I have the same problem.
My sys is:
Processor - Athlon Barton 2600+
MB - Asus a7n8x - deluxe
RAM - 2X512 MB Geil Value DDR400
Power - HyperPower 450 W
HDD - 2XSeagate Baracuda 7200rpm SATA
Video - Leadtek 7600 GS AGP 8X
With this system the problem seem to apear when i shut down the computer (not at restarts). After a few shut downs (5 - 10) when i boot up the computer scandisk starts a check and finds a lot of index's corrupt and a lot of "orphaned files!?". After a period with this scandisks only a format will get it up and running.
After a lot of reading forum posts i find some tricks to get the problem solved. This triks are : upgrading bios and sata bios, lerge disk fix... and last changing from ACPI uniprocesor PC to Standard PC. After a long period with no problems i got to reformat my c:. And guess what? scandisk came back ( For me, seems that bios and sata bios updates dont work like i expected. The only real temporary fix (until next format) is changing in computer manager from ACPI Uniprocesor PC to Standard PC (thing that get ur rig to turn off only by hand from the power button when u see Its now safe to turn ur computer off ()
If somebody find a permanent fix (else than selling the pc) please post here.
try to disable the drives own caching feature. i run my hdd's in raid on a sil controller and it does not like drives with cache enabled. not all drives have the option to turn it on or off, some use bios settings, some software, some hardware.
Something else to keep in mind...One of the major sensitivities in SATA interface drives is electrical noise.
Even improper mounting of the drive can cause excessive retries....these won't necessarily post errors when the retries succeed, but performance will suffer greatly.
E.G.- The SATA specs limit things like the maximum permissable voltage differential between mounting point screws and frame ground. There are other constraints.
When you exceed the noise spec limits, performance is degraded as internal error rates go up dramatically.
I would suspect that truly excessive noise might also result in both detected and undetected data corruption...similar to industry-wide parallel SCSI interface noise data corruption problems I worked to solve some years ago.
Cable and connector issues would aggravate the situation...especially on the SATA2
interface. Proximities to high RF output noise devices like graphics cards, etc might also aggravate the problem. I've not done any testing in this area, so I can't say for certain, but it would be consistent with other technology sensitivities.
@ bot : How do i disable the cache feature? (And this could get performance loss? because i dont mind at all if i shut down the pc from the button)
@ john albrich : I dont think is an electrical noise problem because i think that this ,if it was caused by electrical noise, should manifest either if the Win XP is instaled as ACPI uniprocesor PC or Standard PC. The problem seems to resolve for me if i change that driver from ACPI to Standard (old AT style when Its noe safe to turn your computer apear). That seems to me that Windows shuts down faster than the data from the hdd cache is written to hdd. Or maybe i am wrong?
Once again thanks for the replyes and excuse my bad english.
If all eles fails SpinRite6 from http://www.grc.com works real good for data recovery.
It runs MUCH faster than ever before, can help maintain all of your drives in tip top shape, can warn of impending disaster, and wrestle data from dying and nearly dead drives .
The only problem is the cost $89.00 US. But for me it was well worth it!!
depending on the hdd and controller
for disk: check the label or manufacturers manuel if there is a jumper setting to enable/ disable caching
for controller: onboard controller will usually have in the system bios, controller cards come usually with there own bios which is loaded right after the system bios. if you have a card then you could enter them by pressing: ctrl+A, ctrl+S, ctrl+F, ctrl+F1, ctrl+F4, etc - these keys will also work for some onboard controllers, really depending on the board.
also you can go in windows into the device manager and find your controller under disk drives. either the controller or the disk should show up. right-click on it and select properties. go to the policy tab, here you can select if you want to disable or enable caching or select advanced features.
if the write cache is enable it will enhance performance but in case of power failure, surge or any interuption it can cause loss, fragmented or corrupted data.
disabeling this feature will allow for quick removal, the data is directly written to disk, so in any case of interruption, ideally the data is already on the disk.
@ john albrich : I dont think is an electrical noise problem because i think that this ,if it was caused by electrical noise, should manifest either if the Win XP is instaled as ACPI uniprocesor PC or Standard PC.
I tend to agree. However, I've seen stranger things result in system-undetected data corruption on disks, including factors such as data block size transfer (whether system is transferring 2048, 4096, etc. bytes of data to the disk at a time) The disk buffer boundaries were influenced by interface signal noise under some conditions and not others.
It's possible the differences in the way that ACPI UniprocessorPC, etc handle data transfers might be similar in terms of whether errors are produced.
That's why I thought it was a good idea to mention this, especially as SATA drives are becoming more popular. As end-users we'll likely start seeing things even long-term lab testing doesn't produce.
edited to correct acronym. Having eye injuries suck in terms of trying to proof-read
In my first post i didnt specify that the Seagates are in raid 0 on the onboard sata and raid controler SIL3112a (now i dont remember the chunk size), but even without raid 0 configured the same problem.
@bot: After next format i will try to disable de write cache. (because i already changed to Standard PC after the first data corruption)
@john albrich: And how i can verify that it is or not a electrical noise problem? Should i try with a multimeter that my computer is well grounded?
Thanks for your answers, excuse my bad english and Best regards!