No announcement yet.

Need help with F4-3200C14Q-64GVK

  • Filter
  • Time
  • Show
Clear All
new posts

  • Need help with F4-3200C14Q-64GVK

    As in title, I need some help determining the problem with my system.

    Setting DOCP = 3200 - 14-14-14-34 - 1.35v produces error on memtest86.
    Consistently starting on the 2nd run of test no 7 on random address.
    Sporadic errors on other tests (8 + 12)

    I have the 1950x on Asus Zenith Extreme cooled with a Bitspower monoblock + Aquacomputer Gigant (so there's no lack of cooling capacity).
    Water temp never raise above 33 C, on all burn test (prime95, memtest, aida64), but the Tdie processor temp spiked rather high to around 78 in the first 30 sec, going up to 84 C.

    Prime95 + AIDA64 tests have largely returned no error, but I did not run the tests for more than a few minutes in fear of damage that would be caused by the high temp.
    I did touch the backplate of the CPU to check the temp while it's running burn test (while it's indicating 84 C Tdie). It is barely lukewarm.

    Overclockers forum recommended I try adjusting the CPU On-Die Termination.
    On auto, I can check the default value in Ryzen Master to be 68.6 Ohms.
    Adjusting the value to anything lower resulted in boot loop, where I had to clear CMOS.
    I don't dare to set it to a higher value since Ryzen overclocking mentioned that the higher value needs liquid nitrogen cooling.

    Testing on manual timing setting = 3200 - 16-16-16-36 - 1.35v result in no errors whatsoever for hours.

    Is there anything that I mentioned that indicates a fault somewhere?

  • #2
    If the first memtest run never shows any issues and it is always later on you get errors it might be a temperature related issue. However it might also be the case where there is just a bit more Vdimm needed than specified to lower the timings to 15-15-15-35 and then again for 14-14-14-34 on TR4 with a kit like this (which is probably not on the G.SKILL QVL).

    I have no experience with your water cooling setup, but i would make sure with someone else that the temperatures you did observe are within the expected range. Otherwise you might have to double check cooler seating, WLP application and so on. From my experience a somewhat decent cutom water loop should have no problems keeping a 1950X at stock frequency in the low 60s during Prime95.
    >> The official G.SKILL Memory Frequently Asked Questions (F.A.Q.)! << | [pictures] Show off your G.SKILL products!


    • #3
      That's also my thoughts.

      I am currently also in talks with Bitspower as well to crosscheck whether these temps are within standard parameters.

      As it is, the load temps is rather high compared to the reported temps of other blocks in general by a good 10+ C. The idle temps are about average at around 37 or so, gradually raising up to stable around 40 C.

      I'm afraid there's no other 1950x systems with Bitspower monoblock to compare to. I've been asking on a few other forums.


      • #4
        I am still waiting for Bitspower to give me a proper explanation on their standards.

        One thing to note though. If it were just the temps, it would not explain why memtest86 would be fine running for hours on the lower timing of 16-16-16-36 - 1.35v.

        Isn't this forum monitored by G.skill?
        Could I get some input as to what I should check here?


        • #5
          What memtest86 version are you using?

          By reducing memory timings, you are lowering stress and temp. The memory controller is on the CPU, so it directly affects the processor.

          Have you tried boosting the memory controller Voltage to see if it is lacking due to the large capacity? This can help give us a better idea of whether it is a stress, temp, or other issue.


          • #6
            Memtest86 version 7.4

            Tested on 14-14-14-36 - 1.36v - errors starts on 2rd pass > no difference with the tests at 1.35v
            (not sure how high you'd want me to go)

            Tested on 14-14-14-36 - 1.35v - ProcODT adjust to 80 - errors starts on 3rd pass

            Tested on 15-15-15-35 - 1.35v - 5 passes on test no.7 - NO error > Current setting


            • #7
              Can you try 14-15-15-35 - 1.35v ?

              I have similar issues with my C15 kit and going from 15-15-15 to 15-16-16 gets me stable.


              • #8
                Originally posted by Whispered View Post
                (not sure how high you'd want me to go)
                Try 1.37V then 1.38V then 1.39V and 1.40V until you gain stability. Otherwise try Meehos suggestion and add a bit of airflow for the modules to eleminate a temperature issue.

                ProcODT at 80 might be a bit too high, the ideal value usually is in the range of 50 to 65.
                >> The official G.SKILL Memory Frequently Asked Questions (F.A.Q.)! << | [pictures] Show off your G.SKILL products!


                • #9
                  Tested 5 passes of memtest86 with a big desktop fan beside the case just to be sure it's not a thermal issue.
                  Turned out it's 0 errors on 14-14-14-34 - 1.35v.

                  So I added a couple Noctua 120 fans to blow air from the top of the case and had it retested with 1 error on 5 passes.

                  Increasing voltage to 1.36v finally settled the whole problem.
                  Retested through 10 memtest86 passes with 0 errors.

                  I didn't realize that temp would be the issue.
                  I kept hearing that temp is usually not an issue for RAM that already have heatsink plates.

                  This is quite troubling, because I was not planning to add fans on a water cooled system.
                  The case I am using was specifically chosen with sound dampening materials.

                  I don't think the local distro will honor a warranty claim if I replace the heatsink of these Ripjaws V with something bigger.

                  Overall I'm satisfied with the RAM. I'm just a bit disappointed that I have to add more fans just to run it on spec.


                  • #10
                    Revision for final result:

                    Even with the addition of fans, it would seems that it is not 100% stable at 1.36v over 2 hours of testing.
                    Final voltage is set at 1.38v to have it stable on the 2 hours memtest86.

                    ProcODT is left on Auto *Ryzen master" pegged it at 68.6 Ohms


                    • #11
                      1.38V and 68.6Ohms ProcODT is fine and not a bad result for 64GB @ DDR4-3200 on AMD.
                      >> The official G.SKILL Memory Frequently Asked Questions (F.A.Q.)! << | [pictures] Show off your G.SKILL products!