OCAU Ryzen Overclocking discussion/ Guide / Owners Thread

Discussion in 'AMD x86 CPUs and chipsets' started by mAJORD, Mar 3, 2017.

  1. Jedi Jezz

    Jedi Jezz Member

    Joined:
    Mar 13, 2003
    Messages:
    938
    Location:
    Canberra
  2. MUTMAN

    MUTMAN Member

    Joined:
    Jun 27, 2001
    Messages:
    8,325
    Location:
    4109
    remember the old s939, with the memory dividers?
    sometimes a higher memory frequency would work where a lower one wouldn't
    i dont know how the IMC works on the zen cores but from that drawing in skitz's post you can see something same same but different at work
     
  3. Jedi Jezz

    Jedi Jezz Member

    Joined:
    Mar 13, 2003
    Messages:
    938
    Location:
    Canberra
    My last AMD chip was a Athlon XP 2500+ :)
     
    MUTMAN likes this.
  4. cheddle

    cheddle Member

    Joined:
    Aug 23, 2014
    Messages:
    39
    Location:
    Hobart, Tasmania
    I actually had the opposite experience haha. I couldn't get 3500c16 on 1.52 however I now have it stable(er) on 1.35. These sticks arent the best though... I tried 3600c14 on 6 different CPU's (7900x/2x7800x/7640x/2700x) and couldnt get it stable... I could get 4000c17 though??? *shurg* im not the best at the memory overclocking

    The x470 Taichi is a super capable board with loads of settings exposed - however the BIOS is difficult to work with and dosnt really help you out... some slight 'bugs' bother me but are not really a big deal. I havnt messed around with (mainly because I haven't found it) CLDO_VDDP but I can comment then in my situation 3500 is more stable than 3333

    Im probably going to get either the gigabyte x470 gaming 7 or the asus crosshair VII once my x299/7800x sells - not too sure what board though... advice?

    I do talk to SKITZ0_000 about memory clocking, they know their shit... and the proof is in the pudding with the results they get.
     
    Last edited: Aug 21, 2018
  5. MUTMAN

    MUTMAN Member

    Joined:
    Jun 27, 2001
    Messages:
    8,325
    Location:
    4109
    Asus fanboys will say c7h
    GB fanboys will say G7
    Id lean towards the c7h, but i dont know shit about the subject tbh :)
    Reviews ive read are saying both are a safe bet
    Asus do roll a lot of firmware updates on their top models though ...
     
  6. Jedi Jezz

    Jedi Jezz Member

    Joined:
    Mar 13, 2003
    Messages:
    938
    Location:
    Canberra

    I found CLDO_VDDP last night, I think it was here:

    UMC Common Options >> NBIO Common Options >> NB Configuration >> CLDO_VDDP Control

    It explains in the manual where to find it if that doesn't make sense.
     
    Last edited: Aug 21, 2018
  7. SKITZ0_000

    SKITZ0_000 Member

    Joined:
    Jul 19, 2003
    Messages:
    1,351
    Location:
    Tas
    Post BIOS update (to AGESA 1.0.0.4) I've been tinkering again, I've moved to a Be Quiet Pro 4 air cooler, trying to keep rig quiet at idle/light usage.

    I've been struggling to get 3600-C14 stable on new bios (it's hard enough getting it to post) so for now I'm running at 3500-C14 on the 3466 Zen+ V1 FAST profile, from the Ryzen DRAM Calc, with some borrowed values from the 3533 EXTREME profile.

    This is 4h RealBench Stable (with CPU -50mv) and 10 hour (overnight) memtest with 1.475v DRAM, 1.15V SoC.

    Some interesting science:

    This is my Cinebench with RAM above and CPU @ Stock clocks w/101 BCLK and -50mv offset, XFR boosts to [almost] 4.4Ghz, all-core boost is ~ 4040mhz, temps peaked at 57C (my current daily setup).


    Click to view full size!


    This is the exact same settings as above (101BCLK -50mv) with Precision Boost Overdrive enabled, cTDP lifted to 125 and Processor temperature control on manual with limit set at 80. (these settings are under Advanced\AMD CBS\NBIO Common Options).


    Click to view full size!


    Pretty decent increase, temps peaked at 69C clocks were just shy of 4.2Ghz all cores (41.5 x 101). I'm yet to thoroughly stress test this, and still want to see if there is more in the tank here, but this is looking good.

    Without lifting the Processor Temp Limit, I get a lower Cinebench score using Precision Boost Overdrive as it blows straight through that 60C threshold.

    I can send some BIOS screenies when I get home
     
    MUTMAN likes this.
  8. MUTMAN

    MUTMAN Member

    Joined:
    Jun 27, 2001
    Messages:
    8,325
    Location:
    4109
    Some nice on air clocks there :thumbup:
     
  9. the_antipop

    the_antipop Member

    Joined:
    Jun 11, 2005
    Messages:
    2,279
    Location:
    Gold Coast
    Hey all,
    I've been tinkering with my Overclocks for my CPU, Ram and GPU.
    I feel I have a stable CPU overclock at 3.85ghz @ 1.35v. My GPU overclock seems to be stable without a change to the voltage (I did just reapply thermal paste, and the temps dropped 5 degrees).
    The major issue I'm having is my memory. I have been able to loosen the timings to 18-18-18-42 and get them to run at 2800. Which seemed to be stable. However, I did do a slightly higher overclock to my GPU, and while playing PUBG I got a slight artifact and lock up. After I restarted, Ram and CPU overclock went back to default. What kind of issue is that? Voltage? PSU?
     
  10. SKITZ0_000

    SKITZ0_000 Member

    Joined:
    Jul 19, 2003
    Messages:
    1,351
    Location:
    Tas
    What is the XMP speed of the DDR4? I'm quite familiar with the AB350 Pro4, basically it doesn't have manual SoC control, instead it automatically sets the SoC based off the frequency and XMP setting.

    Are you selecting XMP or are you just manually setting the memory clock frequency? If you select XMP it should automatically run the SoC at ~ 1.08v which should be plenty for up to ~ 3200-C16 which is hopefully the issue you're having. Use HWinfo or something to confirm SoC voltage in windows.

    Otherwise download and fire up Thaiphoon Burner click "Read SPD on SMBUS" for one of your sticks and post a screenshot of the result. We should be able to identify your modules for Ryzen Calc settings.

    edit: some extra

    Visual artifacts are almost always GPU memory, so I would back that off in the first instance.

    Are you stability testing any of this? A good all-rounder CPU/GPU/RAM is RealBench for a few hours (~4), for GPU only testing a few runs through 3dMark graphics tests will give you a rough idea of GPU stability, you could use Furmark or Kombuster as well, though I don't find them as strenuous these days, but are handy for longer periods of testing.

    A big one though is dedicated memory testing, something like HCI MemTest is easy and seems pretty reliable (and quicker to come up with errors) for in Windows (compared to booting into MemTestx86). Best way for you would be to run 16 instances of HCI MemTest with 1GB memory in each, for several hours. If your memory is causing significant issues it'll come up with errors pretty fast, within the first 15 mins or so.
     
    Last edited: Aug 28, 2018
  11. the_antipop

    the_antipop Member

    Joined:
    Jun 11, 2005
    Messages:
    2,279
    Location:
    Gold Coast
    [​IMG]

    Yeah, I tested the stability for the GPU in heaven benchmark... Might run a memtest once I OC my Ram
     
  12. SKITZ0_000

    SKITZ0_000 Member

    Joined:
    Jul 19, 2003
    Messages:
    1,351
    Location:
    Tas
    I'm concerned that your XMP runs at 1.20V, are you increasing your DRAM voltage? Can you check what your SoC is running at? It might not be auto scaling your SoC at all...

    I would try the attached timings, at 2800 then 2933/3000/3200 and see where you end up, then lower the DRAM based off stability. But if your SoC isn't increasing, you'll have next to no hope.
     

    Attached Files:

    darkbastard and MUTMAN like this.
  13. the_antipop

    the_antipop Member

    Joined:
    Jun 11, 2005
    Messages:
    2,279
    Location:
    Gold Coast
    That screenshot is not running XMP. Just Auto and default timings. I will turn on XMP when I get home this afternoon, and increase the Dram voltage to 1.425 (from your image), as well as implement those timings and try 2800 and higher speeds.

    EDIT: Had some real problems with the overclock. Tried the new timings, and got it to go to 2800 using XMP. All good. Tried to go to 2933. No go, CMOS reset. Took quite awhile to boot. All overclocks were reset. CPU overclock was then upstable, and even with the changed timings, I could not get 2800 again... Wish there was an easy guide to Ryzen overclocking for my ram. Should have invested in some better ram :(
     
    Last edited: Aug 30, 2018
  14. TaroT

    TaroT Member

    Joined:
    Jan 18, 2002
    Messages:
    8,723
    Location:
    Hazelbrook nsw 2779
    Hynix is definitely iffy on these things hell even my b die team extreme won,t go anywhere what I would suggest you try is 2400 and 14 14 14 and a few tighter ones see how that goes run wprime cb15 blah blah then up to the stock 2666 and run them again and see if there is a huge difference.
    going over on those sticks I really do not see happening.

    also have you used the ryzen dram calculator I found that pretty useful (can't remember where it is google it :)
     
  15. OJR

    OJR Member

    Joined:
    Jan 19, 2013
    Messages:
    4,952
    Location:
    Melbourne
  16. the_antipop

    the_antipop Member

    Joined:
    Jun 11, 2005
    Messages:
    2,279
    Location:
    Gold Coast
    I'm going to keep prodding away until I can get it to work at 2800 again, and then just accept the fact that it has to stay at that speed.
    Does anyone know the most ideal/'safe' voltages for Ram on a B350 with a 1700? Is it 1.35v? Would that make it stable enough without sucking too much power?
    Likewise for my 1700. I have overclocked to 3.85ghz, what's a safe voltage for that? I think I ran 1.3185 (~1.32v) - should it be higher?
     
  17. Ratzz

    Ratzz Member

    Joined:
    Mar 13, 2013
    Messages:
    10,037
    Location:
    Cheltenham East 3192
    If the 3.85 GHz CPU overclock is stable at 1.32v, I'd just leave it there.

    As far as the RAM goes, my c14-3200 Flare-X runs XMP @ 1.35v, so that's definitely ok. I'd be fine with 1.4v 24/7 personally, which is what I use if I lift the RAM to c14-3600. It works fine at 3600 with 1.4v, no other changes apart from frequency, but gets slightly flaky sometimes..

    Flash in particular seems to crash (occasionally) when I bump it up like that. Its not the volts, its the fact that I've made no changes to the timings.. Just lifted frequency to 3600 and added the slight extra bit of voltage. If I spent the time on getting the timings right, it would be rock solid I am certain.

    I've been known to run RAM at 2v for extreme overclocks on my Intel systems, but I wouldn't recommend going that high too often :lol: 1.4v is pretty safe.
     
    Last edited: Sep 3, 2018
    the_antipop likes this.
  18. Sgt Bilko

    Sgt Bilko Member

    Joined:
    Mar 5, 2015
    Messages:
    1,121
    Location:
    Canberra(ish)
    the_antipop For that board I'd stick to 1.4v as a max recommendation.

    Core voltage you'll most likely be limited by the motherboard itself rather than the CPU but for 1st Gen you should stick to 1.4v or lower for max reliability.
    If you can get 3.85Ghz stable on 1.35v or lower then I'd keep it there :)
     
    the_antipop likes this.
  19. the_antipop

    the_antipop Member

    Joined:
    Jun 11, 2005
    Messages:
    2,279
    Location:
    Gold Coast
    Could the overclocking be limited by the fact my ram is in single channel, rather than dual channel? Would that make a difference?
    For some reason my ram and my motherboard don't like dual channel. It boots, and in bios it shows 16gb of ram - but windows says there is only half the ram. Bizarre error...
     
  20. OJR

    OJR Member

    Joined:
    Jan 19, 2013
    Messages:
    4,952
    Location:
    Melbourne
    huh? That doesn't sound right. Are you sure your board isn't stuffed?
     

Share This Page

Advertisement: