1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

980X Voltage changes

Discussion in 'Overclocking & Cooling' started by forumsviewer, May 19, 2010.

  1. forumsviewer New Member

    Joined:
    Feb 8, 2010
    Messages:
    64 (0.02/day)
    Thanks Received:
    1
    I have an Asrock X58 Deluxe motherboard with an Intel six core 980x processor with 24gb Patriot PC3-12800 1600mhz memory. I am looking for stability, not overclocking.

    With the BIOS settings all on AUTO the CPUz says core was 1.168, core @ 3.466, multiplyer 26, bus speed 133.3 and QPI 3200. I ran Prime95 and after about 4 hours the computer restarted and the minidump said it was "Bugcheck 124" which I believe is associated with CPU voltage.

    so I bumped up the voltage in the BIOS for the processor to 1.26250. Back in Windows 7 in CPUZ it says the same multtiplyer, core speed, bus speed, QPI with the difference showing the voltage as 1.200 (rather than 1.168 as before with AUTO setting). I run Prime95 and keep CPUZ open and the voltage went from 1.200 to 1.152. I turn off Prime95 and it went frolm the 1.152 back up to 1.200. Hrrmm, so I turn on Prime95 again and sure enough, it went from 1.200 to 1.152.

    I'm a bit confused. Shouldnt the voltage stay the same, if not go higher, during 100% Prime95 load?

    Can someone chime in on if this is correct and/or what settings I should have for stability?
     
  2. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    16,086 (3.98/day)
    Thanks Received:
    10,193
    Location:
    Beaumont, Alberta
    Perfectly normal.This is called vdroop. as the cpu consumes current, the voltage drops, so that when the cpu goes idle, the cpu doesn't get flooded with current it cannot use. Using the voltage difference between load and idle, you can figure out how much current the cpu is actually using.

    When you set voltage in bios, you are setting the MAX VOLTAGE, not the minimum. Things like "loadline calibration" reverse this...you set the minimum when using LLC.
     
    10 Year Member at TPU
  3. Bo$$

    Bo$$ Lab Extraordinaire

    Joined:
    May 7, 2009
    Messages:
    5,615 (1.93/day)
    Thanks Received:
    965
    Location:
    London, UK

    you should be stable now, but that Vdroop is supposed to happen :rockout:

    what are your temps like?
     
  4. forumsviewer New Member

    Joined:
    Feb 8, 2010
    Messages:
    64 (0.02/day)
    Thanks Received:
    1
    It has been running Prime 95 for a bit over 2 hours now and 980X temperature highest is 64* C (147* F).

    It usually does its restart within 2 to 4 hours of Prime95.

    Below is a minidump from BEFORE setting voltages manually within BIOS where it restarted. (AUTO showed voltage as 1.168 at idle). Prior I received a several PAGE_FAULT_IN_NONPAGED_AREA (50) but have since replaced all of the memory. (indeed, when I tested each of the modules of memory individually one of them wouldn't post the system, so I replaced all of the memory).

     
  5. Bo$$

    Bo$$ Lab Extraordinaire

    Joined:
    May 7, 2009
    Messages:
    5,615 (1.93/day)
    Thanks Received:
    965
    Location:
    London, UK

    What make is your memory, voltages, specs, rated speed, etc
     
  6. forumsviewer New Member

    Joined:
    Feb 8, 2010
    Messages:
    64 (0.02/day)
    Thanks Received:
    1
    memory is patriot PGS312G1600ELK (using 24GB in 4GB modules). Specs are PC3-12800 (1600MHz) 9-9-9-24 @ 1.65v. CPUZ says it is running at 9-9-9-24 2T 800mhz 2:12. Remember, this memory I replaced all the modules because my mini dump from before running prime95 said PAGE_FAULT_IN_NONPAGED_AREA (50) and later found out by individually testing each module that the system wouldnt post with one of the 6 modules only in. So to not take chances, I replaced all of the modules.
     
  7. Bo$$

    Bo$$ Lab Extraordinaire

    Joined:
    May 7, 2009
    Messages:
    5,615 (1.93/day)
    Thanks Received:
    965
    Location:
    London, UK
    so is your BSOD problem persistant? or have you resovled it?
     
  8. forumsviewer New Member

    Joined:
    Feb 8, 2010
    Messages:
    64 (0.02/day)
    Thanks Received:
    1
    At first I was getting BSOD of PAGE_FAULT_IN_NONPAGED_AREA (50). It would happen within 4 hours of Prime95 running. I then tested each memory module and found one of the six to not post the system if in by itself. So I replaced all of the memory.

    Started running Prime95 again and within 4 hours it BSOD and the minidump is in this thread with a WHEA_UNCORRECTABLE_ERROR (124) issue. Doing research I believe it was the voltage of the i7 980x processor.

    I adjusted the CPU i7 980x voltage higher so that at idle it is at 1.200 while at load it is 1.152. It has been running Prime95 for 5 hours. Resolved? I don't know, I don't believe it has lasted this long without BSOD. If there are any other suggestions or tests that I can run that are fairly quick than I am all ears.
     
  9. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    16,086 (3.98/day)
    Thanks Received:
    10,193
    Location:
    Beaumont, Alberta
    you are running 6 sticks? maybe you need more IMC volts.
     
    10 Year Member at TPU
  10. surfingerman

    surfingerman New Member

    Joined:
    Jan 1, 2010
    Messages:
    209 (0.08/day)
    Thanks Received:
    59
    yeah wow with that much memory, your problem IMO is more likely to lie in memory area.. try lowering your mem speed and manually set timings for the stock value and put the mem volts at 1.65v , then test your CPU after you have CPU rock solid, then play around with increasing mem speeds again

    also set your QPI volts high and leave them so until your done, afterwords when your done try lowering them
     
  11. forumsviewer New Member

    Joined:
    Feb 8, 2010
    Messages:
    64 (0.02/day)
    Thanks Received:
    1
    if not mistaken, the memory is setup at the stock part number values.
     
  12. surfingerman

    surfingerman New Member

    Joined:
    Jan 1, 2010
    Messages:
    209 (0.08/day)
    Thanks Received:
    59
    did you set QPI volts high? how many volts are you running through your DRAM
     
  13. forumsviewer New Member

    Joined:
    Feb 8, 2010
    Messages:
    64 (0.02/day)
    Thanks Received:
    1
    I havent touched anything in the BIOS concerning QPI.
     
  14. surfingerman

    surfingerman New Member

    Joined:
    Jan 1, 2010
    Messages:
    209 (0.08/day)
    Thanks Received:
    59
    you should probably set QPI 1.35v it could mess you up also set mem volts to 1.65v if im not mistaken, test for CPU stability then mem.. and lower your mem volts if you choose. i would leave it at 1.65v personally
     
  15. forumsviewer New Member

    Joined:
    Feb 8, 2010
    Messages:
    64 (0.02/day)
    Thanks Received:
    1

    The specs on the memory say 1.65v for the memory. I haven't adjustd voltage for QPI or memory in the BIOS (i kept them at auto).

    Prime95 has ran for the past 6 hours. It would normally BSOD by now. I will stop Prime95 and start testing with OCCT LinPack 1hr test to see if I can get it to crash. If not, I guess my best bet is to just stick with the settings that it is at right now if I can't get it to crash with Prime95 at 6 hours and OCCT LinPack 1hr?
     
  16. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    16,086 (3.98/day)
    Thanks Received:
    10,193
    Location:
    Beaumont, Alberta
    prime 95 takes about 9.5 hours for a full single blend test @ 3.6ghz...you are using Blend, right?
     
    10 Year Member at TPU
  17. forumsviewer New Member

    Joined:
    Feb 8, 2010
    Messages:
    64 (0.02/day)
    Thanks Received:
    1
    Have been using in-place large.
     
  18. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    16,086 (3.98/day)
    Thanks Received:
    10,193
    Location:
    Beaumont, Alberta
    might be part of the problem(in place large FFTs isn't going to fully test the system). Use blend, it does both small and large FFTs, and seems ot put more load on i7 and Phenom chips. Also takes longer to run, but @ 9.5 hours, that's a good thing to do when ya go to bed. You can surf or whatever when the test runs..I actualyl recommend surfing and running audio or something, to closer simulate real usage....
     
    10 Year Member at TPU
  19. forumsviewer New Member

    Joined:
    Feb 8, 2010
    Messages:
    64 (0.02/day)
    Thanks Received:
    1
    hrrmm, well i've been using inplace large ffts and getting the issue to arise within 4 hours. It ran for 6 hours with the increased CPU voltage and no BSOD. I'm trying to keep variables the same and only change one (such as voltage)
     
  20. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    16,086 (3.98/day)
    Thanks Received:
    10,193
    Location:
    Beaumont, Alberta
    I completely understand. When you are satisfied that Large FFT is OK, be sure to try the blend test as well. I've only come to this conclusion after running prime96 for literally 1000's of hours...

    Also, be sure to run the 64-bit version of prime if you are in a 64-bit OS!!!
     
    10 Year Member at TPU
  21. forumsviewer New Member

    Joined:
    Feb 8, 2010
    Messages:
    64 (0.02/day)
    Thanks Received:
    1
    indeed am running 64 bit version. thanks for the heads up.
     
  22. mastrdrver

    mastrdrver

    Joined:
    Feb 24, 2009
    Messages:
    3,443 (1.15/day)
    Thanks Received:
    702
    Intel i7s are not rated for anything over 1066mhz memory for 3 dimms or more. In other words, Intel guarantees that 3 dimms will run 1066mhz at 1.5v with JEDEC timings. Anything else is ymmv.

    Also, unless you buy them as a kit, memory modules are only guaranteed to run at their speeds and timings depending on how you buy them. So while you bought 1600mhz memory, it will only run those speeds as a 3 dimm kit, 2 dimm kit, or 1 dimm depending on how you bought them.

    Notice how many 6x2 12GB kits actually are rated for anything over 1333mhz? There is a reason for it: its hard.
     

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)