16 Replies Latest reply on Apr 19, 2018 2:31 AM by benman2785

    Biostar X370GTN mITX - 1.0.0.6 A failed flying experiment.

    xobeloot

      Many of you have likely seen my previous thread for my mITX *sweet spot*.  Let me start this off by stating that everything in this thread will be unique to my system.  Your results may vary.

       

      First off, some basic system specs:

      Header 1Header 2
      CPURyzen 7 1700X
      RAMCorsair Vengeance LED 2 x  8gb w/ 3200MHz XMP
      MotherboardBiostar Racing X370GTN (mITX)
      CoolingCPU/GPU custom loop w/ dual 240mm radiators
      Power supplyEVGA Supernova G2 750w Gold
      GPUGTX 1080 Hydrocopper
      BIOSX37AK623 (Microcode release 1.0.0.6)

       

      I am leaving out drives and such as they don't play much (if any role in this).

       

      It took me roughly 24 hours of trial and error to reach this point.  I feel that less became too much less and more did not provide enough to be worth it.

       

      I now introduce you to the *RACING BIOS*  (Please hold back your excitement!)

      main 1.jpg

       

      First off, lets take a look at the easiest part of this adventure.  The fan control.  It leaves a bit to be desired (like specific RPM at specific temp ranges), but it gets the job done.

       

      Go to the advanced tab and select Smart Fan Control

      fan control 1.jpg

       

      Next, Select 4-pin.  I have all 5 of my fans as well as my pump running on the CPU 4-pin header

      fan control 2.jpg

       

      Then highlight FAN CALIBRATE and hit ENTER

      fan control 3.jpg

      After the calibration does it's thing, hit ENTER one more time to close it.

       

      Last (and about the least important) is the fan profile.  You get to choose Quiet, Aggressive or Manual.

      To be perfectly honest here, I am pretty sure they are all doing the exact same thing.

      fan control 4.jpg

       

      OK!  Fans are done.  Lets get that memory running. Go to the ONE tab and select Memory Clock Mode.

      Once there, select your XMP profile and leave Memory Frequency at Auto

      mem 1.jpg

       

      I still need to get my exact DRAM timings for this next step, but I just entered in the basic latency numbers.

      Go to DRAM Timing Configuration and plug in as much as you know.

      *You'll notice that both columns are the same. This is because I had previously entered the top 5 ratings and saved them*

      mem 3.jpg

       

      Lastly, you want to be sure your memory is getting the appropriate amount of power that it is rated for.

      Go back to the main ONE tab, highlight DDR Memory Voltage and add +0.144v as the offset.

      This will put the memory voltage at 1.355v

      mem 4.jpg

       

      Well, Xobeloot...  Why is your memory showing as 3232MHz?

      Great question!

       

      Since Biostar did not incorporate any *underclocked* core frequencies to allow one to set the freq and the multiplier

      without surpassing the memory rated speed and causing CMOS to puke violently, we are forced to use AUTO

      for the CPU frequency *GASP*

       

      Next, I set my CPU Ratio to 38.75 (3875MHz if it ran at 100MHz Freq which it does not)

      oc 1.jpg

       

      Voltage settings on this board are terrible at best.  There is no way to dial in what voltage you want to run at.

      You select an offset and then have to run multiple tests to see what voltage the core is actually running at.

       

      Finding a balance of core speed to core voltage was also a rather painful task.

      After many more episodes of CMOS tossing it's cookies on my shoes, I landed at +0.060v as the offset.

      oc 2.jpg

       

      Here is the Biostar Racing software now showing the CPU at idle.  Note the voltage of 0.935v

      CPU Snapshot no load.JPG

       

      And here is the CPU under load with the 0.060v offset

      CPU Snapshot.JPG

       

      Please don't ask me to try to make sense of how these numbers add up as I have NO CLUE!

      All I know is that this is what worked 100% stable no matter what I threw at it.

       

       

      But Xobe, you still haven't explained why your memory is reading at 3232MHz...

       

      Sorry, I must have seen a squirrel.  Back to that.

       

      Since we had to leave the CPU Frequency at AUTO to attain the desired clock and memory speed,

      we have inherently found our own Achilles heel.

       

      As the motherboard reads the CPU and system load, it adjusts the CPU frequency automatically.

      It seems to fluctuate between 99.3MHz and 100.8MHz.  Somewhere right in the middle there

      is where the system would be at homeostasis, but we can't dial in an exact frequency.

       

      So, in essence, what happens here is that when the CPU Freq flexes upward to the 100.8MHz range,

      It is also bumping up both the CPU Core and the memory.

       

      Some basic calculations to show where the CPU is actually working at:

      CPU Freq
      Ratio/BCLKActual Output
      99.3MHz38.753,847.88MHz
      99.8MHz38.753,867.25MHz
      100.3MHz38.753,886.63MHz
      100.8MHz38.753906.00MHz

       

      I have not figured out the exact range that the memory is operating under, but I assume it is 3200-3232+ under load.

       

      I played with setting, after setting, after setting until I reached a +0.240v offset on CPU_SOC Voltage

      **EDIT** I wound up setting this back to AUTO after my CMOS started puking again

      oc 3.jpg

       

      After that, all was gravy!

       

      Lets see the results!

       

      Temperatures through a grueling CPU, FPU, Cache and RAM barrage of 30+ min.

      (it is tough to read, but it starts and stays in the 66-68c range the whole time)

      If I only run CPU/GPU w/o the extra stress options, it never leaves the mid-high 40's.

      OverclockTemps30m.JPG

       

      Here is a quick screen of my GPU clocks for reference on the following numbers:

      GPU Clocks.JPG

       

      Time Spy:

      Timespy1.JPG

      Timespy2.JPG

       

      Fire Strike Ultra:

      FSultra.JPG

       

      And lastly, the reason that I feel that I have reached the biggest payout at the lowest clocks:

      Sweet Spot New.JPG

       

      Is this the most EXTREME! Ryzen overclock out there?  Absolutely not.

      Is it a fantastic Clock:Voltage:Memory:Performance figure? Absolutely!

       

      Like I opened with: Your mileage will most likely vary, however, if this saves somebody

      with the same motherboard 24hr of sheer agony, then it was a success.

       

      ***DISCLAIMER***

      Neither AMD nor Xobeloot are responsible for damages to your components.

      This write-up is to document my own success w/ overclocking this board/CPU.

      It is always best to research and learn all you can before overclocking.

       

      Thanks for taking the time to read this!  It was quite the task to create!

       

      -Xobeloot

      me.jpg

       

      Message was edited by: Kevin Toole Reason: New system instability and current fix noted in RED

        • Re: Biostar X370GTN mITX - 1.0.0.6 overclocking adventure (pics)
          whiskey-foxtrot

          Those are some nice physics scores! Curious to see what your AIDA64 memory benchmark reads; have you looked into the Ryzen Master app, as it may show a more accurate reading for the  voltages. Not sure if this board has access points, but you can always plug in a DMM to read those voltages.

          1 of 1 people found this helpful
          • Re: Biostar X370GTN mITX - 1.0.0.6 overclocking adventure (pics)
            ajlueke

            Very interesting data here.  I am curious as to why your board jumps to a lower voltage with a fixed multiplier enabled?  Typically that is not something you'd expect outside of p-state overclocking.  And to elaborate for those who aren't aware:  As the system reads the load on the board it will jump to different p-states which are predefined and usually alterable in X370 boards.  In my overclock for example, I also left the base clock on auto as well as the multiplier for the CPU.  In the p-state control, there are numerous different p-states each with their own multipliers and voltages that the system can jump to depending on load.  I found the highest level p-states ( always p-state 0, which is a 36 multiplier and voltage of 1.35V on the 1800X) and changed that to a multiplier of 40.  I also added a bit of voltage offset to give the 4.0 GHz overclock a bit more juice (CPU offsets are applied to all p-states, so I am running idle settings with a bit higher voltage than necessary).

             

             

            What is interesting is since you have set the multiplier to a fixed 38.75, that should override the p-state control and force the system to run constantly in p-state 0 (highest frequency and voltage at all times).  The fact that you see a lower voltage at idle indicates that some sort of power saving is still going on.  Do you have the C6 state enabled?  It is possible there are actual components shutting off when the C state is raised and the system is going idle.

             

             

            As for the voltages, there is definitely something going on with the readout in the UEFI.  With a CPU multiplier of 38.75 it reads 3875 for the CPU (implying a base clock of 100) and 3232 for the RAM (implying a base clock of 101).  The baseclock can't very well be different be different between RAM and CPU, and indeed in your timespy run you see RAM of 3192 MHz and 3865 MHz on the CPU.  Both of these values would correspond to a baseclock of 99.75, which is likely correct.

             

            When you apply the DDR4 timings that whiskey-foxtrot listed manually, have to tried also raising the SOC voltage back up to around 1.1 beforehand?  In my experience that added SOC voltage goes a long way in stabilizing RAM overclocks.

              • Re: Biostar X370GTN mITX - 1.0.0.6 overclocking adventure (pics)
                xobeloot

                Interesting to see your response now.  I have just spent the past 6ish hours trying to get my system back to where it was.  My CMOS puked again and I have yet to get it to hold anything stable.  Yes, ANYTHING!  I have literally had this thing puke my CMOS at stock settings (3400 cpu/2133 mem).

                 

                It has been a fun trip, Biostar,  but I will be purchasing a new motherboard ASAP. 

                  • Re: Biostar X370GTN mITX - 1.0.0.6 overclocking adventure (pics)
                    ajlueke

                    Uffda,  that is never fun.  Not sure if that speaks to the quality of Biostar, but there was definitely something odd going on with that board.  Perhaps the power module was faulty, as it is still unusual that you saw voltage and base clock variation with a fixed multiplier.

                     

                    You could always try an ASRock > Fatal1ty X370 Gaming-ITX/ac.  I haven't tried it personally, but I've had great success with the ATX variant.  Also, make sure to watch your VRM temps with the B350 based board.  There is quite a bit of anecdotal evidence suggesting that overclocking the octocore Ryzen variants on the B350 chipset can lead to some high temps.  The extra phases in the X370 designs definitely will help with that.

                • Re: Biostar X370GTN mITX - 1.0.0.6 A failed flying experiment.
                  xobeloot

                  I just ordered the GIGABYTE GA-AB350N and will be doing a RMA on this biostar board as soon as I have swapped out all of my bits.

                  • Re: Biostar X370GTN mITX - 1.0.0.6 A failed flying experiment.
                    chiko

                    cant you help me with this: i selected ram speed 3400mhz and after that i reboot and got black screen no bios loading nothing, is that sign that motherboard dead? tried to remove ram, nothing...