Announcement

Collapse
No announcement yet.

Seeking 2nd opinions: CPU/Memory issue

Collapse
This topic is closed.
X
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Seeking 2nd opinions: CPU/Memory issue

    My latest build, a Phenom 9750-based system, has been troublesome since September.

    In September I discovered (the hard way) that the memory seemed to be bad.

    Long story short, after replacing both the memory and the motherboard, the sucker is still acting up.

    Replacing the memory at first seemed to make the problem go away, til around mid November (2 months to the day). I decided then that there must be something else at work here, and swapped memory with a friend since I had no other DDR2-capable systems. 24+ hours of memtest86 showed no errors on either system. At that point I replaced my mobo. A few days later, it's crashing, rebooting, and in general misbehaving. Memtest86 actually crashes. (it did before on the old mobo too)

    The cmos doesn't have a voltage monitor, so I don't know what voltage the memory was running at. I manually set it to 2.2, the voltage indicated on the sticker. Still there are problems (but only after ~36 hours). I ran SpeedFan and enabled logging on it. No heat anomalies preceding a crash.

    Previous motherboard was an ECS A770M-A. New one's an Asus (don't recall the model, will check when I get home).
    Memory is Crucial Ballistix DDR2-1066, cas 5-5-5-15.

    My thoughts on it: Could the memory have actually been damaged if it were running under-voltage for long periods? That might explain the disappearance of problems upon replacing it, only to have them reappear later. It only ran on the new mobo for a few days, so that could've been sheer luck with damaged memory.

    Other ideas: PSU is a 550 Watt Antec, I hate to think it might be the problem, but I would suppose it is possible. PSUs are one of if not the most common failure these days.

    CPU might be damaged or defective.

    Maybe both the AMD 770 (on the ECS) and 780 (on the Asus) chipsets suck, too.

    Probably the easiest test would be swapping RAM again, but running the test (or heck, using it) for several days, rather than ~24 hours.

    Thoughts, fellow geeks? I miss my quad core. I'm addicted to Fallout 3.
    Supporting the idiots charged with protecting your personal information.

  • #2
    if you place your memory in your freinds mobo does the issue return? also is there a pc repair place with a power supply testing rig. a bad voltage regulator could allow spikes that would damage your memory or cpu.
    This is a drama-free zone; violators will be slapped. -Irving Patrick Freleigh
    my blog:http://steeledragon.wordpress.com/

    Comment


    • #3
      We swapped memory before I replaced the mobo. I put his ram in mine, my ram in his, and we both ran memtest for ~24 hours. No errors. Either not long enough, or the slower speed (667 vs 800) mitigated (or even prevented) the chance of errors.

      As far as a pc repair place I don't know. I guess I could make some phone calls.

      I checked my spare PSU. It's 430W, has 3 12V rails and according to http://www.extreme.outervision.com/PSUEngine it's more than enough for the machine. I'm gonna try it.
      Supporting the idiots charged with protecting your personal information.

      Comment


      • #4
        Have you looked / googled for that motherboard? There might be issues with that batch of boards.

        I've shopped on newegg and found that some motherboards will not take a specific BRAND of RAM. Have you tried any OTHER RAM yet?

        What version is the BIOS?
        What speeds can the board run RAM?
        What voltages does it want for RAM and CPU?

        Do you have a voltmeter? Test your PSU to check voltage. Antecs are normally very good PSUs that are steady and clean power. Even so, still eliminate the possibility.

        Cutenoob
        In my heart, in my soul, I'm a woman for rock & roll.
        She's as fast as slugs on barbituates.

        Comment


        • #5
          Here's the new Asus mobo: http://www.newegg.com/Product/Produc...82E16813131324

          Bios is an American Megatrands v 2.61. I wish it were an Award bios. This one doesn't have much of a hardware monitor on it.

          Mobo's capable of 1066 for the ram. Don't know what voltage it auto-sets the memory to (lack of monitoring) but I manually set it to what Crucial says the memory wants: 2.2V.


          My dad asked me some questions that really made me think about the symptoms*. (there's the difference between an engineer and a technician, right there) Based on my answers he was leaning toward the PSU, so just to be thorough I swapped power supplies (Old: Antec 550, New: Antec 430, both with 3 12V rails) and encountered the same problems (it was crashed when I got up today, reboot to memtest gave thousands of errors. Powering down for a while then booting memtest again results in no errors, so far), so it's not the psu, unless I have two bad Antec PSUs from two separate product lines.

          Not the PSU, not the memory (though I haven't swapped memory with my friend since I got this particular motherboard, and may not have tested long enough before) , not the mobo.

          I'm thinking my Phenom's bad. I don't want to think that, but it seems the only option left.

          Kinda ironic, I build so many computers out of cheap-ass parts, (ECS mobo, no-name PSU, no-name memory, low-end AMD CPU) and they all work fine. The one I spend some extra money on to get good brands for some of the parts, ends up the most troublesome.

          * He asked about the intervals between symptoms. After I thought about it, I recalled it seemed to follow a pattern of sorts. Two months from initial build to first symptoms, then, crashes and errors fairly often. RMA'd the memory. Computer was off for a week waiting for new memory. Another two months before symptoms reappeared. Swapped memory. Tested mine again on the ECS board. All's fine for a few days. Then I shut it off again to prevent any corruption of disk files, and waited for new mobo. Replaced mobo and ran for only four~ days.
          It seemed that the longer the system was off, the longer it'd be before it crashed again.
          Supporting the idiots charged with protecting your personal information.

          Comment


          • #6
            I'd recheck that memory.

            It's just that I had a really bad experience with Crucial Ballistix memory.

            Short story - I bought a matched pair of DDR 400 for my Opteron system on an ASUS board. Memory fails a couple of years later. Send memory to be RMA'd, buy another matched pair to have the computer operational. Get new memory, doesn't last a day. Took 4 returns to get a working pair, which is not in the Opteron system, but in my HTPC (3 GHz. P4/ASUS mobo).

            I used to use either Kingston or Crucial exclusively. Now, I bought Mushkin for my last system build, and that's been great.

            Comment


            • #7
              Sheesh. That's bad.

              Since RMA'ing the memory is essentially free, I suppose that's what I'll do next, for the second time. My friend only has Wednesdays and Thursdays off so it's not easy to trade with him again. Man, I wish this system had at least one DDR1 slot. Then I could at least cannibalize memory from one of the others temporarily, even if only for testing purposes to absolutely rule out cpu/mobo/psu.

              Too bad my Sempron's not one of the AM2 models. That'd be another way to definitively rule out the cpu.
              Last edited by otakuneko; 12-05-2008, 06:32 PM.
              Supporting the idiots charged with protecting your personal information.

              Comment


              • #8
                Thermal stress fractures in the memory? Didn't catch if you were monitoring in box temps in the posts but working fine on start and then developing problems as use continues often points to heat. Post a follow up if you ever nail down the culprit - always nice to have a new possible solution in the bag of tricks.

                Comment


                • #9
                  Hmm, aside from leaving the case open and occasionally pointing one of those IR thermometers (which I don't have ) at the memory I don't really have a way to monitor the memory itself.

                  I did have speedfan going with logging active on temps last time, temp remained fairly constant throughout.

                  Currently testing with a different brand of memory.
                  Supporting the idiots charged with protecting your personal information.

                  Comment


                  • #10
                    So far, so good with the new (different brand) memory. Then again, last time I replaced the memory, it took two months for any problems to show up.
                    Supporting the idiots charged with protecting your personal information.

                    Comment


                    • #11
                      9 days. was in the middle of exploring vault 92 in Fallout 3 when it froze up, nothing responding.
                      Reluctantly put the memtest disc in and rebooted. Within minutes, it crashed.

                      This is bad. For those keeping score, the only things that have remained constant are the video card and cpu. (well, alright, the case is the same too)

                      The new mobo has a Radeon HD 3400 integrated. It's probably not worth bothering with and integrated video does suck (might play Fallout, but at greatly reduced detail), but given that I don't have a spare socket AM2 CPU, or a PCI-E video card and I'm not going to buy one, it's the only thing left. I might do it, if nothing else just to be thorough. I'd be shocked, absolutely shocked, if it were the video card. But what did Sherlock Holmes say? "Whatever remains, however improbable..."

                      Edit: That settles it. The CPU is the root of the problem. Pulled out the video card, leaving the CPU as the only constant. It crashed memtest. Thrice.

                      Guess I'll be shopping for an AM2/AM2+ CPU. A Phenom 9750 is $150 at Newegg. I may end up downgrading a bit.
                      Last edited by otakuneko; 12-15-2008, 01:46 AM.
                      Supporting the idiots charged with protecting your personal information.

                      Comment


                      • #12
                        Unfortunately, I would guess the CPU rather than the graphics chip. Remember that in all recent AMD systems, the RAM is connected directly to the CPU socket, not via the "northbridge".

                        Can you find someone with another AM2-socket m/board who's willing to test your CPU?

                        Comment


                        • #13
                          You posted as I was editing in my update.

                          Donno anyone with an AMD-based system, but I'm now fairly certain it's the CPU
                          Supporting the idiots charged with protecting your personal information.

                          Comment


                          • #14
                            Do you have any recourse to a warranty?

                            Comment


                            • #15
                              Perhaps with AMD, but not with Fry's.
                              Supporting the idiots charged with protecting your personal information.

                              Comment

                              Working...
                              X