Announcement

Collapse
No announcement yet.

2D Tests Fail -- video memory corruption

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • 2D Tests Fail -- video memory corruption

    I am also getting the 2D Video memory corruption test fail. system setup (I have 3 identical systems with the same errors) :

    Tyan 2895
    dual - dual core opteron 265
    8 gigs ECC memory
    3Ware raid controller with 4 SATA - 250 gig hard drives -- Raid 5
    Onboard serial ATA RAID 1 -- w/ 2 - 400 gig SATA drives
    Jaton PCI express PX6600GT 128 meg video card
    Plextor DVD
    XP 32 bit w/sp2 (we lacked some 64 bit drivers for testing)
    BIT 4.0
    latest nvida drivers --

    it normally throws the 2D video corruption problem immediately in BIT 4.0.
    I had an old BIT 3.1 copy around and I am burning in one system with that
    and so far no 2D video corruption errors. 2 hours and counting.

    the systems have not crashed and seem pretty stable in spite of the 2D video memory corruption errors. but I don't think that I have 3 bad video cards with corrupted video memory. They have seperate loads -- meaning we didn't load from our RIS servers or clone the drives. these were loaded by 3 different people (supposedly using the same process of installation) and we get the same problems. All drivers are up to date. Does burn in test have problems with PCI express video cards.

    I have a customer that needs his systems ASAP so I need to solve this problem immediately.

  • #2
    This is a known issue with some systems.
    See this post for details
    http://www.passmark.com/forum/showthread.php?t=244

    In summary you can ignore a few of these errors. If you get a lot of them, then you should be more concerned.

    -------
    David

    Comment


    • #3
      how many of these errors are a "lot"?

      It was my gut feeling that my P4 3.2Ghz / ATI Radeon 9800AIW system was having video memory issues.

      In any case, is getting about 50-75 of these video corruption errors in a 15-minute test enough to indicate a real problem and not this "pseudo"-problem?

      Thank you.

      EDIT: Feel free to let me know if I need to provide more information with this post.

      Comment


      • #4
        Are you seeing any other effects. In particular texture corruption or display corruption?

        To my mind all these errors are real errors. We write a byte of data to video memory (for example the value 10). When we read back the same byte, some time later, the value is no longer 10 (it might now be 13).

        So either the RAM has failed. Or the value in RAM is still really correct, but the read operation by the CPU has failed.

        So if this was in main RAM it would be an extremely serious problem. In video RAM is isn't so serious because,

        1) It is rare that data is read from video RAM into the CPU. Traffic is normally one way (from the CPU into the video RAM). So if the fault is really caused by a read by the CPU, it won't happen much in real life.

        2) Even if the video RAM is bad, you might not notice it. If your display had a single pixel with subtly the wrong colour for 1/50th of a second, would you notice it? Probably not. But a byte of bad RAM that is being used by a 3D shader on the video card might have a much more obvious (and random) visual effect.

        So no, I wouldn't worry too much about your 75 errors.

        -----
        David

        Comment


        • #5
          The highest amount of those 2D Memory errors I saw was about 250 in a 15-minute test run. (Haven't registered yet so this is as long as I can run it for)

          Ironically I wasn't noticing any actual visual corruption (tearing, etc) so that should have tipped me off that "bad RAM" might not be the problem.


          In any case, I actually did solve the issue. Turns out all I needed was to turn up the DDR/RAM and AGP voltages up by .10 each. After that, I have run every type of RAM and 2D memory test all with no errors. Funny that I didn't think the solution would be something so simple

          Thanks for your time.


          EDIT: If you are curious, the situation was that my GPU likes to waste power, (apparently even though it uses a floppy connector for extra power that is still not enough) and I was running my RAM at 2-2-2-5 timings, and for some ridiculous reason hadn't considered rasing the Voltages on either.

          Comment


          • #6
            I get no errors for the 3-10 minutes of this test, then get several thousand the instant my display shuts off and computer freezes. The problem has been occurring in several games. I'm thinking corrupt video memory or voltage issues...

            The thing is, every error looks like this:
            SERIOUS : 2008-11-25 12:44:55, 2D Graphics, Video memory corruption
            LOG NOTE: 2008-11-25 12:44:55, 2D Graphics, 32 Bit ==> Offset 0xffffffff: Expected 0x55555555: Read 0xffffffff
            All of them the same addresses...

            EDIT: Sorry for being a necromancer, but I didn't want to create a new thread when this one is here to provide background. My video card is a Galaxy Geforce 9800GT 512MB, motherboard is an Abit Fatal1ty FP-IN9, PSU is an Antec EarthPower 500W.
            Last edited by Nebby_99; 11-25-2008, 02:23 AM. Reason: Addendum

            Comment


            • #7
              David's earlier comments, from 3 and 4 years ago, in this post are valid.

              The address show is not correct. We have fixed this in a new revision:
              http://www.passmark.com/ftp/bitpro5.3.1036.0004.exe

              We will include this in the next public build 5.3.1037, planned in a few weeks.

              Regards,
              Ian

              Comment


              • #8
                Thanks Ian. Is that build available for 64-bit systems?

                Thrashing the RAM doesn't seem to duplicate the crash, and the crashes have so far only occurred in games (or in BIT) starting from when I got this video card.

                I have no idea how to check whether or not it's a voltage problem. I'm not inclined to go into the BIOS and crank things up to see what happens. SpeedFan is showing my +12V rails at about 6V, is that abnormal?

                EDIT: Another run through (same BIT version) with both 2D and 3D tests going at 100% rather quickly produced another crash. This time each entry was like:
                LOG NOTE: 2008-11-25 14:26:35, 2D Graphics, 32 Bit ==> Offset 0xffffffff: Expected 0x85ae: Read 0xffffffff
                SERIOUS : 2008-11-25 14:26:35, 2D Graphics, Video memory corruption
                with the "Expected" address incrementing with each error.

                I wasn't able to reproduce the crash running just the 2D test. It can however be reliably reproduced using Local video memory only.
                Last edited by Nebby_99; 11-25-2008, 03:35 AM.

                Comment


                • #9
                  No, the next 64-bit version will be some time off. Does the problem occur with the 32-bit version running on 64-bit Windows?

                  The 12V should be about 12V. Maybe speedfan is wrong, as I can't imagine that your system would work if the 12V really is 6V. Check you have the latest version of SpeedFan.

                  You can buy quite inexpensive power supply Voltage testers (that just test the Voltages of the Power supply, with the Motherboard etc disconnected). However, as you indicate the problem occured from when you added your new Graphics card, I would first suspect that your Graphics card is drawing too much current for your power supply under load. It would be interesting to reduce the test loads in BurnInTest (the duty cycles) and see if the problem still occurs.

                  If you have another (bigger) Power supply, you could try that; or you could try your Graphics Card in another system.

                  You might also check the GPU heat.

                  Regards,
                  Ian

                  Comment


                  • #10
                    Thanks for your help, I'll try that shortly.

                    The system has crashed with the GPU temp at 47 degrees, and has also gone as high as 59 degrees without any issues. Speedfan version is 4.37.

                    Is there another way to check the +12V through software?

                    I've tried it in another PC, but not replicating the exact conditions that caused the crash. I'll try it again when I next get the chance and use BIT.

                    EDIT: Moved updates to separate post rather than just adding to this one again
                    Last edited by Nebby_99; 11-25-2008, 08:04 AM.

                    Comment


                    • #11
                      The first 3 errors of many (after installing the version you linked over the top of the x64 version I was using)
                      SERIOUS : 2008-11-25 18:39:41, 2D Graphics, Video memory corruption
                      LOG NOTE: 2008-11-25 18:39:41, 2D Graphics, 32 Bit ==> Offset 0x2c97c918: Expected 0x55555555: Read 0xffffffff
                      SERIOUS : 2008-11-25 18:39:41, 2D Graphics, Video memory corruption
                      LOG NOTE: 2008-11-25 18:39:41, 2D Graphics, 32 Bit ==> Offset 0x2c97c91c: Expected 0x55555555: Read 0xffffffff
                      SERIOUS : 2008-11-25 18:39:41, 2D Graphics, Video memory corruption
                      LOG NOTE: 2008-11-25 18:39:41, 2D Graphics, 32 Bit ==> Offset 0x2c97c920: Expected 0x55555555: Read 0xffffffff

                      I also noted that the "PC Health Status" page in my BIOS says the 12V rails are 12.20V. Not sure if that's a proper reading though.

                      EDIT: Another run gives these as the first 3 errors
                      SERIOUS : 2008-11-25 21:54:10, 2D Graphics, Video memory corruption
                      LOG NOTE: 2008-11-25 21:54:10, 2D Graphics, 32 Bit ==> Offset 0x2d9c551c: Expected 0x0: Read 0xffffffff
                      SERIOUS : 2008-11-25 21:54:10, 2D Graphics, Video memory corruption
                      LOG NOTE: 2008-11-25 21:54:10, 2D Graphics, 32 Bit ==> Offset 0x2d9c5520: Expected 0x0: Read 0xffffffff
                      SERIOUS : 2008-11-25 21:54:10, 2D Graphics, Video memory corruption
                      LOG NOTE: 2008-11-25 21:54:10, 2D Graphics, 32 Bit ==> Offset 0x2d9c5524: Expected 0x0: Read 0xffffffff

                      The offsets are different. Does that make this more likely to be a power issue than a VRAM issue?
                      Last edited by Nebby_99; 11-25-2008, 11:00 AM.

                      Comment


                      • #12
                        There is other software, but I have to admit that I use SpeedFan for Voltages reporting.
                        Ian

                        Comment


                        • #13
                          I can no longer reproduce this crash in BIT after disabling my top and rear case fans. I'm guessing this meant that more air was pushed through the PSU (at the bottom), helping to stabilise the voltage. I still get the exact same crash at around the same place in Fallout 3, but for now I'm assuming it's an unrelated coincidence (as I can play for an hour or more as long as I stay away from that location).

                          Thanks for your help.

                          Comment


                          • #14
                            I had meet the issue too : 2D Tests Fail -- video memory corruption !!!

                            BIT Version : 5.3 build 1036-0004
                            OS Version : Windows Vista SP1 & XP SP3
                            Run Item :
                            (1)CPU (2)CD-RW/DVD (3)CPU SIMD (4) RAM (5)Video (6)2D (7)3D
                            (Disks (9)Sounds (10)Network
                            Loading 100%
                            4 PCS overthan 15 hrs
                            Rate : 50% (2/4 can see the error message)

                            Dose the BIT5.3 build 1036-0004 fix the issue or not ? Thanks!
                            Or where should we check under our hardware ?

                            Comment


                            • #15
                              BIT5.3 build 1036-0004 does not fix the issue, as it is not a BurnInTest problem.

                              The previous posts in this thread indicate that issues such as slightly incorrect RAM and/or AGP voltages can lead to this problem.

                              Regards,
                              Ian (PassMark)

                              Comment

                              Working...
                              X