Hello,
I'm experiencing a strange issue with my computer. I'm using 2 10K SCSI drives (Seagate Cheetah 10k.6, IBM Ultrastar 36lzx), connected to the onboard SCSI controller of a Supermicro X5DA8 on an XP Pro box. It is known there are issues with the SCSI implementation on XP ( http://forums.storagereview.net/inde...showtopic=7228 ).
Now, I'm trying to find out if my system performs as expected. I ran the virtualdub benchmark (in auxsetup.exe). These are the results for the IBM:
1. Windowsbuffer enabled (checkbox unchecked, default)
peak disk performance (50 MB)
sustained read: 544681 KB/s
sutanined write: 8512 KB/s
videocapture performance:
640x480x16
200 frames
50 buffers
FPS = 25.000
(yields a datarate of 15000)
=> 12 % dropped frames
2. Windowsbuffer disabled (checkbox checked)
peak disk performance (50 MB)
sustained read: 18510 KB/s
sutanined write: 19738 KB/s
videocapture performance:
640x480x16
200 frames
50 buffers
FPS = 25.000
(yields a datarate of 15000)
=> 0 dropped frames
I then tried with XP Cachefilter (got it from Storagereview):
1. windows buffering enabled
peak disk performance (50 MB)
sustained read: 465455 KB/s
sutanined write: 21843 KB/s
2. windows buffering disabled
peak disk performance (50 MB)
sustained read: 18208 KB/s
sutanined write: 28240 KB/s
Both cases yield 0 dropped frames
The differences are similar when benchmarking the Seagate (without windowsbuffering and with the XP Cachefilter, it claims a writespeed of about 54000 KB/s ! )
Does anyone have experience with this XP Cachefilter Tool ? How big is the impact regarding possible dataloss ?
(should I get me a UPS ? )
My system seems to perform better without the Windowsbuffers. Are there ways to disable it to give me better write performance in Windows overall ? Or are there reasons not to do this ? Does it perhaps have tweakable settings ?
What else can I do to get proper results ?
Thanks!
Jörg
I'm experiencing a strange issue with my computer. I'm using 2 10K SCSI drives (Seagate Cheetah 10k.6, IBM Ultrastar 36lzx), connected to the onboard SCSI controller of a Supermicro X5DA8 on an XP Pro box. It is known there are issues with the SCSI implementation on XP ( http://forums.storagereview.net/inde...showtopic=7228 ).
Now, I'm trying to find out if my system performs as expected. I ran the virtualdub benchmark (in auxsetup.exe). These are the results for the IBM:
1. Windowsbuffer enabled (checkbox unchecked, default)
peak disk performance (50 MB)
sustained read: 544681 KB/s
sutanined write: 8512 KB/s
videocapture performance:
640x480x16
200 frames
50 buffers
FPS = 25.000
(yields a datarate of 15000)
=> 12 % dropped frames
2. Windowsbuffer disabled (checkbox checked)
peak disk performance (50 MB)
sustained read: 18510 KB/s
sutanined write: 19738 KB/s
videocapture performance:
640x480x16
200 frames
50 buffers
FPS = 25.000
(yields a datarate of 15000)
=> 0 dropped frames
I then tried with XP Cachefilter (got it from Storagereview):
This program installs a small filter driver, which clears the FUA (Force
Unit Access) bit from SCSI write commands. This has the potential to
improve write performance, at the expense of greater risk in the case of
disk or power failure.
Unit Access) bit from SCSI write commands. This has the potential to
improve write performance, at the expense of greater risk in the case of
disk or power failure.
peak disk performance (50 MB)
sustained read: 465455 KB/s
sutanined write: 21843 KB/s
2. windows buffering disabled
peak disk performance (50 MB)
sustained read: 18208 KB/s
sutanined write: 28240 KB/s
Both cases yield 0 dropped frames
The differences are similar when benchmarking the Seagate (without windowsbuffering and with the XP Cachefilter, it claims a writespeed of about 54000 KB/s ! )
Does anyone have experience with this XP Cachefilter Tool ? How big is the impact regarding possible dataloss ?
(should I get me a UPS ? )
My system seems to perform better without the Windowsbuffers. Are there ways to disable it to give me better write performance in Windows overall ? Or are there reasons not to do this ? Does it perhaps have tweakable settings ?
What else can I do to get proper results ?
Thanks!
Jörg
Comment