Comments Locked

12 Comments

Back to Article

  • ScottSoapbox - Saturday, December 16, 2017 - link

    I prefer the 64 GB 3200 I got for the same price.
  • LordanSS - Saturday, December 16, 2017 - link

    Probably has better timings as well.
  • ddrіver - Sunday, December 17, 2017 - link

    Timings are irrelevant in most cases, especially when it comes to high end products. They're all pushed to the limit of the memory cell anyway. If you count in nanoseconds it's more or less the same. You increase the frequency so the same operation takes more cycles but the same real time.

    But why am I not suprised that AT audience is still at the "chasing numbers" level.
  • willis936 - Monday, December 18, 2017 - link

    Because the numbers are worth chasing. Specifically baud / cas
  • jabber - Monday, December 18, 2017 - link

    I just make sure if 2T is set I change it to 1T and that's me done. I left all the timing tweaks back in the day of my nForce4 DFI LanParty 939 board.
  • ddrіver - Monday, December 18, 2017 - link

    @willis936: "Worth"? Is "within margin of error in some benchmark" your definition of "worthiness"?

    CAS 19 is a lot worse than CAS 14, right? Like 25% slower. I'll let you in on a little secret. Those numbers are relative. They mean nothing by themselves. CAS 19 for DDR4 4800 means ~7.9ns. CAS 14 for DDR4 3200 means ~8.8ns. "Better timings" right?

    You guys know how to count, I'll give you that. But you sure as hell don't know what those timings mean, or what's better.
  • ddrіver - Tuesday, December 19, 2017 - link

    And just as a sidenote, CL19 for DDR4 4000 is about the same as CL15 for DDR4 3200.
  • The_Assimilator - Thursday, December 21, 2017 - link

    Well, duh. That's why you need to calculate the "performance rating", which you get by dividing the maximum bandwidth by the CAS latency.
  • ddrіver - Thursday, December 21, 2017 - link

    Lo, no. You just look at the timings and victoriously conclude that the DDR4 3200 CL15 is 25% better than the DDR4 4000 CL19.

    When talking about binned products you don't need to calculate much. The chips are already pushed to the highest frequency and lowest equivalent timings, which is basically almost always the limit of the silicon. The point where you can't get better timings without lower frequency, or higher freq without worse timings. Any difference at this level is purely academical.

    And you can almost see the wheels slowly turning in the head of the guy who was just saying that "the numbers are worth chasing". Especially CAS, ofc.

    These day I feel like I can pop open a bottle of champagne every time I see an intelligent comment around here.
  • flgt - Thursday, December 21, 2017 - link

    Don't you still get a performance bump with higher data rate even if the transaction latency is constant since you pull large blocks of memory into cache?
  • ddrіver - Saturday, December 23, 2017 - link

    Depends on what you're doing. The point is chasing numbers like "low latency" is very tricky if you have no idea what that latency means. You end up like willis936 or LordanSS complaining about a memory kit like this one because it's DDR4 4000 but the CL is 19, a lot "worse" than the DDR4 3200 CL 15. Because 15 is better than 19 all other factors be damned.
  • levifig - Tuesday, December 19, 2017 - link

    Well, if you bought it a year ago, maybe… I doubt you can buy 64GB of DDR4 3200Mhz without spending almost double ATM!

Log in

Don't have an account? Sign up now