a 64-Bit architecture from the start, performance wasn't hurt, but it wasn't greater either. However, in the patch, and as you can see in the mentioned Hard|OCP article, they added some bonus features that utilize the processor, such as rocks underwater that have better depth and generally look more real. In the future, when games take much better advantage of 64-Bit, we should be seeing many cool features to our games. [PAGEBREAK] [PAGETITLE=Everest] Now to get into some benchmarks that ignore the GPU entirely. We first used all nine of EVEREST's built-in Benchmarks, and put them in a single graph. Each test was run twice, and then averaged off. However, not much averaging took place, as the results were dead on most times. No large differences here once again. CPU Photoworxx is the only notable, with 150 less points with the 64-Bit XP. In SiSoftware Sandra 2005, we start to see more of a fluctuation of scores between the 32 and 64 bit. I should note, that for the x64 XP, I used the 64-Bit version of Sandra. The first three results in the graph are Arithmetic benchmarks, so they are math intensive. The 64-Bit performed better than the 32-Bit in all three, especially with the Dhrystone ALU, scoring more than 500 extra points. The last two results are both Multi-Media tests, and here we see even more scoring differences. The aEMMX/aSSE3 scored a substantial 4,149 points lower than the 32-Bit. Then we can see it performed better with the Floating-Point iSSE2, scoring 2,508 higher. Depending on what you are doing, you can't really win or lose here, between the two versions. [PAGEBREAK] [PAGETITLE=VirtualDub] For the last test, it was more of a test of interest, as there is no real evidence that a true 64-Bit enviroment would improve any of these. VirtualDub actually has a 64-Bit version, but it would not utilize the Divx Codec, so I had to use the 32-Bit version. For the test, though, I took a 105MB MPEG file, and converted it to a Divx AVI, using the highest quality (Slowest processing) setting. All results are in (Minutes.Seconds) The conversion process was 1.1% faster with the x64. Definitely not something to get excited over. Even converting very large videos, should not see much of a difference. When the VirtualDub 64-Bit version uses Divx, we may be able to see some differences then. With ACDSee 7, I converted 771 JPEG images to Bitmaps. They were neck and neck for the full conversion...

Enthusiast Look: Windows XP 32-Bit versus 64-Bit

Page 180 –

Rob Williams