2 Replies Latest reply on May 25, 2015 2:14 AM by gentrycl

    r9 290x higher heat uses lower voltage to stay stable

    gentrycl

      using msi after burner and furmark

       

      i added 100 mv to the core voltage kept the 1000mhz core clock and power play on

       

      ran furmark the card heated up to 94c then started throttling down to 300 mhz until the fan spun up and started becoming stable at about 650-700 mhz

       

      stopped furmark lowed the voltage started furmark again and the clock would become stable at a higher clock each time i repeated the process up until i got to -25 or so mv and it would be come unstable after that "when starting it from and idle"

       

      that me noticed once the card was heated up that i could lower the voltage -100 mv as far as msi

      afterburner would allow and it would maintain 1000 mhz stable

       

      but would artifact with power play on due to the voltage needing to be a little more constant

       

      turned power play off

       

      got up to 94c and i was able to do what i said above and lower the voltage -100 mv and it held the 1000 mhz stable at 93c

       

      it did throttle once until the fan speed picked up due to the fan curve being reset due to applying the lower voltage in msi afterburner

       

      once i stopped furmark the card started to cool down it became unstable due to the -100 mv added to the 2d mode and black screened

       

      one thing i forgot to mention i found this out due to my gpu black screening at -30 or so mv when it went into 3d mode and black screened unless the core was heated up

       

      im going to try to heat the card up again and apply a higher clock with the same voltage

      then if i need to adjust the voltage due to black screens ill post back my findings

       

      i just ran heaven benchmark loop when i got to 90c i lowered it 60 mv was stable went to -80 mv

      scene change usage dropped causing a temp drop and got a black screen

       

      i did not change the core clock due to having to raise the voltage while it was cold which would probably result in a black screen i will try it later

       

      i would change it when i get over 90c but i wanna try to make sure its not a bug in amd's temp reading code

       

      i will also note that i was getting screen artifacts when i lowered the voltage until the temp started raising again

       

      heres a quick pic of heaven in a window being ran i could of lowered the voltage more but max the loop could heat the core up to was 92c it was stable at 1000 mhz when the voltage was lower then in the pic

      it was at 1.09 v when it was at 1000 mhz and throttled down when the voltage increased as seen in the pic

      http://imgur.com/w7r1B8Y

       

      i would like to find out if im messing with one of amd's powerplay formulas or if i can adjust the voltage as the card heats up

        • Re: r9 290x higher heat uses lower voltage to stay stable

          This doesn't look software developer related, I'm thinking it probably belongs here: The specified item was not found.

           

          Let me know, I'm 99% sure this should move to the Radeon forum, which is devoted to hardware, overclocking, game compatibility, etc. In fact, after we get a bit more comfortable I'd probably just do that and send you a message. But we're getting used to having everyone in the same community.

           

          The good news is - before this, I'd tell you "YOu have to start all over again someplace else." This time, I just move it, it appears instantly, and you're good. Thanks for your patience.

          • Re: r9 290x higher heat uses lower voltage to stay stable
            gentrycl

            i wanted to clean this post up and add some more info

             

            well i noticed that as my gpu "a bios modded xfx r9 290 core edition modded to an x" as it heated up that i could apply a lower voltage and keep the same clock speed while still keeping it stable.

             

            using msi after burner and furmark

             

            i added 100 mv to the core voltage kept the  1000mhz core clock and power play on

             

            ran furmark the card heated up to 94c then started throttling down to 300 mhz until the fan spun up and started becoming stable at about 650-700 mhz

             

            stopped furmark lowed the voltage started furmark again and the clock would become stable at a higher clock each time i repeated the process up until i got to -25 or so mv and it would be come unstable after that "when starting it from and idle"

             

            that me noticed once the card was heated up that i could lower the voltage -100 mv as far as msi

            afterburner would allow and it would maintain 1000 mhz stable

             

            but would artifact with power play on due to the voltage needing to be a little more constant

             

            turned power play off

             

            got up to 94c and i was able to do what i said above and lower the voltage -100 mv and it held the 1000 mhz stable at 93c

             

            it did throttle once until the fan speed picked up due to the fan curve being reset due to applying the lower voltage in msi afterburner

             

            once i stopped furmark the card started to cool down it became unstable due to the -100 mv added to the 2d mode and black screened

             

            one thing i forgot to mention i found this out due to my gpu black screening at -30 or so mv when it went into 3d mode and black screened unless the core was heated up

             

            im going to try to heat the card up again and apply a higher clock with the same voltage

            then if i need to adjust the voltage due to black screens ill post back my findings

             

            i just ran heaven benchmark loop when i got to 90c i lowered it 60 mv was stable went to -80 mv

            scene change usage dropped causing a temp drop and got a black screen

             

            i did not change the core clock due to having to raise the voltage while it was cold which would probably result in a black screen i will try it later

             

            i would change it when i get over 90c but i wanna try to make sure its not a bug in amd's temp reading code

             

            i will also note that i was getting screen artifacts when i lowered the voltage until the temp started raising again

             

            heres a quick pic of heaven in a window being ran i could of lowered the voltage more but max the loop could heat the core up to was 92c it was stable at 1000 mhz when the voltage was lower then in the pic

            it was at 1.09 v when it was at 1000 mhz and throttled down when the voltage increased as seen in the pic

            http://imgur.com/w7r1B8Y

             

            i would like to find out if im messing with one of amd's powerplay formulas or if i can adjust the voltage as the card heats up

             

            another note is that the gpu held the clock mostly at 1000 mhz due to the face it couldnt reach 94c its hard throttle temperature

             

            this time increased the clock while the gpu was heated up at stock volts

            http://imgur.com/3Tbnnu4

            started lowering the power in increments on -6 mv and got a black screen at -12 due to the temp dropping to 93c

            ill post more later

            i will also note that that clock was unobtainable under full load at the +65-75 mv it required to start cold

            *note to self show gpu clock graph next time not gpu core temp*

             

            just posted on a fellows solution for black screens using an r9 290x

             

            https://community.amd.com/message/2648737#2648737

             

             

            ill tell you this i was having the same problem think of it like this i have a long road "gpu" a set speed limit "temperature" and a acceleration limit "voltage/clock ratio tempcreation" when the chip is a cool temperature it requires a stronger voltage as the chip cools down the core requires more voltage depending on temperature to get to said "clock". the problem is that amd doesnt want to over volt their cards so its stock at 1.1 volts "average voltage the gpu will operate at under a varying temperature spectrum" because the video card sits at a low temperture while it is idle it requires a higher voltage at idle the it does once it heats up and once it hits 3d mode the voltage is too low for the gpu signal to reach the memory/memory controller on the correct period thus a black screen. r9 290x higher heat uses lower voltage to stay stable  that is a link to what ive been experimenting ive found out why the cards crash i just need to inform amd so they can fix it =D

             

            this more then likely will fix the problem

             

            kind of what i am discribing in my above post should fix black screens when you let the card heat up first/causing higher voltage at a lower heat in comparison ie having my card run at 1100 mhz @ 1.1020 volts once it heated up to 94c or needing 1.2v for it to be stable to clock up to 1100 mhz from 55c

             

            i tried for a higher clock but it would crash because the clock wouldnt go straight to the specified number i set in msi afterburner it would idle/throttle down before creating less heat causing the the commands to be sent on wrong times and had to crash.

             

            ill update more in a few.

             

            i made a gpu sensor log of what happened as i adjusted the clocks manually i would post some of lower temp failures might do that later

             

            the sensor log says edited because i cropped the stuff before stress and after it crashed