cas

Get your dangerous XP Cache Filter here

Recommended Posts

Cas, what do you think about all this. Is it possible to write some fix to make disk.sys behave the same for SCSI as for IDE foer disk-to-self copying. Or may be you have some other ideas.

I really don't know.

Without having independently confirmed the issue, I would guess that differences in the way requests were broken up in the disk or port drivers, could have some effect on the number of seeks required for a disk to self copy.

You may want to try cascopy to set a baseline.

Test with cascopy. It’s really amazing. Look at the results:

WinXPpro

Cascopy and Exploer

(SCSI) C:! to C:=38s (~26.9Mb/s) C:! to C:=3m04s (~5.6Mb/s)-Cascopy 4.5 times quicker

(IDE) D:! to D:=72s (~14.2Mb/s) D:! to D:=2m25s (~7.1Mb/s)-Cascopy 2.0 times quicker

D: to C: =32s (~31.5Mb/s) D: to C: =34s (~30.8Mb/s)-near equal

Now copying with cascopy we at last have normal ratio for C:/D:=26.9/14.2~1.9 throughput.

But how to implement this into WinXP copying mechanism.

Share this post


Link to post
Share on other sites
vltk

I'm wondering if your problem is swap file related.  cas could probably advise if I'm off the mark here, but I believe copying files involves the swap file, so if the swap file is on the disk that's performing the same-disk copy it may impact performance?  Try moving your swap file to the IDE disk and run the same tests.  See how much impact it has on both the SCSI and IDE results.

Tried to enable WinXP swaping and test with swap file on disk c: and dick d:. Have the same results as without swapping.

Share this post


Link to post
Share on other sites

I am having a really hard time reproducing croc’s problem. Despite running a number of memory mapping tests, I have not yet identified any failures.

If you are using the newest version of the filter, and are aware of any problems that you can reliably toggle by installing and uninstalling the driver, please let me know.

Share this post


Link to post
Share on other sites

I measured throughput with memory utilized for system cache (default checked for applications). And have interesting results. Nothing change copying 1Gb file with cascopy. But I receive much better results for SCSI and sufficiently better results for IDE throughput with explorer.

File copying with Explorer.

SCSI:

C:! to C:=3m04s (~5.6Mb/s) – Application checked (was).

C:! to C:=2m16s (~7.5Mb/s) – System cache checked (now).

IDE:

D:! to D:=2m25s (~7.1Mb/s) – Application checked (was).

D:! to D:=1m56s (~8.8Mb/s) - System cache checked (now).

File copying with Cascopy – for comparing.

C:! to C:=38s (~26.9Mb/s)

D:! to D:=72s (~14.2Mb/s)

Again 7200 IDE disk quicker then 15K.3 SCSI one. But both results improved. ATTO results improved also.

SCSI W=R=59.4Mb/s compare to former 58.7Mb/s

IDE W=R=35.5Mb/s compare to former 33.0Mb/s

Before each test I made restart to clean memory and system cache (have OS swapping disabled).

Cas, I consciously choose FAT32 partition for both SCSI and IDE disks. So I don’t need to bother about NTFS files indexes, files meta data, journaling files system I only want to see correct implementation of write back cache for all application and OS copy machine. Maintaining integrity of all transactions via disk cache, adapter cache and system cache forward and back (writes and reads) in case of system disk-to-self copying for all application ought to give me right performance. CXPF.sys – make it for me only for ATTO, but how to make it now for other applications and OS copying. First of all I need it for the case of disk to self copying.

If I can help you in some “Beta” testing, you need only to ask. And many other peoples will help you also.

Share this post


Link to post
Share on other sites

Test also copying with XP native xcopy command with different switches. Use it from the command prompt.

SCSI:

C:! to C:=2m13s (~7.7Mb/s) – without switch.

C:! to C:=2m25s (~7.1Mb/s) – with /v switch. V-verify destin. file’s identity to source file.

C:! to C:=2m23s (~7.2Mb/s) – with /c switch. C-ignore coping mistakes.

Seems that verifying of destination file don’t drastically hits throughput when you using XP xcopy.

Switch /v of DOS copy, xcopy commands mean in former times verifying each file as it is written to destination file to make sure that the destination files are identical to the source files.

Cas, do cascopy have some default flags for verifying/ignoring destination file’s integrity or not?

Share this post


Link to post
Share on other sites
Guest Eugene

cas,

Just wanted to doublecheck here whether you've received my email replies.

Thanks,

Eugene

Share this post


Link to post
Share on other sites

I didn't notice that anyone has brought this up in this thread but the SCSI disc driver that is native on Windows Longhorn has an extra button to select below the "enable write cache" and it said to enable this only if you have a backup power supply and it would give even more performance.

So maybe this issue is being looked into my Microsoft and I wonder if someone could extract that driver from Longhorn and use it with XP?

Share this post


Link to post
Share on other sites
...the SCSI disc driver that is native on Windows Longhorn has an extra button to select below the "enable write cache"...

Tex mentioned the same option above.

I wouldn't waste time trying to mix Windows components from different versions. XP Cache Filter does the same thing for now. Once we have all upgraded, or the new option is made available in a service pack, I can throw my kludgy filter away.

Share this post


Link to post
Share on other sites
...so what's the meaning of this post? well... just to show that there's no real performance increase (at least, for my system)...

Yes indeed, not for your system. Note that You had 40MB/s+ writes before the patch in question. Some are struggling at 20MB/s, and they CAN see the difference with this patch. Note that You also have the "notorius" VIA133a chipset that has some PCI bandwidth limitations without correct patches.

Cheers,

Jan

What patchs for the VIA133a chipset would those me? Just the newest 4in1s?

Share this post


Link to post
Share on other sites

I did the Cache Filter. Thank you Cas, it boosted my scores for my x15 36lp quite nicely. It looks like my reads and writes are capped at 50mb/s for that drive and 40mb/s for my X15. so Im happy. Thank you very much

Share this post


Link to post
Share on other sites

:D The latest version of XPcachefilter works great! No more reboot. Writes are now about 2.5x faster than before according to ATTO and Sandra.

In ATTO writes are now hitting 25MB/s vise 10MB/s.

Thanks cas!

( :cry: It does emphasise the point that my reads are still pathetic at 15MB/s - on to other parts of the forum)

Share this post


Link to post
Share on other sites

How do i know cascopy and the XPfilter is installed. Whenever i click on the excutable in WindowsXP a slight black(dos) box opens in the background for a split second and then closes. Any help would be greatly appreciated.

Share this post


Link to post
Share on other sites
This may be stupid, but will this effect my performance in Windows or just Dos? :?:

Windows.

Share this post


Link to post
Share on other sites

Fantastic! While I never ran any benchmarking programs so I don't have any actual numbers, the difference is insaley noticeable.

I have a Cheetak 15k.2 and an IDE RAID-0. In the past I noticed that copying large files from the raid to the scsi took much much longer than copying the same file from scsi to raid. Now it's about the same speed (or maybe even faster?). So I guess this filter fixed slow scsi write speeds.

FYI, I'm running dual p3 tualatin on a via 266t chipset.

Share this post


Link to post
Share on other sites

I don't want to be rude but perhaps you should learn how to use the command prompt before you mess around with unsupported drivers.

Share this post


Link to post
Share on other sites

about all i can do is Format/Fdisk/copy from the dos prompt and i was able to find files in Win98 dos--Has XP changed. Please Help. Would really like to install this to see if it would affect my SCSI performance as it sucks as it is now.

Thanks Ahead of time

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