Corrupt Images - PCI Latency a cause?

It's well known that various factors such as faulty memory and some USB chipsets can cause corrupt images.
However, there are always users who report that they just can't get TrueImage to work successfully.

In my case ever since I've had TrueImage (version 8 build 786 onwards) I've been consistently getting images which are corrupt. This occurs on one specific computer (I have used on another PC and a laptop without a problem).
Very occasionally I get an image which is valid (validates OK on another computer) although the validation fails on this computer. The point during validation at which the validation fails is random.
Consequently, I have had to move my hard disk drive to another machine whenever I want to use TrueImage.
I have done substantial testing, changed various BIOS versions and BIOS configuration data without any clue to what is causing the issue.
Now I believe I may have found a possible cause.

I have done some testing which indicates that PCI Latency may be a factor.
I have been testing using PCI Latency Tool 3.1 v2 downloaded from
http://downloads.guru3d.com/download.php?det=951
I need to do more testing to confirm my suspicions but I'm unlikely to find enough time to do so in the near future.

I haven't done any machine code programming for about 30 years (before the PC era) and am not very familiar with the Intel x32 architecture.
My questions:
What happens if the PCI Latency timer expires during an operation?
Can this fact be programmed for and appropriate action taken?
What does TrueImage do in this case?
Possibly nothing, hence corrupt images and/or validation failure?

Anyone like to do some testing?
If you don't have a corrupt image problem - can you cause it by adjusting PCI Latency?

Note that apparently setting BIOS PCI Latency is not sufficient since not all PCI cards, and particularly AGP graphic cards, actually use this info for latency timing - they just set their own latency.
This is where the PCI Latency Tool software is useful - you can set the latency for each piece of hardware.

Also note that PCI Express cards do not have latency timers.
Does anyone using PCI Express have unexplained corrupt images?

I'd be very interested in comments by Acronis.

 

Hello tachyon42,

Thank you for choosing Acronis Disk Backup Software.

I've forwarded this information to our Quality Assurance Team and will inform you about the result of their investigation as soon as it will be finished. As this can take a few days\weeks we apologize in advance for any delay with the response.

Thank you.
--
Alexey Popov

 

Hello tachyon42,

I've just received a reply from our Quality Assurance Team and they informed me that decreasing the CPU-to-PCI Write Latency timer value increases CPU productivity, but may cause data corruption. This is usually called "overclocking" and not recommended by our specialists. If your computer is overclocked, i.e. CPU-to-PCI Write Latency timer value is different from the default one or from the one officially provided by the respective hardware manufacturer, we can not guarantee that Acronis True Image will function properly on your system. However, there should not be any problems if CPU-to-PCI Write Latency timer value is set up correctly. If problems occur anyway, thy are most probably caused by the faulty hardware devices.

If you have any further questions concerning Acronis software, please feel free to submit a request for technical support or post any of them on this forum. We will certainly try to help you in resolving any issues.

Thank you.
--
Alexey Popov

 

Hello Alexey,
Thank you for investigating my query.

My BIOS PCI Latency has not been changed from the default setting since TrueImage was installed about two years ago.
However during recent testing I did change it and saw results which indicated that during TI execution the elapsed time until a corrupt image message occured was affected by the PCI setting.
I have not overclocked my computer.
I guess it is possible that the motherboard manufacurer's default PCI Latency setting is just not suitable for my motherboard / CPU / etc hardware combination.
So far I have not had enough time to do comprehensive testing to see if I can find a setting which ensures that no corrupt image messages occur.
I will do further testing when it is possible.

Your QA Team has addressed the questions:
"What happens if the PCI Latency timer expires during an operation?
What does TrueImage do in this case?
Possibly nothing, hence corrupt images and/or validation failure?"

However I would appreciate an answer to the crucial question:
"Can this fact be programmed for and appropriate action taken?"
That is, is it possible to modify TrueImage code to detect the fact that an operation (read, write or whatever) has failed due to PCI Latency expiring?
If such an exception occurs then can appropriate action be taken, such as issuing a warning to the log file and repeating the operation until it completes successfully. If the operation can't complete successfully after a fixed limit, of say five times, then and only then should TrueImage terminate with the corrupt image message.
Of course the details of how to achieve this are in your hands but I hope you get the idea.

 

Hello tachyon42,

Please accept our apologies for the delay with the response.

I'm sorry for not being clear enough with my explanation. Actually, I got the idea and forwarded it to our Development Team. I received a reply saying that they consider the fact of the PCI Latency expiration a hardware problem and therefore do not plan to change the current behavior of Acronis True Image. However, they took this idea into consideration and will probably implement the corresponding functionality in the future.

Thank you.
--
Alexey Popov