Announcement

Collapse
No announcement yet.

Crucial 2 x 16GB DDR5-5200 / DDR5-5600

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

  • #11
    Originally posted by BillBroadley View Post

    I looked and didn't find any DDR5-4800 performance wins. Careful on the graphs, the DDR5-4800 often won on performance per $
    Aw geez, you're right. They're all per-dollar. Sorry about the truly-stupid question.

    Comment


    • #12
      Recent testing by Hardware Unboxed on B650 motherboards produced some interesting results regarding RAM. Memory timings were all over the place on different boards, some of them wouldn't actually run the sticks at EXPO speeds, and boot times for Windows 11 ranged from 23 to 55 seconds depending on the board with the exact same kit of RAM. I trust that Phoronix' results are valid for the test platform, but with all the hosery going on in the video I have to wonder if they can be applied as widely as one would like.

      The video I referenced is here, using a set of 2xAorus 16GB DDR5-6000. What the motherboards did to the memory timings starts at 16:48, and the boot time chart at 18:10.

      Comment


      • #13
        Originally posted by Teggs View Post
        Recent testing by Hardware Unboxed on B650 motherboards produced some interesting results regarding RAM. Memory timings were all over the place on different boards, some of them wouldn't actually run the sticks at EXPO speeds, and boot times for Windows 11 ranged from 23 to 55 seconds depending on the board with the exact same kit of RAM. I trust that Phoronix' results are valid for the test platform, but with all the hosery going on in the video I have to wonder if they can be applied as widely as one would like.

        The video I referenced is here, using a set of 2xAorus 16GB DDR5-6000. What the motherboards did to the memory timings starts at 16:48, and the boot time chart at 18:10.
        probably because many manufactures haven't done qvl testing yet. ram is actually really hard and it doesn't help memory vendors don't really stick to one type of memory dies per model. corsair is absolutely notorious for this. their vengeance kits for example can ship with three or four different dram dies. so, say gigabyte, gets a kit of ddr5 vengeance to test and their kit has micron dies on it, but the kit you get has hynix. and another person gets a kit and its samsung. well gigabyte did qvl testing for micron for that kit. the timings they did are tested around micron, not any of the other brands. resulting in those timings are unstable for non micron versions of that corsair kit.

        then to make matters worse, lets say it's hynix this time, and originally that kit had hynix m-die and now they switched it to hynix a-die. more pain. now some manufactures will go ahead and also look at what brand of memory are actually on the kits and apply settings for that, but not all do. many just go by the actual model of the kit. like for example my teamgroup vulcan ddr5 64gb kit on my msi z690 edge wifi was listed as micron. mine happens to be hynix m-die. luckily, msi put in timings for hynix instead of micron as msi actually will also check the die on the memory as well and gives me three XMP profiles to pick from. default profile being the most compatible, the others offering tighter timnigs. memory simply put is just a massive mess.

        one of the reasons why i was so sad that crucial killed off their ballastix line. high performance ram that was pretty much guarantied to always have micron on them. so at least you knew, if you got a ballastix kit, it was micron. really simple for motherboard manufactures to QVL.
        Last edited by pieman; 08 March 2023, 09:44 PM.

        Comment


        • #14
          Originally posted by Velocity View Post
          Since you can only use two DIMM slots on AM5 socket due to abysmal speeds when using 4 DIMMs
          What?

          Comment


          • #15
            Originally posted by Weasel View Post
            What?
            See: maximum speed of 4 DIMMs used on the 7950X3D chip.

            Comment


            • #16
              I have run DDR5-5600 SO-DIMM with the Ryzen 9 6900HX. It works for the most part and some people OC it to 6000. But for now it is Linux and BIOS release sensitive.

              According to AMD, that 6900HX will support up to DDR5-6400, but the world hasn't gotten there just yet.

              Comment


              • #17
                Originally posted by Velocity View Post
                Thanks. That's a big oof. This CPU is starting to get less and less appealing, and I really wanted it.

                First the 3D cache is only on half the cores, and now the RAM is even slower which means the cores without cache will be even more gimped. And yes I need 4 sticks.

                Nice design.

                Comment


                • #18
                  Well you don't need 4 sticks once there are good 64GiB sticks available. Then you can use 2x64GiB = 128GiB total memory with good timings and frequency.

                  The 48GiB ones appear usable after the AGESA update. But it is unclear when 64GiB DDR5 UDIMM will start entering the market. I hope soon enough. I considered buying 2x8GiB sticks and use it only for gaming in the meantime. My VMs consume up to 64G each, cannot do that on a system with only 64GiB total memory.

                  Comment


                  • #19
                    It could be great to see how much memory speed affects the X3D CPUs.
                    I haven't read up on the clocking of the X3D cache, is it someway affected by memory bus speed?

                    Comment

                    Working...
                    X