Maxman
04-18-2008, 03:48 AM
http://www.ozhardware.com.au/images/stories/Article%20Images/XPVISTACOMP/vistaxpsmall.jpg
Even before Windows Vista was publicly released, there was a lot of concern from people beta testing the O/S. Running Windows Vista seemed to have a large performance drop for their PC when compared to Windows XP. Was it true? Or is it just scare tactics from Microsoft haters?
It is widely known that you need a much faster PC to run Vista than you do to run Windows XP. Microsoft recommend a minimum of a 1Ghz CPU and 1GB of RAM to run anything better than Vista Home Basic (I.e. Home Premium, Business or Ultimate) , and a Directx9 capable video card. I would hate to try and run Vista on a 1Ghz CPU, it would just be painfully slow. So when specing-up your new machine, make sure you at least double the minimum recommended RAM and CPU speed, have a dedicated PCI-Ex16 video card with 128MB of RAM or more, and you will be fine. If your video card supports DirectX9, you should even be able to run Aero.
Vista may include a host of new features, look a whole lot better and tingle your inner geek in all the right places, but most people just want their games to run smoothly, their office applications to open quickly and run without fuss, and the internet to be as fast as possible, so why would you upgrade to an operating system that will serve only to slow you down? Well we’re going to run a few quick tests to either prove or dispel the rumours.
What we’ve done is setup a high-spec machine, and we’re going to do a standard load of Windows XP Professional SP2, run our benchmarks, then a standard load of Windows Vista Ultimate and perform the same tests. The graphics tests will be run using 3DMark06, which runs on DirectX 9, not DirectX 10 as is written into Vista, so whilst it won’t be as true a comparison of possible performance as we would like, it is the best we can do at the moment, and tests current 3D software, which is what matters. 2D application tests will be performed by PCMark05. Whilst these are both ‘artificial’ benchmarks as it were, they perform a wide variety of tests, with 3DMark simulating current and future game rendering systems, and PCMark simulating everything from rendering web pages, to 3D Performance, 2D Image rendering and Hard Disk usage, all in one neat package.
Test Machine Specs:
ASUS Striker Extreme nForce680i SLI Motherboard
Intel Core 2 Duo E6700 CPU
Corsair TWIN2X6400 2GB DDR-2 Memory Kit
2 x Gigabyte GeForce 8800GTX 768MB PCI-Ex16 Video Cards running SLI
Western Digital Raptor 150GB 10,000rpm SATA-II Hard Drive
OCZ GameXstream 850W ATX Power Supply
Certainly a bit higher-end than most home or office PCs, but this will allow us to run much more intense tests and be able to spot any differences more easily.
The Results:
For 3DMark06 we ran 3 different tests, one with the video cards in SLI mode, one using only a single video card, and another in SLI mode using 8x Anti-Aliasing to try to see if there’s more of a difference when the 3D system is pushed hard. All tests were performed at a resolution of 1280x1024.
http://www.ozhardware.com.au/images/stories/Article%20Images/XPVISTACOMP/3dmark06.jpg
As you can see above, there is around a 5% performance increase using Windows XP under both SLI tests, the single card test showed a negligible difference, but still in favour of old-hat XP.
There were a few tests within 3DMark that showed a massive difference between the O/S’s, specifically the shader tests. The simple Vertex Shader test showed almost a 50% drop in performance under Vista, with XP running 369Million Vertices/sec versus 186MV/s under Vista. The shader particles test also showed Vista had a 25% drop from XP, and the pixel shader test a 15% drop in performance.
3DMark also performs Batch Size tests of moving triangle systems, starting with 8 triangles and moving up with 32, 128, 512, 2048 and finally 32768.
http://www.ozhardware.com.au/images/stories/Article%20Images/XPVISTACOMP/3dmbatchtest.jpg
This test demonstrates graphics card memory throughput by testing how many triangles per second can be drawn. In these tests both versions of Windows improve at close to the same rate as the batch size increases, until (as you can see on the graph above) Vista seems to hit a wall, indicating a big problem with memory throughput under Vista. Hopefully this is only a driver issue that can be fixed by nVidia in the future.
Onto PCMark05:
In PCMark, the tests were run at a resolution of 1280x1024, with the video cards in SLI mode.
http://www.ozhardware.com.au/images/stories/Article%20Images/XPVISTACOMP/pcmark.jpg
We can see an overall performance decrease of about 15% in Vista compared to XP, and when looking through the results, the main culprits are the graphics memory tests, where XP was a massive 11 TIMES Better than Vista!!! XP was able to pass through 3600frames per second compared to Vista’s lowly 312FPS. Again, hopefully future driver releases can help fix this issue, as that’s just pathetic.
Video playback of a WMV file also suffered greatly under Vista, giving just 30.8FPS compared to 69.1FPS under XP. This means Vista would struggle to do anything else while a WMV video is playing, quite poor really considering the hardware we’re testing on.
HDD performance was generally better on XP, but only by a few % barely worth a mention really. The only areas Vista excelled in were rendering transparent windows (i.e. like the Aero theme) and audio decompression. For transparent windows, Vista beats XP almost 4-fold, with 2671windows/sec under Vista, to 713windows/sec under XP. Microsoft has obviously optimised this process quite well for Vista to make the Aero theme run smoothly. Audio decompression was twice as good under Vista, working at 2665KB/s compared to 1371KB/s under XP. Unfortunately audio compression is the same under both, so you won’t be able to rip your CDs to MP3’s any faster.
Conclusion:
Vista definitely seems to have a few performance issues at the moment, yet certainly has the potential to be a great performer in the future. We can see from the few tests that Vista has been optimised for, it performed brilliantly, so when games, applications and drivers are made for DirectX10, we may see a big performance increase across the board, well we can hope can’t we?
If you’re into your games I’d hold off upgrading for now, until there is better driver support from the motherboard and video card manufacturers, and until games have DirectX10 support, as the visual improvements are meant to be a giant leap forward. For the meantime we can only hope that upcoming drivers will fix the issues that exist within Vista, but only time will tell.
Even before Windows Vista was publicly released, there was a lot of concern from people beta testing the O/S. Running Windows Vista seemed to have a large performance drop for their PC when compared to Windows XP. Was it true? Or is it just scare tactics from Microsoft haters?
It is widely known that you need a much faster PC to run Vista than you do to run Windows XP. Microsoft recommend a minimum of a 1Ghz CPU and 1GB of RAM to run anything better than Vista Home Basic (I.e. Home Premium, Business or Ultimate) , and a Directx9 capable video card. I would hate to try and run Vista on a 1Ghz CPU, it would just be painfully slow. So when specing-up your new machine, make sure you at least double the minimum recommended RAM and CPU speed, have a dedicated PCI-Ex16 video card with 128MB of RAM or more, and you will be fine. If your video card supports DirectX9, you should even be able to run Aero.
Vista may include a host of new features, look a whole lot better and tingle your inner geek in all the right places, but most people just want their games to run smoothly, their office applications to open quickly and run without fuss, and the internet to be as fast as possible, so why would you upgrade to an operating system that will serve only to slow you down? Well we’re going to run a few quick tests to either prove or dispel the rumours.
What we’ve done is setup a high-spec machine, and we’re going to do a standard load of Windows XP Professional SP2, run our benchmarks, then a standard load of Windows Vista Ultimate and perform the same tests. The graphics tests will be run using 3DMark06, which runs on DirectX 9, not DirectX 10 as is written into Vista, so whilst it won’t be as true a comparison of possible performance as we would like, it is the best we can do at the moment, and tests current 3D software, which is what matters. 2D application tests will be performed by PCMark05. Whilst these are both ‘artificial’ benchmarks as it were, they perform a wide variety of tests, with 3DMark simulating current and future game rendering systems, and PCMark simulating everything from rendering web pages, to 3D Performance, 2D Image rendering and Hard Disk usage, all in one neat package.
Test Machine Specs:
ASUS Striker Extreme nForce680i SLI Motherboard
Intel Core 2 Duo E6700 CPU
Corsair TWIN2X6400 2GB DDR-2 Memory Kit
2 x Gigabyte GeForce 8800GTX 768MB PCI-Ex16 Video Cards running SLI
Western Digital Raptor 150GB 10,000rpm SATA-II Hard Drive
OCZ GameXstream 850W ATX Power Supply
Certainly a bit higher-end than most home or office PCs, but this will allow us to run much more intense tests and be able to spot any differences more easily.
The Results:
For 3DMark06 we ran 3 different tests, one with the video cards in SLI mode, one using only a single video card, and another in SLI mode using 8x Anti-Aliasing to try to see if there’s more of a difference when the 3D system is pushed hard. All tests were performed at a resolution of 1280x1024.
http://www.ozhardware.com.au/images/stories/Article%20Images/XPVISTACOMP/3dmark06.jpg
As you can see above, there is around a 5% performance increase using Windows XP under both SLI tests, the single card test showed a negligible difference, but still in favour of old-hat XP.
There were a few tests within 3DMark that showed a massive difference between the O/S’s, specifically the shader tests. The simple Vertex Shader test showed almost a 50% drop in performance under Vista, with XP running 369Million Vertices/sec versus 186MV/s under Vista. The shader particles test also showed Vista had a 25% drop from XP, and the pixel shader test a 15% drop in performance.
3DMark also performs Batch Size tests of moving triangle systems, starting with 8 triangles and moving up with 32, 128, 512, 2048 and finally 32768.
http://www.ozhardware.com.au/images/stories/Article%20Images/XPVISTACOMP/3dmbatchtest.jpg
This test demonstrates graphics card memory throughput by testing how many triangles per second can be drawn. In these tests both versions of Windows improve at close to the same rate as the batch size increases, until (as you can see on the graph above) Vista seems to hit a wall, indicating a big problem with memory throughput under Vista. Hopefully this is only a driver issue that can be fixed by nVidia in the future.
Onto PCMark05:
In PCMark, the tests were run at a resolution of 1280x1024, with the video cards in SLI mode.
http://www.ozhardware.com.au/images/stories/Article%20Images/XPVISTACOMP/pcmark.jpg
We can see an overall performance decrease of about 15% in Vista compared to XP, and when looking through the results, the main culprits are the graphics memory tests, where XP was a massive 11 TIMES Better than Vista!!! XP was able to pass through 3600frames per second compared to Vista’s lowly 312FPS. Again, hopefully future driver releases can help fix this issue, as that’s just pathetic.
Video playback of a WMV file also suffered greatly under Vista, giving just 30.8FPS compared to 69.1FPS under XP. This means Vista would struggle to do anything else while a WMV video is playing, quite poor really considering the hardware we’re testing on.
HDD performance was generally better on XP, but only by a few % barely worth a mention really. The only areas Vista excelled in were rendering transparent windows (i.e. like the Aero theme) and audio decompression. For transparent windows, Vista beats XP almost 4-fold, with 2671windows/sec under Vista, to 713windows/sec under XP. Microsoft has obviously optimised this process quite well for Vista to make the Aero theme run smoothly. Audio decompression was twice as good under Vista, working at 2665KB/s compared to 1371KB/s under XP. Unfortunately audio compression is the same under both, so you won’t be able to rip your CDs to MP3’s any faster.
Conclusion:
Vista definitely seems to have a few performance issues at the moment, yet certainly has the potential to be a great performer in the future. We can see from the few tests that Vista has been optimised for, it performed brilliantly, so when games, applications and drivers are made for DirectX10, we may see a big performance increase across the board, well we can hope can’t we?
If you’re into your games I’d hold off upgrading for now, until there is better driver support from the motherboard and video card manufacturers, and until games have DirectX10 support, as the visual improvements are meant to be a giant leap forward. For the meantime we can only hope that upcoming drivers will fix the issues that exist within Vista, but only time will tell.