AMD RyZen CPU Architecture for 2017

Discussion in 'PC Industry' started by fellix, Oct 20, 2014.

Tags:
  1. Alexko

    Veteran Subscriber

    Joined:
    Aug 31, 2009
    Messages:
    4,264
    Likes Received:
    682
    How much is that?
     
  2. Clukos

    Clukos Bloodborne Ps4 Pro when?
    Veteran Newcomer

    Joined:
    Jun 25, 2014
    Messages:
    4,224
    Likes Received:
    3,454
    fellix, Lightman, fehu and 2 others like this.
  3. Sinistar

    Sinistar I LIVE
    Regular Subscriber

    Joined:
    Aug 11, 2004
    Messages:
    645
    Likes Received:
    60
    Location:
    Indiana
    At 1440P, on the 4770K with it would have drops to the mid 40's, now never below 60.
    I guess I am assuming its the memory, and not Ryzen, have not tried lower speeds.
     
    #2543 Sinistar, Sep 11, 2017
    Last edited: Sep 11, 2017
    Lightman likes this.
  4. Malo

    Malo YakTribe.games
    Legend Veteran Subscriber

    Joined:
    Feb 9, 2002
    Messages:
    5,895
    Likes Received:
    1,901
    Location:
    Pennsylvania
    I think my PC components for 2018, in addition to a new GPU, will be a CLC for my Ryzen and some Samsung b-die C15 3600 RAM.
     
  5. Alexko

    Veteran Subscriber

    Joined:
    Aug 31, 2009
    Messages:
    4,264
    Likes Received:
    682
    It would be worth a quick bench, methinks, but either way it's very interesting.
     
  6. Sinistar

    Sinistar I LIVE
    Regular Subscriber

    Joined:
    Aug 11, 2004
    Messages:
    645
    Likes Received:
    60
    Location:
    Indiana
    I was looking at Kyle's waterblock article for Threadripper, on the last page he has images of AM4 cooling with the same blocks.
    1504742086n1f4pi37ul_3_7_l.png
    I don't know if he is just trying to see how many watts he can push it to or what? I can run the same test and have a peak power draw of 133 watts on CPU core power, no where near the 222 he is getting.

    EDIT: I am using an H110i for cooling
     
  7. Clukos

    Clukos Bloodborne Ps4 Pro when?
    Veteran Newcomer

    Joined:
    Jun 25, 2014
    Messages:
    4,224
    Likes Received:
    3,454
    Dota 2 with Vulkan ram speed difference in fps:

    2666 auto timings
    [​IMG]

    3466 C14 optimized
    [​IMG]
     
    Grall, fellix, Malo and 2 others like this.
  8. Lightman

    Veteran Subscriber

    Joined:
    Jun 9, 2008
    Messages:
    1,732
    Likes Received:
    404
    Location:
    Torquay, UK
    I will join Clukos and show CS:GO at 2560x1440 max. on Vega56 with my Ryzen 1700 @4GHz
    DDR2666 Auto 14-14-14-34-48 1T:
    [​IMG]

    DDR3200 Optimized 14-14-14-28-42 1T:
    [​IMG]

    As it was live game, FPS fluctuated, ranges were:
    2666Auto: 160 to 250FPS (average 220FPS)
    3200Optimized: 248-310FPS (average 275FPS)

    EDIT:
    Productivity app. gains:

    Frybench 64bit

    2666 Auto:
    [​IMG]

    3200 Optimized subtimings:
    [​IMG]

    Just to show it is not only games which benefit from fast memory and tight subtimings ;)
     
    #2548 Lightman, Sep 11, 2017
    Last edited: Sep 11, 2017
    Kej, jiaolu, fellix and 2 others like this.
  9. Clukos

    Clukos Bloodborne Ps4 Pro when?
    Veteran Newcomer

    Joined:
    Jun 25, 2014
    Messages:
    4,224
    Likes Received:
    3,454
    Even OpenGL apps like CEMU (emulator) which rely heavily on single threaded performance see gains:

    [​IMG]
    [​IMG]
     
    Lightman likes this.
  10. Sinistar

    Sinistar I LIVE
    Regular Subscriber

    Joined:
    Aug 11, 2004
    Messages:
    645
    Likes Received:
    60
    Location:
    Indiana
    Just tried at the same speed ram of my 4770K, which was only 1600 CAS9. The short time I played I saw dips as low as 50.

    Also seems to be what causes stutter on the animation in the games main load screen.
     
    #2550 Sinistar, Sep 11, 2017
    Last edited: Sep 11, 2017
    Alexko likes this.
  11. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,607
    Likes Received:
    1,780
    Location:
    Germany
    As long as the CPUs are good value - who cares
    At the same voltages? Strange, you'd be ~25% below TDP on full-core load with raised OC-voltages. Sure of that? I reach TR's TDP in non-OC'ed conditions already.
     
  12. Sinistar

    Sinistar I LIVE
    Regular Subscriber

    Joined:
    Aug 11, 2004
    Messages:
    645
    Likes Received:
    60
    Location:
    Indiana
    No, vcore at 1.356, llc at level 3
    That was my point, is that voltage nesesary?
     
  13. Malo

    Malo YakTribe.games
    Legend Veteran Subscriber

    Joined:
    Feb 9, 2002
    Messages:
    5,895
    Likes Received:
    1,901
    Location:
    Pennsylvania
    3.9Ghz at 1.440v? Not that I've seen, though he could have a really bad 1700, or he just isn't that familiar with Ryzen OC optimization.
     
  14. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,607
    Likes Received:
    1,780
    Location:
    Germany
    True, I've seen far less volts for 3,8-3,9 GHz, also for TR.
    Maybe he wanted to test the waterblocks instead of the CPUs and put artificially high voltage on the cores in order to create more of a challenge?
     
    Lightman likes this.
  15. Clukos

    Clukos Bloodborne Ps4 Pro when?
    Veteran Newcomer

    Joined:
    Jun 25, 2014
    Messages:
    4,224
    Likes Received:
    3,454
    Putting Agner Fog's findings to the test: http://www.agner.org/optimize/blog/read.php?i=838
    1 core + 1 smt thread (in the same core), Cinebench + CPU-Z test run:

    [​IMG]

    Same scores that a 7600K is getting at 5.2 GHz (without HT). I'm pretty sure it beats pretty much everything when 1 core + 1 smt thread is taken into account clock for clock.
     
    fellix likes this.
  16. fellix

    fellix Hey, You!
    Veteran

    Joined:
    Dec 4, 2004
    Messages:
    3,415
    Likes Received:
    287
    Location:
    Varna, Bulgaria
    Same benchmarks on i7-6800K @ 4GHz, single core + HT:

    CB15 -- 331
    CPUZ -- 340
     
  17. Clukos

    Clukos Bloodborne Ps4 Pro when?
    Veteran Newcomer

    Joined:
    Jun 25, 2014
    Messages:
    4,224
    Likes Received:
    3,454
    Your CB15 score is too high and CPU-Z too low :) Are you sure you are doing this correctly? Setting up Cinebench R15 can be a pain because if you set it to 2 threads it selects two cores instead of 1 core and 1 smt thread. So you have to set affinity manually the moment cinebench starts. CPU-Z is easier to work with, just set core 0/1, or 2/3, 3/4 etc. before starting the benchmark and hit stress test.
     
  18. fellix

    fellix Hey, You!
    Veteran

    Joined:
    Dec 4, 2004
    Messages:
    3,415
    Likes Received:
    287
    Location:
    Varna, Bulgaria
    I set the processor affinity manually for both applications.
    CPUZ behaves weird and does produce lower result with two thread selected in the app, so I left it now by default (12 threads) and the result is 608 while still sticking to the CPU affinity mask.
     
  19. Clukos

    Clukos Bloodborne Ps4 Pro when?
    Veteran Newcomer

    Joined:
    Jun 25, 2014
    Messages:
    4,224
    Likes Received:
    3,454
    Cinebench is resetting the affinity each time you press run (just a quirk of the program) so you have to set the affinity after you press run (I just have the windows open with affinity set and press ok after I hit run) 608 for CPU-Z is the expected result just a bit behind the 625 that I'm getting at 4.0 GHz :)
     
    BRiT likes this.
  20. Clukos

    Clukos Bloodborne Ps4 Pro when?
    Veteran Newcomer

    Joined:
    Jun 25, 2014
    Messages:
    4,224
    Likes Received:
    3,454
    Re-run, this time without Chrome in the background, Core 0 + 1:

    [​IMG]
     
    Lightman likes this.

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...