Post your Ryzen memory speeds!

Discussion in 'AMD Processors' started by bobzdar, Mar 15, 2017.

  1. cybereality

    cybereality [H]ard|Gawd

    Messages:
    1,970
    Joined:
    Mar 22, 2008
    GIGABYTE GA-AB350M-HD3
    G.SKILL FORTIS Series 16GB (2 x 8GB)

    Loaded up out the box at 2666 speed with 18 CAS. I really recommend going FORTIS or Flare X for best compatibility with Ryzen.
     
  2. noko

    noko 2[H]4U

    Messages:
    2,385
    Joined:
    Apr 14, 2010
    The Mortar definitely looks to be a nice board, it is probably either that one or the Biostar B350 GT3 (mainly because of PState overclocking/underclocking which I am not sure the MSI board will do). Now I heard the Asrock board has weak VRMs for that board.
     
  3. Phelptwan

    Phelptwan [H]ardness Supreme

    Messages:
    7,007
    Joined:
    Jul 20, 2002
    Just an update on this. I went ahead and reset bios to stock settings and focused on ram. Tried first to just enable XMP, no dice. I then went ahead and started increasing the memory multiplier (correct terminology?) and was able to max out at 2400mhz (16, 18, 18, 36). Won't boot past that. Once I was stable at 2400, I redid my cpu overclock and am now stable (as far as I can tell) at 4.1ghz cpu and 2400mhz memory.
     
  4. anubis44

    anubis44 I just make shit up

    Messages:
    219
    Joined:
    Jul 30, 2008
    Basically, fastest memory speed I can get is 3147.2MHz using the BLCK multiplier, but this is not stable. Fastest stable memory speed I've been able to get (basically, I can run handbrake to encode h265 in software mode without a crash) is 3080MHz, also using the BLCK multiplier. Relevant bios settings were:

    Ram voltage: Auto (noticed it was hitting 1.41v when running at 3147MHz, and 1.36v max when running at 3080MHz.
    SOC voltage: Auto (never went up above 1.1)

    Even when I tried to o/c the SOC voltage to 1.2, it doesn't make any difference (and I've even gone as high as 1.3v - also no difference).

    I've loosened the timings to ridiculous amounts (20-20-20-20-64, for example), and no matter what I do, I cannot get the system to even post with a memory speed above 3150MHz, come hell or high water, and both of these memory sets are supposed to run at 3200MHz or 3600MHz. I'm beginning to think it's a limitation of the integrated memory controller on my Ryzen CPU, and not my RAM or motherboard that is the problem, as apparently, a guy from Canada Computers was able to dial in 3600MHz memory speed with my model of Asus motherboard (using the same 0083 beta bios) and G.Skill 3600 CL16 RGB RAM!

    System: Asus Crosshair VI Hero
    Bios Rev.: 0082/0083 beta
    RAM: G.Skill Trident Z F4-3200C14D-16GTZR/G.Skill Trident Z F4-3600C16D-16GTZR (so far nearly identical results with both sets)
    Ryzen 7 1700 serial number: 9R64626N70133 - Diffused in USA, Made in Malaysia.

    Special request: I would like to ask those who graciously post their max memory speeds to also try to post their CPU model/serial number/"Made in" country location also so we can attempt to figure out if there's a pattern to which serial number ranges/countries of manufacture lead to the best memory/CPU overclocks? Thanks!
     
  5. JimmiG

    JimmiG 2[H]4U

    Messages:
    2,406
    Joined:
    Apr 3, 2008
    I found this pretty interesting:
    http://www.overclock.net/t/1624603/rog-crosshair-vi-overclocking-thread/9700#post_26020881
    For users of the C6H, make sure you also increase the DRAM Boot Voltage when you increase the DRAM voltage. DRAM Boot Voltage is supposed to track the DRAM voltage (why would you want your 1.35V RAM to boot at 1.2V?), but apparently this doesn't work.

    For those of us whose boards lack this options (such as the Prime X370), I guess we're stuck with the DRAM training happening at 1.2V until AMD and/or Asus get around to fixing it. This explains a lot of the cold boot/POST problems some are having even with RAM that should work at higher speeds. DRAM training happens just before the normal POST screen so if it fails this step, it won't POST.
     
  6. bobzdar

    bobzdar Gawd

    Messages:
    935
    Joined:
    Jun 6, 2003
    If that's the case, that's a pretty huge miss by pretty much all the MB makers..
     
  7. anubis44

    anubis44 I just make shit up

    Messages:
    219
    Joined:
    Jul 30, 2008
    Hi, yes, I've tried boosting the 'initial RAM voltage' also, to 1.45v, and no dice. It does help if you post if you're right on the edge of stability (say, within 20-40MHz), but like I said, no setting allows me to boot this 3600 and 3200-rated G.Skill B-die Samsung DDR4 memory at the DOCP 3200MHz setting, or any manually dialed in memory speed above ~3150MHz, no matter what I do. And it's looking like it's exactly the same cut-off for both kits of RAM, which leads me to the conclusion that my integrated memory contoller is somehow the culprit. If only I had another Ryzen CPU that is known to run at 3200MHz+ memory speeds, I could confirm this. Trouble is, I only know 1 person with a Ryzen 7 (he has an 1800x) but he's 5 hours away in Toronto. I've even contemplated buying another Ryzen CPU (perhaps a Ryzen 5 to put into my HTPC?) just to test this theory, but if IT can't run 3200MHz memory speeds either, I'll be stuck with two Ryzens that can't even hit 3200MHz, although my local bricks and mortar store seems to allow exchanges on the Ryzen CPUs, so perhaps I could just keep exchanging the chip until I hit paydirt? LOL.

    Well, I've also got to take care of the lawn and painting in the house, so I'll just have to see what I do.

    Oh f*ck it. I'll go pick up another Ryzen 7 1700 and try it out. If it doesn't hit 3200+ memory speeds, I'll exchange it until I hit one that does. I'll also post serial#, 'made in' country, etc. I'll start a new thread for this.
     
    Last edited: Apr 23, 2017
  8. CAD4466HK

    CAD4466HK Limp Gawd

    Messages:
    410
    Joined:
    Jul 24, 2008
    Why don't you just return your RAM instead for the G.Skill Flare X 3200MHz kit and and get a guaranteed 3200Mhz?
    I haven't seen a single person not get 3200MHz yet with that kit regardless of mobo.
     
    cybereality and DooKey like this.
  9. JimmiG

    JimmiG 2[H]4U

    Messages:
    2,406
    Joined:
    Apr 3, 2008
    My Corsair Vengeance LPX 3600 (B-die) works at 3200, but only at C16. At C14 I get the cold boot issues, but if I manage to get it to boot (by repeatedly saving and applying the settings), it's rock solid in Windows at C14. I'm pretty sure having the option of booting the RAM at 1.35V instead of 1.2V would make it pass a POST from cold.

    Having DRAM boot voltage working properly would solve those kinds of cold boot problems, but it won't make any difference once the system has POSTed.
     
  10. TsingTao

    TsingTao n00bie

    Messages:
    54
    Joined:
    Nov 22, 2010
    I am running G.Skill Trident Z F4-3200C14D-16GTZR memory in my ASRock Taichi board (BIOS 2.00) at the rated 3200 speed and timings, with a 1700 CPU. I would look to the BIOS or RAM sticks themselves...if it were the IMC I would be hitting the same wall with the same memory kit.

    At work so I can't give you details on the CPU but I'll check when I get home.
     
  11. Nobu

    Nobu Gawd

    Messages:
    984
    Joined:
    Jun 7, 2007
    Gigabyte ax370 gaming k5
    Ryzen 5 1600 (USA/Malaysia)
    G.Skill Ripjaw F4-2800C16D-16GVG

    Booted into windows at 2300 CAS11-11-11-11, but unstable. Currently running memtest86 at CAS14-15-15-15, just under 2600. Good up to test 10 so far (21m)

    BCLK 108, 3.763GHz all cores, +.086VCore, 1.25V Ram
    Edit: this is OOTB, haven't updated the firmware yet. 2400->2600, 12->16->14
    Broke 16.2GB/s in memtest86, probably as fast as I'll get stable with this kit on this firmware version.
    Edit2: Updated to F2 Firmware with agesa 1004a. Can now reach 2800 cas 14-14-14-14 "stable" (occt fails gracefully in first minute, linpack ran 1hr no prob), with bclk at 105.05, but CPU multi won't go above 32 anymore (refuses to pass POST), so I'm stuck at 3.362GHz unless I push the bclk higher. Lower cas settings are unstable. Haven't upped voltages other than vcore (+.082V) and vram (1.25V).
    CPU-Z Validation: https://valid.x86.fr/m7lr32
     
    Last edited: Apr 27, 2017
  12. mdaniel

    mdaniel Limp Gawd

    Messages:
    342
    Joined:
    Jan 8, 2004
    Best I've been able to do with my g.skill is 3162 with a BLCK of 107.8 on the 2933 strap. So far, changing the ram voltage and SOC can't get me a higher ram speed.
     
  13. noko

    noko 2[H]4U

    Messages:
    2,385
    Joined:
    Apr 14, 2010
    DDR4 3500, BCLK 109.4 using strap 3200. 16-16-16-16-38 CR1. Boots right up cold or not. GSkill F4-3200C14D-16GTZKW, Samsung B dies
    https://www.newegg.com/Product/Product.aspx?Item=N82E16820232407

    Both 1700x was able to do DDR 4 3200 without issue, never tested first 1700x with memory above 3300 (which worked also without issue). My other non-used 1700x is going into my HTPC system hopefully this weekend with some Corsairs I got on sale CMK16GX4M2B3200C16W
    https://www.newegg.com/Product/Product.aspx?Item=N82E16820236038

    Anything over DDR 4 2666 is an OC which you may or may not be able to do. that is with two sticks of single sided dies. Hopefully next AMD firmware update will give us more.
     
  14. OpStar

    OpStar Gawd

    Messages:
    586
    Joined:
    Sep 20, 2004
    1600x and Flare X 3200 on Gigabyte B350 Gaming 3. Does 3200 with f6 bios np. Chip does 4ghz @ 1.4v. Haven't tried any other settings tbh.
     
    cybereality likes this.
  15. magictoaster

    magictoaster Limp Gawd

    Messages:
    393
    Joined:
    Apr 6, 2005
    ASUS ROG Crosshair VI Hero (from 1107 to 0003 then back to 1107)

    4x 8GB G.SKILL Flare X Series F4-3200C14D-16GFX

    2666MHz
    14-14-14-34 1.35V --> 14-13-13-30 1.35V
     
    Last edited: May 2, 2017
    cybereality likes this.
  16. egads

    egads n00bie

    Messages:
    6
    Joined:
    Jul 1, 2013
    ASUS ROG Crosshair VI Hero
    BIOS 1002

    Ryzen 1700
    3.85Ghz 100 BLCK
    Anything higher would crash on stress testing

    2x8gb Corsair CMK16GX4M2B3200C16 VERSION 5.39

    3200MHz
    16-18-18-36 1.35V

    Worked with DOCP 5 profile automatically with no extra tweaking
    Haven't yet tried tighter timings
     
  17. Formula.350

    Formula.350 Gawd

    Messages:
    589
    Joined:
    Sep 30, 2011
    It's not the CPU :\ It's all down to the BIOS, and AMD's lack of allowing user adjustment to memory timings. Once they finally (... hopefully) unlock all the Subtimings, we might finally unleash some more performance and speeds. Wish I knew the logic behind their decision to prevent us for changing them :(

    Eitherway, the wall you're hitting clearly seems to be caused by the fact that you're running a much lower memory divisor, and then cranking up BCLK to counter for that. While that is indeed the best way to gain memory performance, it's also going to end up stalling you out at some point. It all comes back to the timings. Lower speeds have tighter sub-timings, and no amount of increase of the primary timings will solve that fact. Chew* @ [XS] forums had suggested to me, while trying to get 14-14-14 stable at 3200 on my 15-15-15 kit, to try less voltage on various things. So perhaps setting DRAM's to 1.35V, and the CPU-NB (memory controller) to 1.0V instead of 1.1-1.2V, might yield better results.There's also the DDR VREF but my knowledge there is non existent. I'm pretty sure that one should be 1/2 DRAM Voltage, which is why at 1.2V it sets it to 0.60V, So for 1.35V it technically should be 0.675V, but again, less could be the ticket.

    Also lets not forgot the LoadLineCalibration settings. It could be that your board, under load, is either over or under volting which may be causing the instability. OR it's just just fluctuating too much (not at a high frequency, ie ripple), and that ends up being enough to throw it off being stable.

    But I digress, I'l be surprised if it's the memory controller to blame, in general or a single case. The most I can speculate there is being since you have a 1700, that you may have a Malaysian made chip instead of a China one. Whether or not the speculation of their being a quality difference on the memory controller was/is true, I dunno. If your's is indeed a Malay, then buy or borrow someone's X chip and see if a China die solves the problem *shrug*
     
  18. noko

    noko 2[H]4U

    Messages:
    2,385
    Joined:
    Apr 14, 2010
    cybereality, egads and CAD4466HK like this.
  19. cybereality

    cybereality [H]ard|Gawd

    Messages:
    1,970
    Joined:
    Mar 22, 2008
  20. waderunner

    waderunner Gawd

    Messages:
    731
    Joined:
    Jun 2, 2013
    MSI B350M MORTAR ARCTIC // BIOS 7A37vA3 (http://www.newegg.com/Product/Product.aspx?Item=N82E16813144046)
    Ryzen 5 1500X (using all AUTO CPU settings in BIOS; I'm in a small case, so not planning to overclock, and why I went with 1500X, to minimize heat; Also, got a good ebay price for $160 :)
    G.Skill F4-3200C15D-16GVK (http://www.newegg.com/Product/Product.aspx?Item=N82E16820232221)
    Rated at:
    DDR4 3200 (PC4 25600)
    Timing 15-15-15-35
    Cas Latency 15
    Voltage 1.35V

    Currently running at:
    A-XMP Profile 1 (15-15-15-35, 2933; CAS 16)
    Auto voltage

    The memory won't currently run well at XMP Profile 2, which is 3200.
    I've gotten it to boot up a couple of times using Profile 2 (3200), but eventually will fail to restart, will cycle through 5 attempts, and eventually drop back to 2133 (I've just discovered that '5' is configurable in advanced bios settings, labeled something like 'memory attempts'

    Have not tried any manually tuning of memory yet. Maybe I'll try 16-16-16-36

    EDIT:

    Now running at 3200, with A-XMP Profile 2, and boosting NB Voltage to 1.1

    Unrelated to memory speed, I'm having issues with the fan curve settings in the BIOS. The fan curves seem to get reset after rebooting, or even simply coming out of sleep mode. Haven't narrowed down the issue. But similar issues have been posted in the MSI forum.
     
    Last edited: May 6, 2017
  21. tangoseal

    tangoseal [H]ardness Supreme

    Messages:
    4,624
    Joined:
    Dec 18, 2010
  22. Nobu

    Nobu Gawd

    Messages:
    984
    Joined:
    Jun 7, 2007
    Update: Now memtest stable at 2933CL14-15-15, 100 bclk, DRAM 1.28V. Must loosen timings + increase bclk to go higher (then bus is unstable), tighter timings fail memtest.
    Interestingly, not samsung--it's Hynix, 1Rank:
    snap582017114200PM.png
     
    Last edited: May 8, 2017
  23. magictoaster

    magictoaster Limp Gawd

    Messages:
    393
    Joined:
    Apr 6, 2005
    With the new BIOS (9945) posted by elmor over at overclock.net :

    ASUS ROG Crosshair VI Hero

    4x 8GB G.SKILL Flare X Series F4-3200C14D-16GFX

    3200MHz!
    16-16-16-39 (1T) - > 14-14-14-34 (1T) 1.35V

    Tomorrow I'll try with more aggressive timings.
     
    Last edited: May 20, 2017 at 7:05 AM
  24. noko

    noko 2[H]4U

    Messages:
    2,385
    Joined:
    Apr 14, 2010
    That is great news for 4 sticks! Will have to try hopefully this morning with 9943 version for 2 sticks. Folks are saying it has a strap for DDR 4000 as well.

    I now run the Corsair ram on the BioStar at rated speed of 3200 16-18-18-36 without issue. Almost as if the ram needed to be broken in a little. Stable as a rock.
     
  25. Nobu

    Nobu Gawd

    Messages:
    984
    Joined:
    Jun 7, 2007
    My g.skill hynix m-die modules posted today at 3272c24 after much training...once. Ram voltage set to 1.4, bios showed 1.412v. Unfortunately didn't get to test if it was stable, as after setting voltage manually to 1.42v it didn't pass training again. :(
    Edit2: Got it to go again! This time exited without saving, running memtest now, so far so gooderr in test 5/6, could be due to bclk...
    Posted at 3204c18, 1.4v, soc @ +.3v (bios displays 1.272), bclk 114 without training, but only the first time. Think LLC will help this ram a lot.
    Started failing at random inversions. Set to cl20 and try again...
     
    Last edited: May 21, 2017 at 6:50 PM
  26. ronan_zj

    ronan_zj n00bie

    Messages:
    28
    Joined:
    Feb 27, 2009
    Gigabyte GA-AX370-GAMING K7
    BIOS: F4A
    Extreme memory Profile: enabled, 3200MHz
    ( 1.35v, G.Skill FlareX)
     
  27. Centauri

    Centauri [H]ard|Gawd

    Messages:
    1,787
    Joined:
    Mar 1, 2003
    Ryzen 1700 @ 3.8GHz
    MSI B350M Mortar
    BIOS 1.3
    3200MHz @ 14/14/14/34 with 1.35v
    G.Skill Trident Z (F4-3600C16D-16GTZ)
     
  28. RaptormanUSMC

    RaptormanUSMC n00bie

    Messages:
    17
    Joined:
    Mar 7, 2017
    Currently running Corsair Dominator CMD16GX4M2B3600C18 18-19-19-19-39 1T @ 3600MHz on Asus C6H. Cannot get beyond 3600MHz no matter what I try though. Seems to be a hard wall for me right now.
     
  29. magictoaster

    magictoaster Limp Gawd

    Messages:
    393
    Joined:
    Apr 6, 2005
    I'm now stable at 3466 @ 16-15-15-15-34 (1T). That's good enough for now!
     
    noko likes this.
  30. JimmiG

    JimmiG 2[H]4U

    Messages:
    2,406
    Joined:
    Apr 3, 2008
    With the new 0612 BIOS for the Prime X370 I can now run my LPX 3600 at 3200 MHz @ C14 (previously C16) without cold boot problems.
    It's still AGESA 1.0.0.4. When the new AGESA comes out for this board, I'll try to run the RAM at its rated 3600 MHz.
     
    Pieter3dnow likes this.