Corrupted large files (Adaptec 1200A)

[Alex]

I built a computer for the sole purpose of it being a backup machine to
my primary workstation. It has a Biostar P4VTG processor with a 1.8
Celeron, 256 Mb of RAM, on-board IDE (bootable hard drive), on-board
SATA (disabled through BIOS). Running Windows XP, SP2.


I put in an Adaptec 1200A PCI RAID controller card, to which I attached
2 Maxtors (6Y250P0, 250 Gigs), and created a RAID-1.


When I copy large files over to the RAID disk (defined as Basic in
Windows Disk Management, not a Dynamic), they get corrupted. I came
across this by accident -- having opened a large Zip file (~250 Mb),
and extraction complaining. I then started experimenting, and confirmed
that this is, indeed the case -- not just the zips, but all large
files, larger than some number (for Jpegs it seems to be ~3Mb).


Oh, and Windows does NOT complain when copying/creating the files. So
from its perspective storing worked fine.

If, using the same RAID card, I define a RAID-0 (defeats the purpose,
but...), handling of files works fine.

Things I tried (including calling Adaptec for tech support):
1. Upgraded motherboard BIOS to the latest
2. Ensure that no IRQ conflicts/sharing exists on the RAID card
3. Loaded latest card drivers from Adaptec
4. Tried various format options on the RAID disks (diff. cluster
sizes... Prefer to use default 4k)
5. Replaced the card through Adaptec's RMA -- same exact issue with
replacement...


And it is still happening....


It almost seems like a RAID card is sending the data to disks. Disks
may have their own write caching? Windows comes on top of it with its
disk management... Somebody is not doing what they are supposed to.

Any ideas? Some BIOS settings? Some hidden Windows config options?
Additional tests I can run to get more info on where and when the data
is lost?

Thank you in advance!!!

--Alex

 

[Rod Speed]

I built a computer for the sole purpose of it being a backup machine to
my primary workstation. It has a Biostar P4VTG processor with a 1.8
Celeron, 256 Mb of RAM, on-board IDE (bootable hard drive), on-board
SATA (disabled through BIOS). Running Windows XP, SP2.


I put in an Adaptec 1200A PCI RAID controller card, to which I attached
2 Maxtors (6Y250P0, 250 Gigs), and created a RAID-1.


When I copy large files over to the RAID disk (defined as Basic in
Windows Disk Management, not a Dynamic), they get corrupted. I came
across this by accident -- having opened a large Zip file (~250 Mb),
and extraction complaining. I then started experimenting, and confirmed
that this is, indeed the case -- not just the zips, but all large
files, larger than some number (for Jpegs it seems to be ~3Mb).


Oh, and Windows does NOT complain when copying/creating the files. So
from its perspective storing worked fine.

If, using the same RAID card, I define a RAID-0 (defeats the purpose,
but...), handling of files works fine.

Things I tried (including calling Adaptec for tech support):
1. Upgraded motherboard BIOS to the latest
2. Ensure that no IRQ conflicts/sharing exists on the RAID card
3. Loaded latest card drivers from Adaptec
4. Tried various format options on the RAID disks (diff. cluster
sizes... Prefer to use default 4k)
5. Replaced the card through Adaptec's RMA -- same exact issue with
replacement...


And it is still happening....

It almost seems like a RAID card is sending the data
to disks. Disks may have their own write caching?
Windows comes on top of it with its disk management...

Very unlikely indeed that that's the problem.

Somebody is not doing what they are supposed to.

Most likely adaptec.

Any ideas? Some BIOS settings? Some hidden
Windows config options? Additional tests I can run
to get more info on where and when the data is lost?

Demand that adaptec fix the problem or provide a full refund.

 

[Peter]

I built a computer for the sole purpose of it being a backup machine to

my primary workstation. It has a Biostar P4VTG processor with a 1.8
Celeron, 256 Mb of RAM, on-board IDE (bootable hard drive), on-board
SATA (disabled through BIOS). Running Windows XP, SP2.


I put in an Adaptec 1200A PCI RAID controller card, to which I attached
2 Maxtors (6Y250P0, 250 Gigs), and created a RAID-1.


When I copy large files over to the RAID disk (defined as Basic in
Windows Disk Management, not a Dynamic), they get corrupted. I came
across this by accident -- having opened a large Zip file (~250 Mb),
and extraction complaining. I then started experimenting, and confirmed
that this is, indeed the case -- not just the zips, but all large
files, larger than some number (for Jpegs it seems to be ~3Mb).


Oh, and Windows does NOT complain when copying/creating the files. So
from its perspective storing worked fine.

If, using the same RAID card, I define a RAID-0 (defeats the purpose,
but...), handling of files works fine.

Things I tried (including calling Adaptec for tech support):
1. Upgraded motherboard BIOS to the latest
2. Ensure that no IRQ conflicts/sharing exists on the RAID card
3. Loaded latest card drivers from Adaptec
4. Tried various format options on the RAID disks (diff. cluster
sizes... Prefer to use default 4k)
5. Replaced the card through Adaptec's RMA -- same exact issue with
replacement...


And it is still happening....


It almost seems like a RAID card is sending the data to disks. Disks
may have their own write caching? Windows comes on top of it with its
disk management... Somebody is not doing what they are supposed to.

Any ideas? Some BIOS settings? Some hidden Windows config options?
Additional tests I can run to get more info on where and when the data
is lost?

Thank you in advance!!!

Is there a chance that there is something overclocked in your PC?
Did you run extensive memory test?

Try it in a different PC.

 

[Alex]

I checked all the BIOS settings -- it does not appear overclocked. It
is running with the factory defaults....

Trying a different PC would be nice if I had a spare one
If it must work per documentation, it must work -- I am not about to
run around looking for a spare PC just to prove it not working. Also,
since I have been through the RMA process with Adaptec already, this is
the second card with the same problem, chances are --- it is either the
card design or something else....

Do you think Adaptec would give a refund if I bought from compUSA 5
months ago?

 

[Peter]

...chances are --- it is either the

card design or something else....

I think it is something in your PC. That motherboard is really cheap.

Do you think Adaptec would give a refund if I bought from compUSA 5
months ago?

Unlikely, but it doesn't hurt to ask.

 

[Odie Ferrous]

I checked all the BIOS settings -- it does not appear overclocked. It
is running with the factory defaults....

Trying a different PC would be nice if I had a spare one
If it must work per documentation, it must work -- I am not about to
run around looking for a spare PC just to prove it not working. Also,
since I have been through the RMA process with Adaptec already, this is
the second card with the same problem, chances are --- it is either the
card design or something else....

Do you think Adaptec would give a refund if I bought from compUSA 5
months ago?

I had a 2400A Adaptec card for a while. Caused me endless grief.

I managed to get Adaptec to swap it for the 2100s (I decided to go
SCSI).

They advance shipped it from CA to the UK, which impressed me no end.

Unfortunately, the 2100s is also a POS.


Odie

 

[Rod Speed]

I checked all the BIOS settings -- it does not appear
overclocked. It is running with the factory defaults....

Trying a different PC would be nice if I had a spare one
If it must work per documentation, it must work -- I am not about to
run around looking for a spare PC just to prove it not working. Also,
since I have been through the RMA process with Adaptec already, this is
the second card with the same problem, chances are --- it is either the
card design or something else....

Do you think Adaptec would give a refund if I bought from compUSA 5
months ago?

compUSA 5 is legally obliged to give a full refund since its never worked
properly.

The time is irrelevant given the subtle nature of the problem.

Adaptec is likely aware of the design problem and it
shouldnt be hard to force them to provide a full refund too.

 

[Bruce T. Berger]

"

----- Original Message -----

Any ideas? Some BIOS settings? Some hidden Windows config options?
Additional tests I can run to get more info on where and when the data
is lost?
--Alex

I've had data corruption problems with several 1200A's under both WIN 2000
and XP and simply won't use then anymore.

btb

P.S. I realize this won't help your current situation but for what it's
worth in future I've had nothing but good results with the integrated RAID
1-0 SATA controllers in Intel brand 865 chipset boards.

 

[Alex]

Thanks, Bruce )
Well, at least, I know I am not the only one.

Adaptec went back to CompUSA... I think I will pick a cheapo card (like
a SIIG or a Promise) -- just to prove the point

Regardless of the motherboard, it should work -- it is a PCI device
that should comply. So cheap or not cheap, it could be the components
and the workmanship, but in the long run, we are not talking about
moving parts and high quality bearings.

 

[Alex]

Follow-up:
The SIIG controller behaves the same identical way. It does, however,
provide more flexibility than the Adaptec in specifying the mode, in
which the hard drives that make up the mirror should be accessed (PIO
0-4, or UDMA 0-6). I noticed that downgrading UDMA 6->5->4.... -> PIO 4
helps some, in that the number of errors in large files is reduced or
eliminated, but not entirely.

So the issue seems to be not with the RAID controller card (as
expected), but more with how the information travels to the disk. I am
still buffled by what the solution might be

 

[Peter]

Follow-up:

The SIIG controller behaves the same identical way. It does, however,
provide more flexibility than the Adaptec in specifying the mode, in
which the hard drives that make up the mirror should be accessed (PIO
0-4, or UDMA 0-6). I noticed that downgrading UDMA 6->5->4.... -> PIO 4
helps some, in that the number of errors in large files is reduced or
eliminated, but not entirely.

So the issue seems to be not with the RAID controller card (as
expected), but more with how the information travels to the disk. I am
still buffled by what the solution might be

Sad to hear.

Did you try to run "Windows Memory Diagnostic" ?
http://oca.microsoft.com/en/windiag.asp

I did put it on DOS (98SE) boot floppy disk and ran for a day or so.
It helped me to pinpoint an intermittent error I attributed to a hard
disk before. It was a memory issue.

 

[Eric Gisin]

If you are using UDMA modes, corruption over the IDE cable is unlikely.

The lower UDMA modes put less traffic on the PCI bus, so this suggest you
have a chipset or memory problem. Are the advance chipset settings set to the
defaults?

 

[Alex]

Eric,

Everything is set to defaults that the computer came with from the
assembler.
I tried tweaking the BIOS settings as well, but there are so many that
could be affecting the behavior, and not by themselves, but in
combination -- that I eventually gave up.

The problem, obviously, is in something with the timing. It does not
explain, however, why striping works and mirror does not. My
undertstanding was that mirrored sets are transparent to the OS -- it
sees them as one drive only, and the RAID controller hardware is
responsible for making sure writing is done on both... Now, since I
tried two cards (well, three -- counting the RMA from Adaptec) from two
different manufacturers, to me the issue is deeper, while still not
explainable

 

[Alex]

I also ran the diagnostics -- the standard and extended. As I
suspected, it came out clean. The machine is operating just fine, it is
only the mirrored RAID configuration that makes it go "coo-coo".

 

[Eric Gisin]

It is firmware RAID. It uses a standard PCI IDE chip and bios/driver do RAID.
So two writes are issued by software.

I don't know who makes the chip in 1200A, but SIIG uses SiliconImage.

If you do a file compare with "fc /b", any pattern in corruption will help.
For example, bad RAM shows at the same lower 12 bit file offset
(usually), because of page alignment.

 

[Alexander Grigoriev]

My Memtest can also run I/O diagnostics for this kind of problems. I suspect
the same kind of problem resurfaced, which plagued some IDE chipsets a few
ears ago. Bug in the arbitration between primary and secondary channels,
when the total data rate is more that PCI can handle.

Download it from http://home.earthlink.net/~alegr/download/memtest.htm