Announcement

Collapse
No announcement yet.

DDR4 Memory Scaling & DDR4-3600 Testing With AMD Threadripper On Linux

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

  • oooverclocker
    replied
    Originally posted by coder111 View Post
    Hmm, it's a bit of a shame highest speed ECC memory I can find is DDR4-2666.

    I'd love to have ECC, but if does look like it would be a substantial hit on performance... And probably my wallet as well.
    I have 2x 16 GiB DDR4-2400 ECC Memory @3200 MHz. The RAM clocks are really important for Zen CPUs. Not just for EPYC, the same counts for Threadripper and Ryzen.
    There are many benchmarks that draw a weak picture for Ryzen when it comes to gaming. But usually when you see which memory has been utilized you have the explanation for the weak performance.
    With the right memory clocks you have about the same performance per clock like Intel CPUs offer for a much lower price.

    Leave a comment:


  • rrohbeck
    replied
    Originally posted by coder111 View Post
    Hmm, it's a bit of a shame highest speed ECC memory I can find is DDR4-2666.

    I'd love to have ECC, but if does look like it would be a substantial hit on performance... And probably my wallet as well.
    ECC memory overclocks too and you can see exactly when you're going too far. On my FX-8350 I got 1866 with decent timing out of garden variety DDR3-1333 Kingston ValueRAM DIMMs, it just took some time to find the right settings.

    Leave a comment:


  • AndyChow
    replied
    Originally posted by coder111 View Post
    Hmm, it's a bit of a shame highest speed ECC memory I can find is DDR4-2666.

    I'd love to have ECC, but if does look like it would be a substantial hit on performance... And probably my wallet as well.
    Faster ECC sticks should start appearing Q12018. There are two chips currently missing to make it happen. One can be sampled, but only pre-ordered, the other should be available for sampling soon. I forget what they are called.

    Leave a comment:


  • Luke_Wolf
    replied
    Originally posted by Kemosabe View Post
    Does this mean any 4x configuration > 2x?
    Within the confines of Threadripper, yes. This is not generalizable to anything that has less than 4 memory channels such as the normal Ryzen parts. You basically get a lot more bandwidth by filling channels than you do from increasing the frequency of a particular channel.

    Leave a comment:


  • Mark Rose
    replied
    The 4GB sticks are almost certainly single rank modules. Do you know if the 8 GB modules are single or dual rank? You can check by running `dmidecode | grep -oP "Rank: [1-4]+"`. Dual rank allows for more rank interleaving which can impact memory bandwidth.

    Leave a comment:


  • droste
    replied
    Originally posted by Kemosabe View Post
    Does this mean any 4x configuration > 2x?
    If you run them in quad-channel mode, yes. (and you don't try to run a 4x64MB configuration instead of 2x8GB configuration of course)

    Leave a comment:


  • droste
    replied
    Originally posted by Ehvis View Post
    "The memory timings were not tweaked between the different frequency levels."

    Can you clarify what that means? Because if it means that the timings were kept the same for tests, then the test would be a bit silly.
    Pretty sure he means they were not changed from the default value each stick has. As in "I didn't change anything in the bios".

    Leave a comment:


  • Kemosabe
    replied
    Does this mean any 4x configuration > 2x?

    Leave a comment:


  • coder111
    replied
    Hmm, it's a bit of a shame highest speed ECC memory I can find is DDR4-2666.

    I'd love to have ECC, but if does look like it would be a substantial hit on performance... And probably my wallet as well.

    Leave a comment:


  • carewolf
    replied
    Many of those "CPU" benchmark were better memory speed benchmarks than the official memory speed benchmark!

    Leave a comment:

Working...
X