Defiler

Terrible SCSI performance in Windows XP

Recommended Posts

allnighte, If that's a raid board, you MUST disable the raid controller on the board for the scsi controller to work properly in my experience.

Everyone slse, the variour latency patches around, Via's included, change the default settings to attempt to maximize the settings for performance, so adding Powerstrip last won't show much if any gain, at least in my experience. I haven't tryed the Via raid patch as I've got my onboard raid disabled and am using the softraid in XP Pro. Getting around 70-80 reads and writes. I think my limiting factor is the PCI Bus, no windblows problem. My 19160 is in a 33/32 Pci slot. I'm waiting to get a dual board with a 66/64 bit slot to fully crank it up.

Good Luck,

Flint

Share this post


Link to post
Share on other sites
allnighte,  If that's a raid board, you MUST disable the raid controller on the board for the scsi controller to work properly in my experience.

Everyone slse, the variour latency patches around, Via's included, change the default settings to attempt to maximize the settings for performance, so adding Powerstrip last won't show much if any gain, at least in my experience.  I haven't tryed the Via raid patch as I've got my onboard raid disabled and am using the softraid in XP Pro.  Getting around 70-80 reads and writes.  I think my limiting factor is the PCI Bus, no windblows problem.  My 19160 is in a 33/32 Pci slot.  I'm waiting to get a dual board with a 66/64 bit slot to fully crank it up.

Good Luck,

Flint

not a RAID board

just your standard cheapo U160, single channel, 1 external and 1 internal connector, 64/66 pci compatible (but running in 32/33)

and i looked up the raid performance patch, and it looks like it's really only for IDE

Share this post


Link to post
Share on other sites

Allnighte, the RAID performance patch is not specifically for IDE, it was just developed as a response to many people with lower than expected performance with IDE RAID. I've tested and proven it helps even with a SCSI controller in my setup.

FredF, you have not yet reached the limits of 32/33 PCI if you're only getting 70-80MB/s throughput. My setup peaks at 120MB/s with software RAID 0 on my 4x 10k Fujitsu MANs running on the Adaptec 39160.

Share this post


Link to post
Share on other sites
FredF, you have not yet reached the limits of 32/33 PCI if you're only getting 70-80MB/s throughput.  My setup peaks at 120MB/s with software RAID 0 on my 4x 10k Fujitsu MANs running on the Adaptec 39160.

it is possable he is running into the via pci latency issue. i've got the same problem with my board. my bursts never go about 60-65MB/s, even with the pci latency patch and the via raid preformance patch.

Share this post


Link to post
Share on other sites

Hey Chew, I don't have your funding source and as such have only a single channel controller and 2 drives connected. Mines on a KT3Ultra ARU with 1gig of pc2700 with an XP1500+ running at 166mhzx10 or 1667 Mhz (XP2100+roughly). Maybe I'm being limited by the oc or something or the 4 partition raid versus the 2 partition raid I've got set up. Anyway, I haven't taken a lot of time trying to optomize it. Maybe I'll try to get up with you for some help when I have the time to screw with it.

Thanks,

FredF

Share this post


Link to post
Share on other sites

okay

when i get home, i first want to run ATTO with the smallest file size, to see if even cache/burst rates go above 45mb/s (if they don't.. i guess it's time for some researching)

then i'll give the raid performance patch a shot

i think i need to double check ALL my settings (bios, scsi card especially) to make sure i have it tweaked for performance

i also had a thought. for those running their scsi cards in a 64/66 slot, were they seeing perfect (ideal) results? i can't find someone that was running 64/66 and had this performance problem

though i don't know what that would mean if it were true :?

Share this post


Link to post
Share on other sites

:?

I believe the 11-02 Adaptec U160 SCSI drivers are just the official release of the 03-02 candidate; I just did quick look and don't see much difference in the files.

croc

Share this post


Link to post
Share on other sites

Is this an Ultra 160 problem only? I have a 29160LP and it seems to work fine (altough writes sometime seem to beat reads in ATTO). I have a POS LSI 10101-33 (free from Hypermicro :P ) that only gets about 16MBps read and write. Personally I don't give a hoot about the 16MBps because I rarely use that drive and I get 47MBps from EZ-SCSI.

GA-7DXC

384 pc1600

1.1Tbird

Radeon 8500LELE

Adaptec 29160LP (w/)

-Seagate X15-36LP (36GB)

-Atlas III (18GB)

-Fujitsu MAN3735 (72GB)

LSI 1010-33 (w/)

-IBM 36Z15 (36GB) [external]

Anyway I don't seem to have this problem at work. The Windoze XP workstations work fine when accessing out SAN. I'm not sure if any WinXP machines are on the SAN directly. We are running fourty 36GB 10k drives (@ RAID 5 I think) on four SUN Fire 4800s. SUN servers are maxed out with 96GB of RAM and 12 processors each. I think we are getting total of 107,000 IOPS on our normal load (50% usage). Dunno what max load is yet.

Share this post


Link to post
Share on other sites

So where's this patch that was supposed to be released by Microsoft? My SCSI performance still sucks and I'm getting tired of waiting. Has anyone got any news on this?

Thanks,

Celeron

Share this post


Link to post
Share on other sites

Please help. Want to place file scsiport.sys from SP2 W2000 to WinXP Pro, with purpose of attempt to increase Writes to SCSI disk (ATTO testing). After replacing existing driver in folders:

rootsystem32drivers

rootsystem32/dllCasche

rootServicePackFiles -

- WinXP always restore it back to native file. Tried to “unzip” SP1.cab then change file and “zip” it back. But Microsoft defend SP1.cab, maybe by using CRS of files.

Tried registry {HKLMSoftwareMicrosoftWindowsNTCurrentVersionWinlogon} SFCDisable=0 change to “1” by no avail. Who can advice other ways. I have WinXP on

FAT32 so with full access to system files. Read attentively all thread, but don’t find solution.

Share this post


Link to post
Share on other sites
Check out the post from another forum that I found today...

It claims that the benchmarks in 2K are wrong and not a performance difference between 2k and XP and now  XP now shows what it's supposed to be.

http://forums.2cpu.com/showthread.php?s=&t...&threadid=28860

Have a little faith, this is Storagereview.com. Several readers have reported timed file copies being significantly slower in Windows XP. There's no more accurate benchmark than reality.

Share this post


Link to post
Share on other sites

Yea... I don't know what the hell I was thinking even suggesting such a thing??? :oops:

I was actually posting something that I had seen to see if anyone else had heard such a thing. I have not but this is storagereview, you'd think that if there was an ounce of truth someone here might have seen it...

I thought I'd post it and see. I have done times copies too, then again I posted my time and no one ever posted back to confirm weather or not the time was a good one.

Becides timed copies ar not perfect either...

The copy time will only be as fast as the drive you are copying to so if you are copying to a slower drive than the array then the times will show that. Unless you load the file into memory you can't ever copy from a 100 or 133 hd to a 2 or 4 drive array at the speed of the array just can't be done... The only way to get a true timed copy is to copy from and to seperate arrays of the same size and stripe in the same system.

"g"

Share this post


Link to post
Share on other sites

I was using PCMark2002, not sure how great it is for a disk benchmark, but my 15K Cheetah was giving a very average score.

Someone posted this article on google, which seems to suggest the problem was fixed???

http://support.microsoft.com/default.aspx?...b;en-us;Q308219

So the verdict is that the actual performance is good, but the benchmark tools can't accurately report it??

Share this post


Link to post
Share on other sites

Look at last pages of the thread. Microsoft "repair" with new ntfs.sys (always included in SP1) don't work. More over I have the same problem in NTFS and in FAT32. Still want to try scsiport.sys from SP2 W2K.

Share this post


Link to post
Share on other sites
Look at last pages of the thread. Microsoft "repair" with new ntfs.sys (always included in SP1) don't work. More over I have the same problem in NTFS and in FAT32. Still want to try scsiport.sys from SP2 W2K.

A while ago I came up with this idea and thought I'd be the clever one and try it. I think I made a note of it somewhere in this huge thread. Makes no difference.

If you still want to try it, what you need to do is delete the file from the dllcache folder. I think it's a hidden folder somewhere in WINDOWSSYSTEM. Until this is done it keeps copying the file from here back over any system file that you try to copy over.

Share this post


Link to post
Share on other sites

To Chew,

look at my initial message - few posts above. WinXP don't permit you to delete or change *.sys files not in dllcache nor in system32driver folders. You think that you tried but WinXP restore back their original file after 2sec in driver folder BUT it restore the same file in dllcache folder longer - (5-10)sec. So you think that you something change but in real you testing ATTO with native WinXP SP1 SCSI driver. At first I was catch by the same.

Share this post


Link to post
Share on other sites
To Chew,

look at my initial message - few posts above. WinXP don't permit you to delete or change *.sys files not in dllcache nor in system32driver folders. You think that you tried but WinXP restore back their original file after 2sec in driver folder BUT it restore the same file in dllcache folder longer - (5-10)sec. So you think that you something change but in real you testing ATTO with native WinXP SP1 SCSI driver. At first I was catch by the same.

Not so. I was able to confirm I had successfully overwritten the file by checking the file version. I even rebooted the system in case the file in use was still the original version in memory somewhere.

As a sidenote, this was done on XP without SP1 installed, so perhaps the same method no longer works under SP1.

Share this post


Link to post
Share on other sites

To Chew,

I tried to overwrite this file in three above mentioned folders booting from DOS diskette but WinXP after rebooting restore it all back. Don't think it's SP1 poblem. I will try to reinstall my backup WinXPpro on IDE disk without SP1. We need to clear this situation.

Share this post


Link to post
Share on other sites

Good god, so after all of that, and even that knowledge base article which contains the supposed fix in SP1, the problem still exists?

Amazingly I went to Seagate's web site and they have a FAQ link about slow SCSI performance. Guess what the link is? YES! The same KB article I posted before.

Have we been duped?!?! Is this fixed in SP1 or not?!

Share this post


Link to post
Share on other sites

Good god, so after all of that, and even that knowledge base article which contains the supposed fix in SP1, the problem still exists?

Amazingly I went to Seagate's web site and they have a FAQ link about slow SCSI performance. Guess what the link is? YES! The same KB article I posted before.

Have we been duped?!?! Is this fixed in SP1 or not?!

Share this post


Link to post
Share on other sites

Dang, sorry for the double post. That wasn't my fault ;-).

I did have a question. Where were you guys getting this information about copying files from 2000 to XP and such? Is this documented somewhere?

Share this post


Link to post
Share on other sites

In my case, and probably vltk's as well, it was just an idea. I've read no documentation that suggests it. It just seemed that if the problem occurred in XP and not 2K, why not try to copy the 2K SCSI files to XP?

vltk, unfortunately I no longer have XP on any of my systems so I can't retest what you are trying.

Share this post


Link to post
Share on other sites

To Chew,

I made it. But it didn't help with SCSI perfomens under WinXP. You wre right. So we ought to wait for something from MS or try some other idea.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now