Announcement

Collapse
No announcement yet.

problems with test 13 "row hammer"

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

  • problems with test 13 "row hammer"

    Hi,

    I've recently bought a new computer. The RAM kit is constituted of a pair of ddr3 sticks (2*4GB). The product name is "Ballistix sport low profile", the following kit:

    http://www.amazon.fr/gp/aw/d/B00A14ZT6U/ref=mp_s_a_1_5?qid=1424090108&sr=8-5&pi=AC_SX110_SY165_QL70&keywords=ballistix+low+pr ofile&dpPl=1&dpID=41V9Igy%2BPzL&ref=plSrch

    I've tested each stick separately, in each slot, in parallel/round robin/singlecore mode and I have errors on both sticks. Curiously I have more errors with CPU 0.

    Is my ddr3 kit defective ? Should I have to send back the
    kit to Crucial for a replacement ?

    Thank you.

  • #2
    How many errors per pass?
    Are all the errors just on Test 13?
    Are the error over a wide range of addresses or tightly grouped?
    Are the errors single bit errors?

    Comment


    • #3
      How many errors per pass?
      - First test, in parallel mode :
      During 4 passes I had around 15-16 errors on CPU 0 with the first stick, then I decided to swap with another slot. I had errors again. Then I changed the stick with the second one.
      - Second test, in parallel mode, with the second stick, in the second slot,
      during 2 passes I had around 10 errors, on CPU 0.
      -Third test, I select a logical core (for a single core test), maybe the third one, I don't remember (for testing the second stick, again). As soon as I had errors, I select another logical core, I had 2 or 3 errors during one or two passes, I don't remember exactly.
      - In the last test I performed, I have chosen the logical core numbered 5, and I let the computer performing tests during the night (maybe 10 hours). I had 5 errors during 44 passes, something like that. That appears curious for me, since I had the impression that I have more errors on CPU 0.

      Are all the errors just on Test 13?
      Yes.

      Are the error over a wide range of addresses or tightly grouped?
      I don't know. I've not looked at that.

      Are the errors single bit errors
      Sometimes I had a single byted changed, for instance FFFFFFFF -> FFFFFBFF.
      Sometimes I had something that appear random, for instance 00000000 - > 203A5C9F.

      Is it necessary to perform more, and accurate, tests ? I can come back on this thread in some days, with more results...

      Thank you for your help.

      Comment


      • #4
        The fact that you are only getting errors in the Row Hammer Test means that your RAM modules likely contain inherent defects making it susceptible to disturbance errors caused by charge leakage, as described in this paper. According to the paper, a significant number of RAM modules are affected by this defect though we haven't encountered nearly as many in our testing.

        Because disturbance errors are only exposed in extreme memory access cases, it is less likely to manifest itself during typical home PC usage (eg. web browsing, word processing, etc.). It may be of greater concern if you were running highly sensitive equipment such as medical equipment or database servers.

        You can send your RAM back for replacement but chances are that the same model of RAM is susceptible to the same defect. It may be better to switch to another model that is known to have passed the Row hammer test (eg. see this thread:
        http://www.passmark.com/forum/showthread.php?5077-How-to-relate-to-errors-in-Hammer-Test-13)

        Comment


        • #5
          it is less likely to manifest itself during typical home PC usage (eg. web browsing, word processing, etc.)
          And what about video games ?

          It may be better to switch to another model that is known to have passed the Row hammer test (eg. see this thread: http://www.passmark.com/forum/showthread.php?5077-How-to-relate-to-errors-in-Hammer-Test-13)
          I've seen this post, but I need low profile ram sticks.

          Thank you for your help.

          Comment


          • #6
            HyperX Savage are low profile sticks

            Comment


            • #7
              I am also having a similar problem, see this thread :
              (http://www.passmark.com/forum/showth...-is-plugged-in)

              and I would also like to know the likelihood of this affecting heavy gaming sessions. I do alot of video editing also and so would consider this much a more senstive use of the equipment than "web browsing, word processing".....
              There are nights when the wolves are silent and only the moon howls. - George Carlin

              Comment


              • #8
                I would also like to know the likelihood of this affecting heavy gaming sessions
                It is impossible to predict with any accuracy. It is by its nature random and it isn't known which games and how many games contain algorithms that provoke the RAM faults.

                However it is safe to say that,
                • Row hammer errors in your RAM won't cause consistently low frame rates
                • Row hammer errors won't be the cause of non random errors. e.g. a game crash on one specific map
                • Row hammer errors won't cause consistently bad behavior. That is to say that if you can reproduce exactly the same error, time and time again, in a game then the fault is more likely to be elsewhere (bad device drivers, over heating, bad video card, power supply issues, or just a plain old garden variety software bug in the game itself).

                Comment


                • #9
                  Originally posted by klmd View Post
                  HyperX Savage are low profile sticks
                  You are correct, but just a little warning to OP. The HyperX Savage are branded as 'low profile'. But the marketing geniuses have been very good at mixing up the meaning of 'standard' vs 'low profile'. What many would call 'normal' height is usually markeded as 'low profile'. I am not 100% sure of the RAM sticks in OP's Amazon link. Tried to Google BLS2C4G3D1609ES2LX0 against Crucial.com. No hits. Closest I get is BLS2K4G3D1609ES2LX0. Which is branded as 'low profile', but looks to be 'very low profile (VLP)'.

                  My point to OP. If you go for new RAM sticks. Make sure you double-check the actual physical height of existing vs new ones. The HyperX Savage PDF spec says: 133.35mm x 32.8mm. Which is actually 2.8mm extra on the top because of the heatspreader, vs the 'normal'/'low profile' 30.0mm.

                  Comment


                  • #10
                    Yes the the Kinsgston Savage sticks are labbeled low profile but have standard dimmensions. The only difference with other sticks non labelled LP is that the heatsink height is short.

                    Today there exist two products really "low profile":
                    - My faulty model "Ballistix sport low profile" from Crucial
                    - and the model from Corsair, named "Ballistix Tactical low profile", which seems to have problems with the Hammer test. Look at this post:
                    http://forum.crucial.com/t5/Crucial-...il/td-p/164100

                    Maybe, the compact form, with modules close one to another, are subject to problems with this test 13 ?

                    Kevin

                    Comment

                    Working...
                    X