EDIT: Updated with Doom 3 results.
I just reinstalled Vista Home Premium 64-bit and thought I'd run some tests. These are just a bunch of quick benches I did to satisfy my own curiosity, but I thought I'd share them with you guys. The test config is as follows:
Core 2 Duo E6600 @ 3.2GHz (8*400)
Gigabyte GA-P35-DS3
8GB DDR2-800 4-5-4-15
GeForce 8800GT 512MB (stock)
Western Digital Caviar SE16 250GB <--- Vista Home Premium SP1 RC 64-bit
Samsung SpinPoint T166 320GB <--- Windows XP Professional SP2
[See the end for some comments on the results.]
The tests:
File copy (6.3GB ISO from Samsung drive to WD drive):
Vista: 2min 48sec
XP: 3min 07sec
File extraction (extracting the Crysis patch zip file with WinRAR):
Vista: 14s
XP: 13s
File extraction (extracting 3.9GB RAR archive file using WinRAR):
Vista: 2min 16sec
XP: 2min 22sec
File extraction (extracting 1.7GB ISO file using 7zip):
Vista: 2min 05sec (64-bit version of 7zip)
XP: 2min 18sec (32-bit version of 7zip)
Program load times:
Vista:
Photoshop CS3: 2 secs
OpenOffice: 1.5 secs
Crysis: 26 secs
XP:
Photoshop CS3: 8.5 secs
OpenOffice: 6.5 secs
Crysis: 33 secs
Photoshop CS3 "Retouch artist's benchmark":
Vista: 30s
XP: 29s
3DMark06
Vista:
3DMarks: 11297
SM2.0: 5227
SM3.0: 4942
CPU: 2772
XP:
3DMarks: 11706
SM2.0: 5391
SM3.0: 5150
CPU: 2869
Call of Duty 4 (1280x1024, MAX settings, except no AA and AF slider at ~half):
Vista: 104fps (avg), 54fps (min), 255fps (max)
XP: 102fps (avg), 56fps (min), 274fps (max)
Since CoD4 contains no built-in benchmark, I had to measure with FRAPS and play the same sequence a few times to get somewhat reliable results. I played from the beginning of the level "The Bog" and until you pick up the Javelin, approximately 3 minutes of gameplay. The results were very consistent, so they're probably pretty accurate despite the inaccurate metodology.
F.E.A.R. (1280x1024, MAX settings, 4xAA, 16xAF):
Vista: 97fps (avg), 48fps (min), 242fps (max)
XP: 95fps (avg), 45fps (min), 242fps (max)
Doom 3 (1280x1024, Ultra quality, no AA, no AF):
Vista: 181 fps
XP: 191 fps
Doom 3 (1280x1024, Ultra quality, 8xAA, no AF):
Vista: 54 fps
XP: 55 fps
Crysis GPU-test (1280x1024):
Vista (”High”, DX10, 64-bit): 35fps (avg)
Vista (”High”, DX9, 64-bit): 37fps (avg)
Vista (”High”, DX10, 32-bit): 35fps (avg)
Vista (”High”, DX9, 32-bit): 36fps (avg)
Vista (”Very High”, DX10, 64-bit): 20fps (avg)
XP (”High”): 39fps (avg)
Crysis CPU-test 2 (800x600):
Vista (”Low”, except ”Physics” on high, DX9, 32-bit): 60fps (avg)
Vista (”Low”, except ”Physics” on high, DX9, 64-bit): 54fps (avg)
XP (”Low”, except ”Physics” on high): 69fps (avg)
Comments:
It's interesting to see that Vista's performance seems to have progressed since it was released. The general usage tests are either faster or equal to XP. It should be noted that the program load times may be influenced slightly by the fact that the OSs reside on different drives (Samsung/WD). However, Vista is on the slower drive (the WD), so it only makes the results even more impressive. Talking about the program load times, it's obvious that Vista's SuperFetch feature works wonderfully. Even though I've only launched Photoshop and OpenOffice a few times, Vista has already picked this up and loads them into mem right after boot up. Crysis was almost certainly not cached into RAM during the test, but still loaded faster than in XP.
The horrendous file copy performance has been fixed. Performance is definitely higher than in XP. Also gone is the sometimes irritatingly long file delete times. File deletion seems instantaneous now, just like it's always been in XP.
Now, we've come to the less stellar part of Vista's performance: Gaming. 3DMark06 did perform okay, with numbers within 3-4% of XP. Although a performance decrease is never welcome, this is tolerable. Moving on to Crysis, however, things look a little more bleak. Performance is down 5-8% when looking at DX9 32/64-bit benches and DX10 is even worse off. The CPU test is also curious with a WinXP performance lead of 15%! This was definitely unexpected and something that would be interesting to research further.
EDIT: With the addition of the Call of Duty 4 results, things are more of a mixed bag. CoD4 actually performed very well and Vista was able to match the performance of XP. It's clear that more data is needed to come to a conclusion here. From the two games I've tested, Vista does atleast seem to do an "okay" job.
EDIT: Okay, with F.E.A.R added to the mix we can see that Vista doesn't seem all that bad for gaming. Not much else to say, really. The more data the better and with these three tested games, Vista performance seems good enough.
I haven't really benchmarked enough apps to show any clear trends, but it seems general Vista performance is very good and caching to RAM is beneficial. Overall feel of Vista is very good and subjectively an improvement over its performance at release. Thus far, it feels slightly faster than XP, even when stuff is not cached into memory. This is not just imagined either. Loading the image for the Photoshop test was about twice as fast in Vista compared to XP and it was the first time that file was loaded so it couldn't have been cached. With all that said, gaming performance is still a let-down and it seems we'll have to live with it.
Hope you all appreciated this and found it interesting! Thanks for reading.
------------------------------------------
EDIT: Vista has a feature called SuperFetch, which treats the system RAM as a giant disk cache. Whatever memory not used by another application gets stuffed with data that you use frequently, so that it can be loaded quickly if you need it. Some people believe this feature to be useless, so I set out to test this too. Results below (in seconds):
http://www.etek.chalmers.se/~brunnis/superfetch.png
As can be seen, SuperFetch has a very large impact on program load times.
