The AMD Ryzen All In One Tread /Overclocking/Memory Speeds & Timings/Tweaking/Cooling

Discussion in 'Processors and motherboards AMD' started by chispy, Feb 22, 2017.

Thread Status:
Not open for further replies.
  1. thatguy91

    thatguy91 Ancient Guru

    Messages:
    6,643
    Likes Received:
    99
    GPU:
    XFX RX 480 RS 4 GB
    That's a good point! It's possible that they'll add back the other options when the next update is released. Mine came with that bios as well, so there would be a lot of boards out there without the ability to update if they had only a bios option from now on.

    It's quite possibly they're making a better Windows update method.
     
  2. Razorblade

    Razorblade New Member

    Messages:
    9
    Likes Received:
    1
    GPU:
    R9 290x 4 GB
    New Bios for Asus Rog crosshair VI Hero, can't post links yet. :):banana:
     
  3. Odellot

    Odellot Master Guru

    Messages:
    708
    Likes Received:
    69
    GPU:
    TUF RTX 3090 FO48U
    Sold The 1700x and MSI B350 Tomahawk on our Family PC and Bought 1600 and MSI X370 SLI Plus...

    Is this a decent result?

    [​IMG]
     
  4. Kanuki

    Kanuki Member

    Messages:
    47
    Likes Received:
    4
    GPU:
    RTX 3080 10GB
    I had to use the Device Manager to search for most of the PCI Drivers
    from the Drivers folder. The installer did not install the PCI Drivers
    automatically for me.
     

  5. thatguy91

    thatguy91 Ancient Guru

    Messages:
    6,643
    Likes Received:
    99
    GPU:
    XFX RX 480 RS 4 GB
    youtu.be/vZgpHTaQ10k?t=3m34s

    It's pretty basic stuff, but it is intersting some of the stuff you hear directly from AMD such as the SOC voltage and what you can actually set the DDR4 voltage to. I think the 1.5 V would be pushing it to the extreme! The SOC at 1.1 V is probably high as well, but it is the suggestion going around so it's at least somewhat comforting that it's not a non-suggested solution :) if that makes sense! It's interesting the stock voltage for the SOC is so low. Maybe they'll increase it to say, 1 V - 1.1 V at some stage automatically when using higher RAM speeds.

    The most useful thing was probably the mention that overclocking the memory and optimising that should be done first then overclock for best performance and results. Simple, but it made me revisit my settings and I can easily get 3200 at CAS 14 stable. I am encoding something in the background, watched some videos and gamed (including encoding in the background at idle priority, which you can easily do), no issues.

    Anyways, basically I recommend doing just that. Memory first, then CPU, SOC a tad over 1 V and maybe a little more DDR4 voltage like 1.36 V. I might not even need the 1.36 V, but it's there for potential added stability since it seems 'recommended' as such.
     
    Last edited: Apr 28, 2017
  6. Razorblade

    Razorblade New Member

    Messages:
    9
    Likes Received:
    1
    GPU:
    R9 290x 4 GB
    Any one tried the new 1106 bios for Asus Rog Crosshair VI Hero?

    Can't get my ram stable at 3200 MHz 14 14 14 34:cry:
     
  7. thatguy91

    thatguy91 Ancient Guru

    Messages:
    6,643
    Likes Received:
    99
    GPU:
    XFX RX 480 RS 4 GB
    Reset bios, overclock and tweak RAM first, then overclock the CPU.
     
  8. Hully

    Hully Member

    Messages:
    25
    Likes Received:
    1
    GPU:
    EVGA GTX970 SSC
    Just upgraded to 1600 from the tiring FX8320...been playing about with it for couple days and hit a nice sweet spot with 3.9Ghz with 1.25v. However i can't get Ram higher than 2400 atm


    [​IMG]
     
  9. Clouseau

    Clouseau Ancient Guru

    Messages:
    2,794
    Likes Received:
    470
    GPU:
    ZOTAC AMP RTX 3070
    NVMe (960 EVO) drive arrived yesterday. Swear the bios was written assuming one was on board. Got a stronger heartbeat at 3200. It is so close, just a matter of voltage. Want the May bios now.

    After watching that video, felt better about trying 1.4v on the ram. 1.5v though has to be nuts. Was also happy to hear the range for ProcODT and really not to exceed 80 Ohms. Still at "Auto" though. Funny thing, already had SOC at 1.10625. Right at the wall though with the Vcore @ 1.4125.
     
  10. Odellot

    Odellot Master Guru

    Messages:
    708
    Likes Received:
    69
    GPU:
    TUF RTX 3090 FO48U
    Woah...Not Bad...My 3.9ghz 1600 can keep up to my old 7700K @ 5.1ghz on TimeSpy..:D

    TimeSpy Comparison between my 3.9ghz 1600 vs 7700K @ 5.1ghz with both Ram @ 2666 and RX 480 @ 1400..:D

    [​IMG]
     

  11. AsiJu

    AsiJu Ancient Guru

    Messages:
    7,978
    Likes Received:
    2,771
    GPU:
    MSI 6800XT GamingX
    ^ nice! Also cool to have (at least one) direct comparison between Ryzen and Kaby Lake clock-for-clock, including memory, and with a more mainstream GPU.
     
  12. thatguy91

    thatguy91 Ancient Guru

    Messages:
    6,643
    Likes Received:
    99
    GPU:
    XFX RX 480 RS 4 GB
    So interestingly, 3200 for me is mostly stable. It appears stable in any testing, however general stability got a bit iffy at times, such as the desktop freezing but mouse still moving etc, among other things. I was playing Mafia III for about 3 hours and it dropped back to the desktop, but that could have been related to a bug in the game or using reshade.me etc.

    Anyways, I decided to go back to 2933 and it seems perfect now. I'll probably try higher again once later AGESA are released. For instance, I'll try again once the May update is released which is meant to greatly improve compatibility with faster RAM speeds.

    I am now running at 2933, with 14-14-14-14-32 timings. The RAM I have is DDR4-3733 17-17-17-17-37, which I bought a little while ago for a really good price, and before the issues with memory speeds became really known. Even second hand I could sell these for a profit if it came to that!

    Why 14-14-14-14-32? The timings are proportional to the RAM speed, so 14 cycles at 2933 is much longer than 14 cycles at 3733.

    I selected the speed I wanted, 2933, divided it by 3733, which comes to .7857 (rounded). Of course, leaving it in the calculator without rounding. Mutliplying this by the timings at 3733 will give you the equivalent timings in terms of time spent in that action at 2933. For the timings of 17 at 3933, it gives 13.56 at 2933. Choosing 13 for one of those timings could possibly work, but that would require more testing and chance of random issues. I therefore chose the next full figure, 14. Would be cool to have 13-13-13-13 at 2933 stable though :). At 3200 this would mean timings of 15 when using the next whole number, but you also have to consider this method of timings is mathematically correct, but the tech will behave differently. The 3733 RAM is (supposedly) just really good binned slower RAM, so 14-14-14-14 timings at 3200 wouldn't be unrealistic if later AGESA code does improve things 'significantly' as suggested. Before you ask, yes I did even try 16-16-16-16-40 to eliminate timings as an issue with the 3200 compatibility.

    Quite honestly, most people would have run it at 3200 and not been the wiser with it being a bit finnicky. It becomes more apparent in a 24/7 system like I run it. Also because of running 24/7 I have not pushed the SOC, RAM, or CPU voltage to the extent that others have to get it stable at higher speeds.

    Now, for that TRAS of 32. That is quite low when I look at other memory of around 3000 MHZ etc. The actual figure when I multiply it out is 29.07 (so 30), but decided to go 32 just in case. So even though I'm not running 3200, I am quite happy, at least for now and in comparison to other people, to be running 2933 at 14-14-14-14-32. It's actually surprising that TRAS is so high on the lower speed RAM with decent other timings. Using the old (whether accurate or not) tRAS = tCL + tRCD + 2 gives 30, which is the same as the speed equivalency rating would give. Funny that :). There is also another formula tRAS = CL + tRCD+tRP, which would make tRAS 42, which obviously isn't right. However, the direct calculation doesn't quite work for RAS since you already rounded off the other values to the next highest full number. Using 30 as a base, dividing the rounded timing by the multiplied timing, then multiplying that by 30 gives you 31.4, so the 'ideal' TRAS calculated by equivalency would be 32.

    Just for comparison price wise, 2x8 GB is over AUS$100 more for the AMD spec F4-3200C14D-16GFX than what I paid for mine. This is mainly due to the awesome deal I got, however even now 3733 RAM is cheaper than the AMD spec 3200 RAM.

    BTW, in the latest Asrock bioses you can finally set the CPU voltage as an offset instead of a fixed value. They just need to label the CPU and SOC offset as mV, and not V like they currently do!
     
    Last edited: Apr 30, 2017
  13. AsiJu

    AsiJu Ancient Guru

    Messages:
    7,978
    Likes Received:
    2,771
    GPU:
    MSI 6800XT GamingX
    ^ good that cpu offset voltage is available now! I updated to 2.30 and reapplied my overclock the old way, didn't realise you can set an offset now.

    3200 memory freq is a bit strange for me. If the system boots it seems to be fully stable as well in games and benchmarks.
    Getting the system to boot is finicky though. Plus rebooting seems to always lead to a boot loop and revert to 2133.

    I don't think our mobos/bioses have a setting for DRAM boot voltage? Gaming K4 and Killer SLI seem nearly identical specwise and have the exact same bios versions too.
     
  14. thatguy91

    thatguy91 Ancient Guru

    Messages:
    6,643
    Likes Received:
    99
    GPU:
    XFX RX 480 RS 4 GB
    Yes, I think there are some small changes between them. I think the AGESA code probably made things easier for them in providing multiplie bioses. There are a few little differences between the two.

    I'm still surprised 32 TRAS works fine, I'll probably know it back to 34 simply because that's probably about where the non-so overclockced version of this RAM would do, just to be on the safe side I guess.
     
  15. AsiJu

    AsiJu Ancient Guru

    Messages:
    7,978
    Likes Received:
    2,771
    GPU:
    MSI 6800XT GamingX
    ^ yep, there's no offset setting for Killer SLI yet, checked, still only fixed/manual (or Auto/default).

    The version numbers were different too, 2.20 for Gaming K4 and 2.30 for Killer SLI even if the description is the same.
     

  16. Clouseau

    Clouseau Ancient Guru

    Messages:
    2,794
    Likes Received:
    470
    GPU:
    ZOTAC AMP RTX 3070
    Head is stuck in the past. What is the advantage of using an offset instead of just inputting the voltage say as 1.3250? Does the offset allow the voltage to down clock on lesser loads?
     
  17. thatguy91

    thatguy91 Ancient Guru

    Messages:
    6,643
    Likes Received:
    99
    GPU:
    XFX RX 480 RS 4 GB
    In principle offset is just as it sounds, the base voltage plus the offset. This means that at what ever the voltage would have been, the offset is added to it, so yes :). I believe at higher overclocks it simply won't clock down currently, which means if the CPU speed is fixed the voltage is mostly fixed. Due to the SenseMI the voltage may vary at a given clock rate in normal circumstances. I think if they allowed downclocking even at overclocked speeds in later AGESA it would be beneficial, it would certainly be reflected in off-load temperatures and power use.

    Next time I reboot I might take a screenshot. I'm currently encoding, it's 4.15 am Monday and only up due to a noisy possum in the tree outside my window... lols.
    One of these. Random internet pic of a brushtail possum like outside my window. I had to reupload it to imgur as the original site was censored (word press)
    [​IMG]

    Basically if I did do a screenshot it won't be until tonight. If they release a new bios before then I probably won't bother, the only reason why I would do it is that I'm pretty sure it wasn't in the previous bios, and I don't even recall seeing it in the latest bios until I came across it playing around with the memory speed and timings.

    Two big differences between the Gaming K4 and Killer is the Gaming K4 does not come with wifi, but there is provision for it (including antenna places at the back). It also has ALC1220 audio and Purity Sound 4, the Killer only has ALC892. There other small differences like 4 included SATA data cables versus 2 for the Killer, and has Dr Debug LED.
     
    Last edited: Apr 30, 2017
  18. Elder III

    Elder III Ancient Guru

    Messages:
    3,731
    Likes Received:
    333
    GPU:
    6900 XT Nitro+ 16GB
    Your possum's are allot cuter then the fugly ones we have in North America...

    [​IMG]

    I have my Asrock X370 Fatal1ty Professional Gaming (damn big mouthful of a name) motherboard installed and working for a few days now. I have not had time to tinker very much yet, but the Bios is many many steps above and beyond what my Biostar X370GT7 had to offer. Preliminary results indicate that it will do 4.0 Ghz @ 1.35v just like the last motherboard, so it looks like it's lack of a need for higher voltage is due to the CPU and not the motherboard. Memory testing is still a work in progress at this point.... need more time to focus on it and see what the limitations are.
     
  19. __hollywood|meo

    __hollywood|meo Ancient Guru

    Messages:
    2,990
    Likes Received:
    139
    GPU:
    6700xt @2.7ghz
    opossums are definitely a lot uglier than possums :D

    thatguy, your calculations are good for eyeballing a proportional equivalent of stable timings at slower speeds, but remember that the latency scaling for timings/subtimings isnt always linear as clockspeed changes. experimentation with stability testing & bandwidth benches is the only way to pin down exactly wat your ICs are capable of.

    id also recommend raising your SOC (memory controller!) voltage if you want to hit 3000+ youre gonna need to get it over 1v. dont loosen tRAS if you dont need to! its the biggest performance hit next to higher CAS latency itself.
     
  20. -Tj-

    -Tj- Ancient Guru

    Messages:
    17,478
    Likes Received:
    2,132
    GPU:
    Zotac GTX980Ti OC
    The most unstable and highest bandwidth efficiency is in tertiary timings (3rd last),
    [​IMG]

    tRTR, is for read speeds e.g. aida64 read test
    tWTW is for copy and write, but mostly copy speed.

    Asus usually set them too high with negative bandwidth if at higher OC.
     
Thread Status:
Not open for further replies.

Share This Page