Announcement

Collapse
No announcement yet.

F4-3600C18Q-128GTZN won't boot with XMP settings on ASUS ROG Crosshair VIII Hero

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

  • F4-3600C18Q-128GTZN won't boot with XMP settings on ASUS ROG Crosshair VIII Hero

    A computer I have assembled is using the following components:
    ASUS ROG Crosshair VIII Hero
    AMD Ryzen 9 3950X
    G.Skill F4-3600C18Q-128GTZN

    When setting the default XMP settings in BIOS (latest version, 2103), I can't even get past post: it tries a couple of times, then enters BIOS safe mode. Here I can adjust the XMP settings by lowering RAM clock frequency to 3200 MHz (from 3600 MHz), which enables me to boot to OS, however after 20 or so hours of operation the computer suddenly shut down without warning. A single complete pass of MemTest86 v8.4 reveals no errors - since it took four and a half hour, I haven't been able to run more extensive tests.

    It boots to OS just fine using default JEDEC settings (no further stability tests done), so that's what I'm using right now.

    I am very disappointed that the RAM kit is unable to run as advertised under optimal conditions. I am posting here to ask for help in troubleshooting eventual errors on my side.

    I also have another question: The G.Skill website, the retailer's website and the stickers on the individual RAM sticks display the latencies 18-22-22-42 and the voltage 1.35V. A sticker on the kit box instead displays the latencies 18-25-25-45 and the voltage 1.40V. Why is this? Is this memory kit being falsely advertised as having better latency profile due to some mistake?

  • #2
    Try each module individually to see if they all have the same maximum results.

    Memtest is only accurate with one module at a time. If there is any problem, the issue should arise more quickly when a stick is performing on it's own.

    For that memory model there is a label misprint, however the actual product is correct. Check to see if each module can reach DDR4-3600 18-22-22-42 on it's own with DOCP enabled.

    Let us know how it goes.

    Comment


    • #3
      Originally posted by GSKILL TECH View Post
      Try each module individually to see if they all have the same maximum results.

      Memtest is only accurate with one module at a time. If there is any problem, the issue should arise more quickly when a stick is performing on it's own.

      Check to see if each module can reach DDR4-3600 18-22-22-42 on it's own with DOCP enabled.

      Let us know how it goes.
      Thanks for your attention to this matter. I will proceed with testing individual sticks this evening and report results.

      Originally posted by GSKILL TECH View Post
      For that memory model there is a label misprint, however the actual product is correct.
      I assume this means the latencies and voltage listed on the webpage is correct?

      Comment


      • #4
        Yes, XMP Profile, SPD programming, everything should be same as shown on web page.

        Comment


        • #5
          I now have the following to report:

          Using XMP settings, all four sticks boot to OS when inserted individually in the 1-stick slot of the motherboard, no crashes.

          Using XMP settings, all four sticks have no errors when testing them individually, using 2 complete passes of Memtest (I have not been able to run more tests due to time constraints).

          It would seem there is no issue with the individual sticks.

          Comment


          • #6
            A new BIOS dropped for the board quite recently, you might want to try that.

            Version 2204 2020/08/14 19.24 MBytes

            ROG CROSSHAIR VIII HERO (WI-FI) BIOS 2204
            "-Update AMD AM4 AGESA V2 PI 1.0.8.0
            -Improve system performance.
            -Improve system stability
            -Improve M.2 storage compatibility

            CH8 https://www.asus.com/Motherboards/RO...HelpDesk_BIOS/
            CH8 Wifi https://www.asus.com/Motherboards/RO...HelpDesk_BIOS/
            Last edited by RedSector; 08-16-2020, 03:33 AM.

            Comment


            • #7
              Originally posted by RedSector View Post
              A new BIOS dropped for the board quite recently, you might want to try that.

              Version 2204 2020/08/14 19.24 MBytes

              ROG CROSSHAIR VIII HERO (WI-FI) BIOS 2204
              "-Update AMD AM4 AGESA V2 PI 1.0.8.0
              -Improve system performance.
              -Improve system stability
              -Improve M.2 storage compatibility

              https://www.asus.com/au/Motherboards...HelpDesk_BIOS/
              Thanks for the advice and your interest in this topic.

              I updated my BIOS to this newest version and tried again using all four sticks, leading to the same results as previously: JEDEC settings boot to OS and seem stable, XMP settings makes the system repeatedly shut down during the POST sequence, eventually reaching safe boot.

              (To not confuse other people reading this topic, I would advice against posting a link to the BIOS for the Wi-Fi version of the ROG Crosshair VIII Hero when this pertains to the non-Wi-Fi version)

              Comment


              • #8
                My bad I own CH8 wifi. No real point getting your knickers in knot over it. I just added the other. Not really that much of an issue, they have moved together for every BIOS to date from launch and you can't flash one on the other in any case. Most people who own them, would know their sister boards and not identical sisters.

                Did you try 1302 BIOS at any point ?
                Last edited by RedSector; 08-16-2020, 04:51 AM.

                Comment


                • #9
                  Originally posted by RedSector View Post
                  Did you try 1302 BIOS at any point ?
                  Before I updated BIOS to 2103 I tried the sticks using some older version, but sadly I don't remember which one.

                  Comment


                  • #10
                    Originally posted by GSKILL TECH View Post
                    Try each module individually to see if they all have the same maximum results.

                    Memtest is only accurate with one module at a time. If there is any problem, the issue should arise more quickly when a stick is performing on it's own.

                    For that memory model there is a label misprint, however the actual product is correct. Check to see if each module can reach DDR4-3600 18-22-22-42 on it's own with DOCP enabled.

                    Let us know how it goes.
                    Do you have any further advice in response to the results?

                    Comment


                    • #11
                      With DOCP enabled, manually lower DRAM Frequency until the system can run properly. We need to know what works to get an idea of limitations. Since all modules work fine individually, seems like the CPU internal memory controller may be overloaded with full slots. Try a higher DRAM Voltage to see if that can help.

                      Comment


                      • #12
                        Originally posted by GSKILL TECH View Post
                        With DOCP enabled, manually lower DRAM Frequency until the system can run properly. We need to know what works to get an idea of limitations.
                        The testing I did initially would suggest 3200 MHz is at least semi-stable. I don't know if the spontaneous shutdown I mentioned was due to RAM or some kind of transient fault in the power supply network (I have had to deal with those in the near past). I can explore this further when the computer becomes available for me to perform tests again.

                        Originally posted by GSKILL TECH View Post
                        Since all modules work fine individually, seems like the CPU internal memory controller may be overloaded with full slots.
                        Yes, this is my strong suspicion as well. If you look here and here, AMD officially lists no 128GB kits as compatible with Ryzen CPUs.

                        As you write, the likely explanation is that the IMC or IF can't handle the load from large amounts of RAM when pushing the clock frequency of the IF, if you don't happen to have won the binning lottery.

                        My hope when buying this kit was that G.Skill had sorted this out on their end somehow, and that the 128GB kits (marketed as being engineered for the Ryzen 3000 series CPUs) were sold with confidence that they would work with advertised specifications with enough Ryzen CPUs.

                        Originally posted by GSKILL TECH View Post
                        Try a higher DRAM Voltage to see if that can help.
                        I will try this and report back, if that's what you recommend. Please tell me if you need any further data.
                        Last edited by fetch_my_hat; 11-30-2020, 08:08 AM.

                        Comment


                        • #13
                          Originally posted by GSKILL TECH View Post
                          With DOCP enabled, manually lower DRAM Frequency until the system can run properly. We need to know what works to get an idea of limitations. Since all modules work fine individually, seems like the CPU internal memory controller may be overloaded with full slots. Try a higher DRAM Voltage to see if that can help.
                          My apologies for dragging this out with my late reply, but I can now report the following:

                          Using the XMP settings for all other values, the highest clock frequency that allows boot to OS is 3333 MHz (the next increment is 3400 MHz). With 3333 MHz, the system has restarted twice just after OS login, and frozen once unexpectedly without being under heavy load. These errors have been spread out over ~60 hours of operation, so I would say it's semi-stable. I don't have any logs to corroborate these are memory errors, but I am confident they are since the system is perfectly stable with JEDEC settings or with the RAM sticks it had before.

                          I used the Karhu memory testing tool for 8 hours, 2500% coverage according to the tool, no errors. According to this thread however, stability of the IMC is best tested with traditional stress tools like prime95. Maybe I can see if I can catch any errors using that method next week.

                          Originally posted by GSKILL TECH View Post
                          Try a higher DRAM Voltage to see if that can help.
                          Raising the voltage to 1.395 V did not enable boot to OS with 3400 MHz. Maybe not surprising if the IMC/IF is the issue.

                          Since many of the RAM-related settings in BIOS are seemingly not touched by XMP and set to "auto", I have used Ryzen Master in order to see what the values actually are during operation:

                          Click image for larger version

Name:	ryzen_master.png
Views:	973
Size:	95.4 KB
ID:	165230

                          With this in mind, what would you suggest I do next?

                          Comment


                          • #14
                            Adjust CPU SoC Voltage to see what values may improve stability especially at a higher DRAM Frequency. Full slots at a high frequency can be stressful for most IMCs.

                            Comment


                            • #15
                              Hi, was this issue resolved as I was looking into buying same kit but with AMD 5950x processor.

                              Comment

                              Working...
                              X