cas

Get your dangerous XP Cache Filter here

Recommended Posts

Forgive my ignorance, but how do I install the xpwritecache fllter??

I downloaded and then run it but it does not do anything, I also checked on device manager non-plug and play drivers and it doesn't apper there.

Share this post


Link to post
Share on other sites

run it from a command prompt (instead of double-clicking it) in order to see the output. if you have problems installing it, see the advanced instructions in this thread.

Share this post


Link to post
Share on other sites

Would XP Cache Filter improve my write performance:

Tyan Thunder K7 board

LSI Logic MegaRAID 320-2 SCSI RAID controller

All Seagate 36GB 10.6K U320 drives

RAID 1 (one on each channel)

RAID 5 (six total, three on each channel)

Dual 2100+ Athlon MPs

1.5 GB of RAM

Adapter is in 64/33 slot

Is their a new version of XP Cache Filter that has improvements in it?

Share this post


Link to post
Share on other sites
Would XP Cache Filter improve my write performance:

Probably not. RAID controllers such as yours with controller cache typically disable the write cache on the drives anyway, and use the caching functionality of the controller instead. XPCacheFilter improves performance on SCSI drives that have write cache enabled, by filtering out commands to bypass the write cache.

Share this post


Link to post
Share on other sites

No change. I am going to remove it tonight. I guess I keep on the Microsoft tech and LSI Logic to get me a solution. My next step is replacing my Tyan Thunder K7 with a Supermicro X5DAE.

Share this post


Link to post
Share on other sites

Hi

What program are you guys using to benchmark your drives? I would like to test this patch out myself but haven't found a program that shows read/write data only thorough output. Also, here is my configuration do you think it will help me? I have noticed that my SCSI drive always benchmarks lower than my standard ATA drive and my SATA drive. This is aggravating since my SCSI drive is Ultra 320 and 15K RPM...

Windows XP Pro

P4 3.0

SCSI Ultra 320

1 GIG 400 Mhz DDR dual channel

Share this post


Link to post
Share on other sites

there are a few things being discussed here:

meta and file data synchrosity

scsi cmd w/ forced write

drive write cache

raid controller cache

if drive write cache is enabled, the os must use write barriers (eg, ordered cmd tags). last time i checked, linux doesn't do this (there was a big thread about this on lk). that's why my drives have write cache disabled.

raid controllers (often) have battery backed cache onboard. in that case, drive cache should be disabled (for maximum reliability). the raid ctrl will take care of flushing its cache. that's how raid ctrls get a big win on writes over software raid.

if your drives have write cache enabled and the os doesn't support write barriers, or some driver disables it somehow, then you're at greater risk of corruption if the power is suddenly cut.

(i think i've covered everything)

Share this post


Link to post
Share on other sites

Please significantly reduce the size of that signature graphic. It is larger than most posts! :)

Share this post


Link to post
Share on other sites
Is it possible that sata raid controler affect with same problem ? System sees sata controlers as scsi devices.... and currently i have strange problem with WinXP and performance ( more http://forum.abit-usa.com/showthread.php?s...40&pagenumber=1 ). Any advice ?

If you read the following FAQ entry

XpScsiProblems

you'll see the issue this driver resolves is the reduced performance of writes due to the now correctly functioning requests to bypass write caching. This doesn't sound related to your problem.

To the specific question of whether SATA RAID controllers would be affected, it's possible but only if they support the relevant commands which is very unlikely.

Share this post


Link to post
Share on other sites

I already readed that article. And its very possible its related problem i though its about read speed..... but when i disabled "write cache" list was makeing not 30 but 4 mins. So it might be problem with writing. Now if its affect sata raid controlers that are seen by system as scsii devices were stucked :(

Share this post


Link to post
Share on other sites

Well, there's a checkbox to enable write caching, ticked by default on my system, but if you want to know how it affects SCSI performance, someone with SCSI, who was previously experiencing the write performance issues in certain applications and benchmarks, will have to report back for us.

Share this post


Link to post
Share on other sites

Hey folks,

It's been a VERY long time since I have posted on this board (although, I read the forum just about every day). I used to be a member before the board database went down, and everyone had to redo their profiles. That was back in the day when Tannin, Supercaffeinated, and Clocker used to rule the roost. Are these guys still around?

Anyway, I thought the whole XP SCSI issue was resolved? At least, it has been for me since SP2 came out. Somebody on this forum posted the fix on one of the 'several' thread topics on this issue. I don't know which one, since a few of the threads are a bazzillion pages long (and quite frankly, I don't feel like sifting through them to find it). The user posted it, but nobody replied back. I guess everyone was pretty sick of the topic.....or, everyone just missed it. Anyway, I work on quite a few SCSI-based systems, all of which ran XP Pro, and all have completely different hardware (ie....different MB's, SCSI cards, SCSI drives, etc.). As in the case with most of you SCSI users out there, I was extremely frustrated with the SCSI bug. I know that several "schools of thought" regarding this issue have been bounced around on this board for quite some time. Some believe it's an aweful mistake made by Microsoft with the "Power Protected Write Cache Option" and the whole "FILE_FLAG_WRITE_THROUGH" issue.......causing horrible write-performance (the SCSI bug also rears its ugly head in Win2k SP3). Others believe that it's not really a bug at all, that it doesn't make any noticeable difference in performance, and that it's only a benchmark anomaly. In my case, it was DEFINITELY an EXTREMELY frustrating bug!!! A benchmark anomaly is one thing, but having degraded "real-world" application performance is a totally different ball of wax. My biggest pet-peave was HORRIBLE write-performances when transferring files in Explorer (by testing with a stop-watch). You know something is wrong when an IBM 75GXP kicks the pants off a Maxtor Atlas 15k in write performance. Switching to Dynamic Disk worked on 'some' of my systems, but didn't seem to be a real "fix" to the issue (for a single drive, I'm not a big fan of Dynamic Disk anyway). In addition, Cas's filter only worked on a few of my systems.

I can attest to the fact that this was in fact a performance problem......and the bug was in fact for real. As with all you SCSI folks, I hoped and prayed that MS would come out with a fix in XP SP2. Well, they did......sort of. I'm sure all of you have read the MS article Slow Disk Performance When Write Caching Is Enabled . It informs us of a hotfix tool that will deal with the bug in Win2k SP3 (can be found here: Obtain the Dskcache.exe Tool to Configure the "Power Protected" Write Cache Option ). Just for grins and giggles, I tried this tool on XP SP1, and of course, it didn't work. I then gave up, and reverted all my work systems to Win2k SP2, just to get the full capabilities of the SCSI hardware. When XP SP2 came out, I was very dissapointed that MS didn't include an option for the fix (as they did with Win2003). Then, an SR user posted that they tried the Win2k hotfix on XP SP2, and it worked. On my own personal system (which was running XP SP1), I installed SP2, and then retried the Win2k "Power Protect Write-Cache" tool from MS.......and IT WORKS!!!!!!!! With SP2, it now allowed me to disable the power protection.......and the write performance kicked butt. It worked both in the benchmarks, and in real-world applications (via comparing with a stop-watch). The difference was phenomenal. I saw amazing peformance gains in ATTO and IOMETER after the hotfix. It makes me wonder how the fix would effect SR's benchmark scores on SCSI drives......hmmm. I thought that I might have just gotten lucky with my system.....however, I re-installed XP on all my work machines.....then installed SP2. It worked on EVERY SINGLE ONE of those systems!!!! The stop-watch test also worked on those systems (in addition to huge benchmark increases). It basically brings the XP SCSI write performance back up to par with Win2k SP1 & SP2. All of these systems have been running with the fix for quite some time, without any problems. Again, the Win2k hotfix worked for ALL my SCSI-based systems, but of course, I can't guarantee it will work for you. Use it at your own risk, and don't blame me for any problems you might run into!!!

Obviously, the draw back to all this is......in the case of a power failure......there is a good chance that you may lose data. I HIGHLY recommend the use of a UPS battery backup with PC connectivity (which will automatically sefely shut the computer down in case of a power failure).

SCSI is back baby!!!!!

CC

Share this post


Link to post
Share on other sites
Guest Eugene
It makes me wonder how the fix would effect SR's benchmark scores on SCSI drives......

As has been explained in the past, the primary SR single-user measurement, the DriveMarks, are not affected by the SCSI "bug" because of the way they were created. Results do not significantly change whether run in Win Server 2003 with the appropriate settings, nor Windows 2000 SP2, nor with any XP hacks.

Share this post


Link to post
Share on other sites

Hey Eugene,

I would definitely have to agree. After I thought about it for a second, my inclination was that the fix won't help SCSI drive scores with SR's benchmarks, since we are talking about "accesses" via WinTrace32 captures. The XP SCSI bug only effects writes, not reads......therefore, there would be no differences in benchmark scores after the fix was applied.

Cobra Chi

Share this post


Link to post
Share on other sites

Although, I'll have to admit, after applying the fix........my applications defininetly feel MUCH snappier, my computers boot faster, and transferring files across drives is now AMAZING!!!

CC

Share this post


Link to post
Share on other sites

Hmm, I found a dskcache.exe that self-extracts. It was in an old digest that another member hosted for us. I tried this link and ran the exe; it actually worked for me (XP PRO SP2).

http://www.baldingape.net/Q811392.zip

Here's the abbreviated results:

D:\>dskcache +p

Current settings:

Write Cache is enabled

Power Protect is disabled

New settings:

Write Cache is enabled

Power Protect is enabled

...

Did M$ give us a freebie fix and not tell anyone? :)

This comes just before I have to do lots of digital scanning this month. (Good timing).

Share this post


Link to post
Share on other sites

Hey Krusher,

That is the tool I'm referring to. It definitely works for XP SP2. It was released by Microsoft to allow one to change the power protect feature of Win2k SP3. I didn't have any success with it when I tried to apply it to XP SP1.......but it has worked on every XP SP2 machine I've tried. Funny thing is, MS's documentation is very confusing when it comes to this tool. In addition, I think they "might" have changed the wording in these documents since XP SP2 was released (If my memory serves me wrong folks, please don't hesitate to correct me). First, the Slow Disk Performance When Write Caching Is Enabled article states:

"This fix is included in Windows 2000 SP3. A similar fix is also included in Windows XP."

Before SP2 was released, I'm pretty sure it didn't used to say "A similar fix is also included in Windows XP". But in the same article, they state:

"Because of design differences between Windows 2000 and Windows XP, there are currently no plans to release a supported update for this behavior for Windows XP."

A contradiction of words maybe? If they did in fact change the wording of these articles after SP2 was released, then they should have nixed that statement. In the next article Obtain the Dskcache.exe Tool to Configure the "Power Protected" Write Cache Option it states:

"In Windows XP, the Power Protect mode is not displayed in the output of the Dskcache.exe tool."

Funny, I don't remember that XP blurb being there before SP2 was released. They go on to say:

"Therefore, if your system is not severely affected by the lack of this feature, Microsoft recommends that you wait for the next Windows 2000 or Windows XP service pack that contains this feature."

I remember people emailing MS about this, and MS first claimed that they 'might' fix it in SP2, but other people have said that MS stated that they do not plan to fix this with XP......only win2k. At the bottom of all these articles, there is a section to show which operating systems the hotfix applies to. Call me crazy, but I don't remember XP being listed in any of these sections before SP2 was released.

Finally, in the article List of fixes included in Windows XP Service Pack 2 we see "811392 Obtain the Dskcache.exe Tool to Configure the Power Protected Write Cache Option" being listed. So there ya go......they did in fact include it with SP2. This contradicts what they said before about "there are currently no plans to release a supported update for this behavior for Windows XP." They kinda just slipped it in there, then "perhaps" changed the wording of some articles to cover their tracks. I think it's a conspiracy :ph34r: ......hahaha.....just kidding. I really don't care if those articles are a little confusing or misleading......I'm just glad they finally gave us an option to fix the bug.

CC

Share this post


Link to post
Share on other sites

Oh yah, I think I remember some of you ATA RAID folks out there being affected by the XP bug (ie....low write scores), since XP lists ATA RAID cards as a SCSI device in the device manager. I wonder if the hotfix will work for you guys?

CC

PS - remember, using this hotfix.....a power failure will most likely result in data loss. I highly recommend a UPS battery backup.

Share this post


Link to post
Share on other sites

The SP2 + dskcache patch did not fix my problem. I am still experiencing slow write speeds when I enable write-back caching; compared to write-through under atto. From all the information that I have heard, write-back should provide a definite improvement over write-through. I am switching between both settings under storage manager 3.14. My setup is as follows: 2x 15k.3 Seagate drives on a RAID 0 array connected to an adaptec 2100s card.

Any thoughts?

Share this post


Link to post
Share on other sites

I had to register/post a reply given this situation has been my nightmare for over 1.5 years now. It just didn't figure a PC that was fast as heck with Win 98se and $400+ worth of UW hardware performed horrible as soon as you tossed XP on it. I've gotten the run-around for months from MS, other forums and locals. Done a million things, converted disks basic/dynamic.. all to no avail.

I couldn't get the XpCacheFilter tool on the onset of this thread to install properly- on XP Pro/SP2, running the -i and rebooting never had it show up in Device Manager/Non Plug-n-Play section.. and of course, no improvement occurred.

The dskcache.exe DID solve my problem! I'm so grateful for this information!

To give some indication of the gain for me:

ATTO Write before/Write After

64.0: 11335 -> 23011

128.0: 13687 -> 23414

256.0: 12157 -> 21970

512.0: 12041 -> 23910

1024: 13784 -> 23519

Sandra's Filesystem benchmarks:

Before -> After

Buffered Write: 10 MB/s -> 27 MB/s

Sequential Write: 11 MB/s -> 28MB/s

Random Write: 10 MB/s -> 24MB/s

Gains in how the system performs are huge. Pagefile access is normal speed again! Defrag runs in 1/2 the time as well.

And as far as "power safety"- I've only had power fail once and only crashed/froze up once as well. In both cases, the drive was totally FUBAR with all that protection crap on (disk repair tool and chkdsk wouldn't restore stability first trial)... so I'm seeing NO drawback to this. I've gotten used to Dynamic Disk/NTFS basically meaning- "Reformat and lay data back down" in reality anyways, so it's of no circumstance.

Thanks again guys and I hope you will help spread the word to other SCSI owners.

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