Announcement

Collapse
No announcement yet.

Max temperature for XMP?

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

  • Max temperature for XMP?

    Hello everyone!
    First, I really appreciate G.Skill brand because they provide the most extreme performance memories!
    But, summer is coming in my country, and temperatures are rising (30c).
    And I am facing memory errors on XMP setting with my new DDR4 kit: 4000MHz CL15-16-16-36 1.50V 2x8GB (Ripjaws V).
    These errors always appear when modules reach 58c.
    I tested at CL16 (16-16-16-36) and no more errors, even as high as 63c (can't go higher).

    So my question is simple: what is the max safe temperature to run XMP without any errors?
    Asking it another way: what maximum temperature G.Skill use during stress tests to guarantee stability on their rated XMP settings?








  • #2
    From my experience as an overclocker: The higher the frequency/voltage and the tighter the timings, the lower you may have to keep the temperature. The specific temperature threshold for certain settings can vary from one kit to another. In general you want to keep B-Die based kits below 50C for very tight overclocks and the XMP of your kit is already pretty extreme

    Temperature based instabilities are easy to confirm though, just give the modules some additional airflow and the stability should return with the temperature reduction.
    Last edited by emissary42; 06-04-2020, 02:07 PM.
    >> The official G.SKILL Memory Frequently Asked Questions (F.A.Q.)! << | [pictures] Show off your G.SKILL products!

    Comment


    • #3
      Thank you both for your answers.
      My point was: temperature is a key point for RAM stability, but it's not mention anywhere on the spec.
      Something like this for instance XMP: 4000Mhz at 15-16-16-36 tested max 60c.
      I mean you can kill your OS and risk to loose all your data if you reach the temperature limit of your memory, and it depends from one kit to another, and the buyer must know it.
      I have tested mine to be stable until 58 at stock XMP settings before risking to loose my system.
      Am I the only one to be concerned we are so far from JEDEC standard 85c?
      I am not saying I will not buy a fast memory kit even not stable below 85, I am just saying we need to know max safe temperature for each memory kit when buying, to adapt sufficient cooling and avoid loosing our data...





      Comment


      • #4
        1. G.SKILL DDR4 memory modules are not designed with temperature sensors. This applies to all G.SKILL DDR4 memory modules.

        Therefore, the temperature data shown through the software may not accurately reflect actual IC temperature.

        2. The memory IC chips are rated for operation up to 85C, when measured directly at the IC.

        What this means is, even if there is a temperature sensor on the module, it's not measuring the IC chip temperature. It would be measuring temperature from a different location. So the DRAM temperature reported in the software is not reliable enough to conclude that the memory temperature is the source of error.

        3. With any PC hardware, cooler temperatures mean better stability. We understand this, which is why each memory kit is hand tested under XMP specifications on actual systems to reflect real world usage. No additional memory cooling is used during validation.

        During these validation tests, memory IC chips need to maintain stability at the rated settings regardless of the temperature. So even if it fails at a temperature lower than 85C, then it fails and won't be made into a retail kit at that specification. Only the memory kits that pass will become retail memory kits.

        This also means that it is also important to keep other components cool as well, such as the CPU temperature, which may also be a point of failure. Based on our experience, we would be more concerned about maintaining CPU temperature in order to stabilize the memory kit operation.

        Based on the system information we received from you through eurotech@gskill.com, the i9-9900KS is a processor that may also run very hot. In the screenshot you have provided, it seems to indicate that the CPU temperature is running at about 75C, with a maximum peak of 85-87C, which is pushing toward the high end. Such high CPU temperatures would likely lead to issues with memory stability, more so than memory temperature. We have noticed that the fans seem to be operating only at 800rpm to a max peak of 1400rpm, which is on the low end in terms of fan speed, especially when CPU is on load.


        So there are a few things we'd like for you to try.

        1. Increase fan speed for your CPU cooler to lower the CPU temperature. Based on the screenshot, you might be using a NZXT cooler, and on some NZXT coolers, their CAM software is required to set the fan speed and pump speed to higher speeds, otherwise it will default at a low base speed. If the fans are connected to the motherboard fan headers, then you may need to change the fan speed through the BIOS. If the fans are on a 3-to-1 or 2-to-1 cable splitter, then the fans may also be sharing power, which is not optimal. It's recommended to connect fans to its own power connector.

        2. We do not know if you're overclocking the CPU or if the CPU settings are at default settings. Default CPU settings are highly recommended. Overclocked CPU would increase CPU temperature, and high CPU temperature would affect memory stability.

        3. Ensure you have proper case air flow. Please make sure to have at least 1-2 fans for pushing air outwards. If you are using a liquid AIO cooler, please also make sure that the radiator is set up so that the fans are blowing outwards. If you have additional case fans at the front or on the side panel, those can be set up to blow air into the case. This is to prevent warm air being trapped inside the case, and to make sure that air is properly cycled throughout the system.

        Please check and do the above steps and let us know the results.

        Thank you.

        Comment


        • #5
          Originally posted by G.SKILL View Post
          1. G.SKILL DDR4 memory modules are not designed with temperature sensors. This applies to all G.SKILL DDR4 memory modules.

          Therefore, the temperature data shown through the software may not accurately reflect actual IC temperature.

          2. The memory IC chips are rated for operation up to 85C, when measured directly at the IC.

          What this means is, even if there is a temperature sensor on the module, it's not measuring the IC chip temperature. It would be measuring temperature from a different location. So the DRAM temperature reported in the software is not reliable enough to conclude that the memory temperature is the source of error.

          3. With any PC hardware, cooler temperatures mean better stability. We understand this, which is why each memory kit is hand tested under XMP specifications on actual systems to reflect real world usage. No additional memory cooling is used during validation.

          During these validation tests, memory IC chips need to maintain stability at the rated settings regardless of the temperature. So even if it fails at a temperature lower than 85C, then it fails and won't be made into a retail kit at that specification. Only the memory kits that pass will become retail memory kits.

          This also means that it is also important to keep other components cool as well, such as the CPU temperature, which may also be a point of failure. Based on our experience, we would be more concerned about maintaining CPU temperature in order to stabilize the memory kit operation.

          Based on the system information we received from you through eurotech@gskill.com, the i9-9900KS is a processor that may also run very hot. In the screenshot you have provided, it seems to indicate that the CPU temperature is running at about 75C, with a maximum peak of 85-87C, which is pushing toward the high end. Such high CPU temperatures would likely lead to issues with memory stability, more so than memory temperature. We have noticed that the fans seem to be operating only at 800rpm to a max peak of 1400rpm, which is on the low end in terms of fan speed, especially when CPU is on load.


          So there are a few things we'd like for you to try.

          1. Increase fan speed for your CPU cooler to lower the CPU temperature. Based on the screenshot, you might be using a NZXT cooler, and on some NZXT coolers, their CAM software is required to set the fan speed and pump speed to higher speeds, otherwise it will default at a low base speed. If the fans are connected to the motherboard fan headers, then you may need to change the fan speed through the BIOS. If the fans are on a 3-to-1 or 2-to-1 cable splitter, then the fans may also be sharing power, which is not optimal. It's recommended to connect fans to its own power connector.

          2. We do not know if you're overclocking the CPU or if the CPU settings are at default settings. Default CPU settings are highly recommended. Overclocked CPU would increase CPU temperature, and high CPU temperature would affect memory stability.

          3. Ensure you have proper case air flow. Please make sure to have at least 1-2 fans for pushing air outwards. If you are using a liquid AIO cooler, please also make sure that the radiator is set up so that the fans are blowing outwards. If you have additional case fans at the front or on the side panel, those can be set up to blow air into the case. This is to prevent warm air being trapped inside the case, and to make sure that air is properly cycled throughout the system.

          Please check and do the above steps and let us know the results.

          Thank you.
          Hi sir,
          thank you very much for your time to write a proper reply, very appreciate!
          Indeed, very fast memory can be tricky, especially depending on the trio: [RAM quality]+[CPU IMC]+[Motherboard capacity]
          And I like to test the limits of my new hardware.
          So I slow down my fans on purpose, to let memory temperature increase and find its stability limit.
          And I noticed it was always around 58c, based on this sensors called "DIMM[]" and regardless of my CPU temp.
          You said G.Skill modules have no temp sensors inside, then I assume they are provided by my Asus motherboard (Nuvoton NCT6798D ISA Sensor Device), but that doesn't make sens because all monitoring softwares call them DIMM temperature, and not "DIMM slot" temp.
          In this case, if I have no possibility to monitor the real temperatures inside the RAM modules, only I can guess they must be higher.
          What concerns me, is that when I play games, this DIMM[] temperatures increase to around 55c because of the GPU hot air inside the case and hot weather.
          It means I am getting too close to the limit I found (58) where I can encounter some memory corruption on my system and risk to loose data.
          It's why I tried to relax CL just by one unit from 15 to 16, in order to have more 'appropriate' 16-16-16-36 at 4000Mhz, and thanks to this timings it become stable, even as high as 64 (max tested, can't go higher), no matter CPU temperature.







          Comment


          • #6
            Add the screenshot about my errors starting from 58c at default XMP:

            Click image for larger version

Name:	Stabilite defaut error 3.png
Views:	122
Size:	2.34 MB
ID:	164589

            Comment


            • #7
              I did some testing again, and I get rid of the errors just by increasing RAM voltage from 1.50 to 1.52v.
              No more errors, even as high as 64c, and even with tighter secondary timings!
              It means that either voltage was not enough, either such high speed and low latency was too optimistic (4000Mhz CL15)
              My point is not to blame G.Skill, but RAM stability is critical, in order to avoid silent corruption and data loss.
              It should remain stable at least until 60c (not mentioning Jedec 85c)
              It's why it's very important to add the key information of max safe temperature for running XMP in the spec, such as:
              XMP1: 4000Mhz 15-16-16-36 max 55c
              XMP2: 4000Mhz 16-16-16-36 max 60c
              XMP3: 3800Mhz 16-16-16-36 max 65c


              Thank you to consider my recommendation to improve your nice products even more compare to other brands...


              Click image for larger version

Name:	stress 4000 15-16-36 tuned 1.52v IO1.15 SA1.20 - 6XX.png
Views:	81
Size:	755.1 KB
ID:	164639filedata/fetch?filedataid=1589

              Comment


              • #8
                Just to say this modules are beasts!
                So fast, even at default primary timings, and with my personal touch for secondary timings!
                Moreover, it is safe until 64c at 1.52v!
                Good job G.Skill, I've ordered a second set of this very same modules, to have 32GB of crazy fast memory


                Click image for larger version  Name:	RAM 4000 CL15 - default primary - optimized secondary.png Views:	0 Size:	1.65 MB ID:	164641
                Last edited by Reflex78; 06-15-2020, 08:45 AM.

                Comment


                • #9
                  I received my new set of the exact same kit for a total of 32GB (4x8GB)
                  They work perfectly fine all 4 together

                  Except for the same problem of too low temperature to run default XMP which starts to throw errors above 50c.
                  In my case, my dimms can reach 55c during heavy gaming (big air GPU)
                  I solved this problem just by increasing a little bit the RAM voltage from 1.50 to 1.52v, which gives about 10c more safety room until 61c

                  Click image for larger version  Name:	stress 15-16-36 opti 1.52v 1.20-1.25 61-err mem.png Views:	0 Size:	723.9 KB ID:	164674


                  About bench, in addition to the bigger RAM size, I can see some improvements for copy bench (54->57GB/s) but a slight worse latency (37.5->38.5ns)

                  Click image for larger version  Name:	bench.png Views:	0 Size:	213.8 KB ID:	164673

                  Nonetheless, I beat my record at Time Spy CPU bench with a score of 13 455 !
                  https://www.3dmark.com/spy/12542919
                  (with my daily usage high stability setting)
                  This benchmark scales very well with faster RAM.
                  Just check the ranking: https://www.guru3d.com/articles-page...review,18.html


                  Overall, I am very happy with my purchase, especially for Ripjaws with 2 things I like more vs Trident:
                  - no need of background process just to turn off RGB
                  - low profil size in the middle (made on purpose?) providing better clearance to the first dimm slot which is just next to the CPU AIO cooler (like my NZXT X62)


                  Click image for larger version

Name:	20200617.jpg
Views:	67
Size:	422.4 KB
ID:	164678
                  Last edited by Reflex78; 06-19-2020, 11:08 AM.

                  Comment


                  • #10
                    Just to add last information about my topic subject of RAM temperature, and as said previously by G.SKILL Administrator, the thermal sensor (which is located somewhere on the PCB), is not giving the exact temperature inside the memory chip, which should be higher. But how much higher? Only the maker know, because it depends on its exact location.
                    I found some technical information below in case some people are interested in this subject.
                    Again, RAM temperature is a critical point to monitor, in order to avoid system corruption and data loss.
                    And the only tool we have at our disposal, is the report DIMM thermal sensor.
                    In my case, default XMP with 1.50v can rises rapidly above 50c during heavy gaming, because of big air GPU and high ambiant temperature.
                    And the difference/delta is too close to the limit where my memory IC chips start throwing errors because of the degraded signal caused by higher temperature.
                    For peace of mind, I recommend people to check their own safety temperature limit, especially with 1.50v modules that can increase faster.
                    And if too close to the limit, just rise the voltage a little to have more safety room (in my case +10c limit with 1.52v)

                    Memory binning is very interesting process, and I wonder if someone at G.Skill can provide more general information (not confidential) like a ranking of their binning chips to know for instance if my new 4000 CL15 at 1.50v is better than the old 4000 CL17 but running lower voltage 1.35v (can it get to the same CL15 just by increasing voltage to 1.50v?)
                    Because sometime, prices don't always reflect the real performance ranking...
                    Thank you G.Skill!


                    http://ww1.microchip.com/downloads/e...Doc/22327C.pdf


                    Click image for larger version  Name:	ddr4_temp_sensor.png Views:	0 Size:	610.0 KB ID:	164701
                    Last edited by Reflex78; 06-21-2020, 12:25 PM.

                    Comment

                    Working...
                    X