HDD corruption in GA-970A-D3 v1.1

[uKER]

I'm experiencing a rare issue of data corruption on my newly-purchased GA-970A-D3 v1.1.

It happens that for every 5 GB I copy, ocasionally I get around 1-4 mis-copied bytes.
Strange thing, is, the mismatched bytes are always within one digit of their intended value. Say, if it was supposed to write hex F3 it may write F4 or F2.
Also strange is that it doesn't always happen. Yesterday I managed to copy about 30 GB with no errors.
Occasionally, when performing intensive storage operations, like several simultaneous copy operations or copy+verify, the system may BSOD or crash altogether.
On the other hand, apart from the storage issues the system seems stable, and I'm even to play demanding 3D games with no stability issues.

My setup is:
- GA-970A-D3 v1.1
- AMD FX-4170 (125W)
- 2 x GSkill DDR3 1600 2 GB modules in dual channel (slots 1 and 3)
- Samsung Spinpoint F3 1 TB SATA2 HDD
- Crucial M4 128 GB SATA3 SSD
- GeForce GTX560
- Thermaltake TR2 600W PSU

I don't overclock. Anything.
Having an SSD and a NCQ-enabled HDD, I'm obviously running the SATA controllers in AHCI mode, which I'm coming to think may be the culprit.

Now, here's stuff I've ruled out:
- Hard drives: does the same thing in both my SSD and HDD.  USB drives seem unaffected.
- SATA cables: replaced the cables for both the SSD and HDD.
- SATA connectors: tried switching the cables around the six SATA connectors in the motherboard.
- Memory: ran Memtest86 for a whole night, no issues. Also tried swapping all of my memory with other modules just in case.
- Processor: ran Prime95 for a whole night, no issues, temperature stays below 55ºC.
- Operating system: did clean install of Windows 7 and Windows 8, installing no software whatsoever before doing the tests.
- Software issues: corruption has been verified by WinRAR, uTorrent, CDCheck, and commandline FC /B.
- AHCI drivers: tried both the Microsoft built-in ones and the AMD-provided drivers
- BIOS: tried reverting to version F10, as F11c is marked as "beta BIOS" and coincidentally updated the AHCI BIOS.
- CMOS settings: reset several times.
- PSU: tried a brand new one.
- Video: tried a GT210 to make sure the GTX560 wasn't too much of a power hog.
- Power connections: triple-checked all motherboard, video, processor and drives' power connectors to make sure they were properly seated.
- Case: tried running the PC without it to rule out interference or any short-circuits in the mountings or whatever.
- Memory running at 1600 MHz: run them at 1333, the default after CMOS reset.

I tried to RMA the motherboard and the technician tried it in front of me and he seemed able to copy a 16 GB load with no errors, but when I got home I saw he had reset the CMOS and had been running with AHCI disabled, so that could confirm my suspicion that I'm a victim of an AHCI bug in the chipset.
I have yet to perform conclusive testing with AHCI disabled (will do it later today), but I'm running out of things to try.

If this was the case, what would be my options?
Running without AHCI is obviously not a viable solution.

In any case, any suggestions or further things to try will be greatly welcome.

[Vezina]

Set the AHCI ports ,where you connect the HDD/SSD , to be removable ,a different driver will be used.
The setting is in the BIOS.
Or disable write caching in the Device Manager for Disk Drives and see what happens.
If you can ,disable NCQ and see what happens.I have read somewhere that IDE Mode does not use NCQ SATA feature.NCQ is also useless to a SSD if it features it.  

[uKER]

Thanks for your reply.
I'll try your suggestions.
They may be prove useful as troubleshooting measures, but neither flagging as removable or turning NCQ off are things I'd like to do on a daily basis.
Here's some info that explains NCQ does actually provide benefits in SSD drives.

Well, thanks again.

If anyone else has any other suggestions, keep them coming.
I'm really lost ATM.

[Vezina]

Do you think you will ever notice those IOPS in home PC applications ,like when you are playing a PC game !?

If the controller on the mobo has some issues with the NCQ then maybe you should try disabling it and see what happens.Otherwise keep the NCQ enabled and watch your data getting  corrupted

ATI SB 600 south-bridge had issues with NCQ so i wouldn t be surprised that the newer AMD SATA controllers to feature such bug.
I don t even know if my WD HDD has NCQ and i don t care

As SSD-s do not encounter fragmentation i really see no problem having it off in spite of the paper advantages.

You can very well RMA the mobo ,it may be that the motherboard has issues at hardware level if  NCQ works fine for other people with similar configurations.
Nevertheless try disabling the caching features at driver level and see what happens.
Also report to Gigabyte support the issue.

[uKER]

Even if the improvement was 0.01%, I don't think I should have to keep a motherboard that requires workarounds to work properly.

[Vezina]

Even if the improvement was 0.01%, I don't think I should have to keep a motherboard that requires workarounds to work properly.

You are perfectly right but first you have to identify if the problem is with your motherboard or you get the same with a replacement or a similar one.
So you should try the workarounds first.

You can be one of the "lucky" guys to have a motherboard with some manufacturing issue.
I have the boards you see in the signature and they seem to work properly.

I have had corruption issues in a case when some Corsair memory gone bad ,so maybe you should change more ram brands ,maybe it s an un-matched memory kit.Not all bad memory kits show off in Memtest.

Additionally to what i have already suggested (3 things easy to do) ,try and use RAM at slower speeds see what happens.Put it in different slots
 ,use it as Ganged ,use standard JEDEC timings per frequency ,etc.

Anyway test and come with feedback ,it would be interesting to see what you find out.

Is it new or refurbished ?!

[uKER]

You are perfectly right but first you have to identify if the problem is with your motherboard or you get the same with a replacement or a similar one.
So you should try the workarounds first.

I sure will.

I have had corruption issues in a case when some Corsair memory gone bad ,so maybe you should change more ram brands ,maybe it s an un-matched memory kit.Not all bad memory kits show off in Memtest.

I'm having a hard time getting my head around that.
Memory able to withstand being thoroughly hammered all night by Memtest86 but failing to keep consistency during a file copy?
Stable enough to run a 3D game for hours but unstable during file copy?
Unstable only when writing to SATA but stable writing to USB?

Additionally to what i have already suggested (3 things easy to do) ,try and use RAM at slower speeds see what happens.Put it in different slots
 ,use it as Ganged ,use standard JEDEC timings per frequency ,etc.

As my report says, I already tried the memory's default timings, set when I reset the CMOS settings.
These are 1333 MHz at the SPD-provided timings for that speed.
I'll try ganged mode, but for all the reasons I mentioned before, I REALLY don't think this is memory-related.

Anyway test and come with feedback ,it would be interesting to see what you find out.

Sure will.

Is it new or refurbished ?!

It is supposed to be new.

[Vezina]

Take a look here as well ,even if it s not even an AMD based motherboard ,it s a similar issue , RAM related  -> http://forum.giga-byte.co.uk/index.php/topic,9014.0.html

[uKER]

Whoa!

"it read/wrote badly from/to the ssd or any hard drive introducing random 1 bit errors to each 1GB read."
"the RAM is super tested and has absolutely no problem. No overclock."

Those REALLY hit home!
It's EXACTLY the same issue I'm having.

I didn't think that was even possible.
I'll try some more RAM modules when I get home.

Thanks a lot for the heads up!

[uKER]

Well, update...

I tried a third set of RAM (a single 4 GB stick to be precise) and it didn't make any difference.

Disabling AHCI (which effectively disables NCQ) however, made the errors less frequent, but they are still there, and I even got a ntfs.sys BSOD at a time when AHCI was disabled (performing two simultaneous copy/check processes).

Now it seems like I'm out of options.

Last step will be to try my whole system with a motherboard I'll have to borrow from a friend, and if that works, I will demand a replacement.

[uKER]

More stuff I just tried:
- Setting all drives as removable
- Disabling Windows' write caching
- Disabling SATA 3.0 support (fallback to SATA 2.0 I guess)

None of that helped either.

[Vezina]

You ve tried everything.
You should RMA the motherboard.

If the warranty is done by the reseller , then ,if in your country is like in mine :0  ,they may refuse you because most of the time the guys that work in this shops consider the board ok if it boots 
If you got it online the law may allow you a time gap  to send it back with no questions asked .

If you manage to send it back get a UD3 instead.

Something seems to be wrong with the motherboard in my opinion ,it may be related to the memory slots , or CPU socket or SATA sockets or ,or ...

[uKER]

Well, I just tried the motherboard on a friend's system, and guess what.
It had no problems. AHCI enabled and all.
I'm as puzzled as I've ever been.

What I did notice is that there is SERIOUS heat buildup in the Northbridge's heatsink.
Like so-hot-you-cannot-touch-it serious.

Now, is there any reason that could lead my FX 4170 (4.2 GHz, 125W) to cause the NB to overheat more than a 4100 (3.2 GHz, 95W)?
My theory is that it might be just due to the processor being faster and therefore putting more demand on it (I'm running several simultaneous copy/compare/delete loops).

I'm now trying to find some software that lets me query the NB's temperature. Will report when I do that.

[Vezina]

The same chipset cooler was installed when you were running yours.The real NB is in the processor die itself ,nowdays the memory controller is no more in the NB.
It may be that the 4170 CPU to have issues ,either the mobo does not supply some voltages correctly with it ,either the CPU has issues on memory controller side (you ve see the other guy issues related to RAM)
You could try lowering the NB speed in the BIOS ,the 4100 has 2000Mhz while the 4170 has 2200 Mhz if i am not mistaking.
Could be a CPU related issue of some sort.
The NB chips that heats up has the main job to supply PCI-E lines.You could try and replace the crap between the heat-sink and NB chip but don t expect to much from that.
How hot is the SB chip ?!

[uKER]

Well, I saw this in the morning, and while I had already found the culprit, you were spot on too.

The NB would overheat horribly when working under load at 2.2 GHz.
At 2.0 GHz it would build some heat, but at least it kept working.
This explains why the MB had worked with the 4100 and 6100 (both use the NB at 2.0).

Went to the retailer earlier today, and presented with this they took the return and I got a UD3 which seems to work just fine.

Thank you for all your help.