Posted By: Michael Kleef [MSFT] | May 20th @ 12:22 AM
Its interesting that people seem to think that Vista underperforms in every area of the system which is quite an incorrect perception. In this demo I show how Vista outperforms Windows XP and I show the under the covers process traces of just how it achieves it.
Demo environment consists of two images. One Vista SP1 and the other Windows XP SP2 both on the same HDD IO and communicating across my home wireless network to a Windows Server 2008 box on my main LAN. Once the two images get going latency gets introduced and things start to slow down....except Vista doesnt slow down. Watch and see!
Rating:
0
0

Mike, are both images the same architecture? (i.e. 32-bit for both, or 64-bit for both).

Not able to view the video in Opera or IE7. What software do you need to view it?

Anyway, this was a network copy test right?  I am also curious about the copy speed from one physical disk to another on the same PC (say, both internal SATA drive) as in backing up la few large files (~4-5 GB) and thousands of smaller files (~20 MB).  I wonder if there is a difference between Vista SP1 and XP SP3 in such a scenario.

Also, for sequential I/O such as copying/writing files, fragmentation does makes a big difference. On my home XP box with 4 drives (1 is a backup drive) , I've found it  very beneficial to keep the drives defragmented to obtain the best copy speeds as well as overall disk performance. Ofcourse, I have an automatic defragmenter running on it, so I don't have to schedule defrags regularly for 1.5 TB (used to be a PITA) and waste my time. I suspect many have fragmented files and then complain about poor copy speeds in Vista or XP without eliminating that problem.

I've found that Vista is faster for copying than XP even before SP1 but I'm still upset that the 32-bit version can't recognize the full 4GB of RAM I had installed. I really need it for video editing!

There is a great channel 9 video that talks about the improved TCP/IP stack in Vista/Server2k8 (Longhorn), unfortunately, I spent about 15 minutes searching for it without finding it.

It is important to note that the share you are copying to is on a Server 2008 box, you only touch on it briefly. If I remember correctly from the channel 9 video, Longhorn is required on both ends of the file copy to take full advantage of the new network stack. This is actually a test that I have been wanting to perform, I think it would be interesting to see the speed difference when copying from Longhorn-to-Longhorn, Longhorn-to-XP/2003 and XP/2003-to-XP/2003. I am also curious to see if the file transfer utilities in programs like VNC and PC Anyware will benefit from this. I imagine they will because this all happens at a lower layer, doesn't it?

I think that this has the potential to save great deal of time for companies that do a lot of transferring to remote sites over WAN links. Thanks for the great screencast Michael...


EDIT
 
The video i was thinking about was actually a edge video, watch it here. This channel9 video talks about it too..
Oh, and kestrel - Keith Combs has a small post on his blog where he talks about disk to disk copies, it can be found here and a update here.
Title is inaccurate. Correct one would be 'Windows Vista SP1 outperforms Windows XP SP2 in file copy in certain scenarios' because it's still few orders of magnitude slower in others. In some cases it is actually faster to write on ancient floppy disks in DOS than it is to write on modern USB keys in Vista. Fire IO department, grrr. Perplexed

http://channel9.msdn.com/ShowPost.aspx?PostID=390897