Defiler

Terrible SCSI performance in Windows XP

Recommended Posts

All three of my SCSI drives exhibit ridiculously bad write performance in Windows XP. (Yes, write caching is enabled.)

They're connected to an Adaptec 19160 card. At first I thought it must be the driver that ships in XP, so I tried the Windows 2000 driver. Same performance.

A WD 1200JB connected to the same system doesn't have any trouble turning in good ATTO numbers.

Check out this screenshot from an X15-36LP:

atto-xp.png

My X15 and Cheetah 18XL both seem to be "capped" at the same 12MB/sec. Nasty business. Same hardware in Windows 2000 didn't have a problem.

Does anyone have a suggestion? I'm tempted to swap in my Tekram Ultra160 card and see if it solves the problem.

Share this post


Link to post
Share on other sites

If you have a look around in the forum you will notice that many other members have comparable problems under Winxp (with any SCSI controller which excludes a driver probkem). The question: is it ATTO or does it have something to do with the OS itself.

While ATTO reports appropriate results with my IDE drives (under XP) this is what i get with my Atlas 10KIII (on Tekram DC-390U2B):

ATTO10KIII.jpg

I gave it a try with ACPI disabled and i still get the same results while everything is fine in WinME.

Share this post


Link to post
Share on other sites

Windows XP Professional

Athlon XP 1800+

EpoX 8KHA+ mainboard (Via KT266A-based)

Adaptec 19160 v3.10

etc.

I tried forcing the "write-back cache" setting in the card's BIOS to "Yes", but it didn't solve the problem. Grr.

Share this post


Link to post
Share on other sites

I don't think this is merely a benchmarking artifact.

I used Total Copy to watch the transfer speed as I copied a 270MB file.

From X15-36LP to WD 1200JB: 37MB/sec or so

From 1200JB back to X15-36LP: 12.5MB/sec or so

Seems like the performance problem is "real", not just an ATTO issue.

Share this post


Link to post
Share on other sites

Well, from my D540X Raid0 array to the Atlas 10KIII i just get 8mb/sec with Total Copy while i have 33mb/sec with Explorer. In case you have an explanation i am open to suggestions...

Share this post


Link to post
Share on other sites

By the way SR uses an XP testbed. A performance degradation does not seem to come out in the various benchmarks results.

A significant difference relatively to my own system is that the testbed is formatted with NTFS while i use FAT32. On the other hand i can hardly imagine that this would be the root of the problem.

Share this post


Link to post
Share on other sites

Laurent.. First you say that it doesn't seem to have anything to do with drivers/SCSI card, then you post a screenshot of your system unaffected by this issue.. Can you explain this apparent contradiction? I'm ready to blame the interaction between Adaptec's card and WinXP.. I guess I'll know more when I swap in the Tekram..

Share this post


Link to post
Share on other sites

I am not sure i understand.

Did you notice my transfer rates on small blocks ? Even my D540x Raid0 array transfers up to 10times more data with these. While your screenshot shows poor writes only it seems i have a mix of both. By the way you should really have a look around in the forum; i am sure the other ATTO screenshots posted by other members are still available.

Share this post


Link to post
Share on other sites

I see what you mean. I tend to ignore the tiny-block ATTO results, since they vary so much. I'll trawl the forum for more ATTO on XP tests.

Thanks.

Share this post


Link to post
Share on other sites

My ATTO scores on Win XP are also low. This is on a 3Ware 6200 with two IBM 75 GXPs. (FAT32)

I'll run ATTO from the Win98 dual boot as soon as get a chance.

Bozo :D

Share this post


Link to post
Share on other sites

Hi,

Don't know if this is the problem with your system, but when I installed my Adaptec 2940U2W with an Atlas 10k III, I had same kind of problems, low write performance. Naturally I tried enambling write cache for the drive -> No help. Then I noticed the write cache checkmark keeps popping of every time I reboot. Checked 2940U2W bios, Write cache was set on.

When I set write cache setting to default (NC or auto or something like that) and booted to windows, I could set the write cache on so that it sticks. And now my results are OK.

Are there some issues with Adaptec card forcing write cache on or is it just my setup?

Hege

Share this post


Link to post
Share on other sites
My X15 and Cheetah 18XL both seem to be "capped" at the same 12MB/sec. Nasty business. Same hardware in Windows 2000 didn't have a problem.

Does anyone have a suggestion? I'm tempted to swap in my Tekram Ultra160 card and see if it solves the problem.

Just curious, why did you choose the Adaptec over the Tekram as your primary SCSI card? Was it performance or driver related? I'm thinking of buying a new SCSI controller.

Share this post


Link to post
Share on other sites
Just curious, why did you choose the Adaptec over the Tekram as your primary SCSI card?  Was it performance or driver related?  I'm thinking of buying a new SCSI controller.

Neither. I needed an external 50-pin port.

Share this post


Link to post
Share on other sites

atto.jpg

Original X15 on a 2940U2W on a Tiger MP with dual XP1500+, almost full C: in NTFS, fairly fragmented (if that makes a difference).

My writes seem fairly consistant with reads....

Share this post


Link to post
Share on other sites

I have absolutely the same thing, but not only with SCSI (also with IDE RAID). The read speeds are just where they should be, but the writes speeds are dismal.

Apparently, this is a problem with ATTO. Another thing to try (which I can't) is to delete all your partitions and try running ATTO on the unpartitioned drive. ATTO even has disclaimer that it doesn't guarantee correct results for write performance on a partitioned drive.

Leo

Share this post


Link to post
Share on other sites

Just curious, why did you choose the Adaptec over the Tekram as your primary SCSI card?  Was it performance or driver related?  I'm thinking of buying a new SCSI controller.

Neither. I needed an external 50-pin port.

Tekram has an external 50-pin port. It comes as an extension card.

Leo

Share this post


Link to post
Share on other sites
I have absolutely the same thing, but not only with SCSI (also with IDE RAID). The read speeds are just where they should be, but the writes speeds are dismal.

Apparently, this is a problem with ATTO. Another thing to try (which I can't) is to delete all your partitions and try running ATTO on the unpartitioned drive. ATTO even has  disclaimer that it doesn't guarantee correct results for write performance on a partitioned drive.

Leo

On a second thought, my guess might be wrong. I'll try Total Copy... But if the problem is real, then I wonder what is the common denominator? You have a problem with Adaptec, I have a problem with Tekram and Escalade, other don't have problems with either. Hence, the problem is probably not with the controllers.

Windows XP? There are people who do not have the problem under Windows XP (if I understand correctly). So, how about we tell each other the motherboard brand we're using? Mine is Asus A7N266-E.

Leo

Share this post


Link to post
Share on other sites

Mine is posted above:

EpoX 8KHA+ (Via KT266A-based)

I'm going to plug in the Tekram controller tonight. I'll report back with the results.

Share this post


Link to post
Share on other sites

I have the same problem, I thought it was just me earlier. I mention somthing about this in an earlier thread, but not much feeback. I have a 2940UW, with an Ultrastar 18ES. I too get crappy ATTO results. 3MB/s read!!! When I disabled the cache, it went up to 19MB/s.

Strange. I want to go back to Win2K Pro, to see if I have any problems. I think it is XP related, because now, with XP, I get lockups. When I check the event log, it shows aicxxx errors everywere. Obvoiusly SCSI related. (termination, and term power, and cables are all good). With Win2K, I NEVER had that problem.

I am going to try Win2K tonight, or tomorrow. I'll let you guys know my findings......

Bud

Share this post


Link to post
Share on other sites

Defiler, a good idea would be to use a substitute benchmark. I suggest you give a shot to Sandra (yes Sandra) and look at the individual sequential write score . It usually provides very reliable results.

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