Solved

My 1st OC 8350 SUMOUT. Need advice.

I am currently trying to stabilise an overclocked 8350. I have run prime 95 for an hour (FTT small) with no issues but as soon as I run a memory test the system falls over with SUMOUT errors. I'm currently running voltages at 1.36 on the CPU and 1.25 on the CPU Northbridge. I have all the power saving facilities off and cool and quiet off.

Multiplier is at 21.5
FSB is at 200 (stock)
Cpu/NB and HT frequency is set at 2200. I use the D.O.C.P. ram profile to set ram timing as I could get the 1866 stable at stock. This was when I noticed the asus bios set cpu/NB and HT speeds to 2200 for my crucial ballistix tactical 1866 9-9-9-27 to run stable at stock clocks. It was the only way to get the ram stable with everything at stock. To try to overclock I have reduce ram frequency to 1600 but tightened timing to 8-8-8-24.

All my FTT small runs are stable for anything up to an hour. I get no rounding errors at all. The SUMOUT errors happen during blend and custom tests ultilising 75-90% of my ram.

I have tried bumping NB and NB HT voltages up 4 integers but this doesn't help. I am not checking if RAM voltage set to 1.525v will help but I am at a loss. I just cannot get the ram to stabilise (if it isn't he ram. I presume it is from the SUMOUT errors).

My system is:
M5A99X evo 1.0
FX 8350 4.0ghz
Crucial ballistix tactical 1866 1.5v
1060 gtx
Corsair 850 bronze power supply.

Temps are
Package 52c under load
Socket 59c
Idle package 11c!!
Idle Socket 35c

The error always comes at 58 minutes. No matter what i do?

1.525 test just fails at 58 minutes on same core. The failure varies though but is often core 8, 7, or 6.

I know this can go higher but I just cannot get the ram stable. Any suggestions? Should I wrap the Northbridge footage or the Northbridge HT voltage? Should I try increasing cpu voltage further (I have done this before but it didn't help). I am lost!

I am now trying a 2 integer bump in cpu volts to 1.381v to see if this stabilises the ram but I am not hopeful. Will edit post if this fixes my problems.

This isn't a link to google photos for all my BIOS setting and ram timings.

https://goo.gl/photos/DhoRUQSrtDRkFZpD9
28 answers Last reply Best Answer
More about 1st 8350 sumout advice
  1. Best answer
    Just another reason for not using Prime95 with AMD system, it's made for Intel and it's manner of work and measurements. Use AOD and or OCCT. First one is made for AMD only and second can be adjusted better. Prime95 should have been in a museum long time ago.
  2. AOD causes black screen with ANY adjustment.
  3. Don't use it for OC, that should be done in BIOS, use it for temperature and other readings but there's also benchmark and stability test better suited for AMD platform.
  4. Thanks mike.

    Should I run all tests at once to check stability or one at a time?
  5. I usually turn Logging on and run all tests, if it fails than one test at time to see what really failed.
  6. I also have OCCT which seems to work more like Prime but with test halt options so I can set stresses to end if temps are breached. This is really nice. Any idea why I would be getting sumout errors though when cpu stresses are stable after 4 hours? Seems to be ram related but I cannot work it out. My ram is even under clocked @ 1333mhz! It's 1866 but just won't hold that with a multi of 21.5... I am scratching my head for any voltage or setting that would stabilise the ram. I have manual applied stock settings and tightened them from 1866 9-9-9-27 to 1600 8-8-8-24 and still not stable. Voltages seem like that should be high enough for such a small OC. 1.38v on CPU. Should I raise cpu/fb voltages from 1.25v?

    I have no idea what to do...
  7. I adapted opposite strategy with OC and voltage, I first raise voltage as much as possible to test cooling limits, raise frequency thru multipliers and than lower the voltage to see where it could run lowest and most stable. Because of FX chips sensitive IMC, I use FSB just for fine tuning afterwards. Surprisingly highest frequency may not bring best overall performance, slightly lower than top OC achieved. I also almost exclusively use Kingston RAM and it loves higher voltage up to 1.65v. and seems more stable.
  8. Well OCCT worked a charm. Discovered I cannot run this chip @4.5ghz on air but am now doing an 8 hour stress on 4.4ghz at 1.75v (stock voltage) No error picked up on 30 minute burn small or large. Max temp was 57c compared to 61c on 4.5 at max (safe temp) voltage of 1.875. Didn't want to push further that that for everyday use. The fact that I am @1.75v and if this turns out to be stable I can do some bclk increases to get closer to 4.5. I have a feeling I may end up at a bclk of 202-203 or 4.48ghz.

    Now the question will be can I get may RAM up to 1866 or 1600 will hour errors. If not I will work on tightening timings to 8-8-8-24 @1600 or 7-7-7-21 @1333. Will have to wait and see. Frequency and HT are 2400. Don't really know why but was 2200 with HT 2600. Thought I would match them as I will be messing with BCLK and read it was better to have them matched. Will let you know how it goes.

    Many thanks Mike
  9. OCCT gives readings as you test so you can see what is possibly failing.
    But... 1.75v is too much for air cooler. AMD says it should not go over 1.5v (real voltage at core, that is "Vid0" in AOD) but I did go with 1.6+v with my 6350 without ill effects for at least couple of years and 24/7 of which some times it was cooking some video transformations whole night at full load.
    When playing with FSB/BCLK, make sure you put PCIe buss on manual and at 100MHz or your GPU and/or SATA may start making problems.
    It also doesn't hurt to put everything on manual even if you keep default values.
    Another thing while testing, turn off all power saving and turbo modes in BIOS or that may make your measurements inaccurate and may even interfere with OC. Once you get everything stable at your best OC, you can turn some of them back on.
  10. All I am willing to go for on air is 4.3ghz. I was able to get it stable at higher voltage at 4.4 but was hitting 65c.

    As stable now I am 21.5 multi and FSB 200 and ram is set to 1600 9-9-9-24. HT is at 2600 cpu/NB is 2200. Cpu or Vcore is 1.375v and cpu/Nb is 1.225v.

    All was stable after 18 runs on intel burn at 90% ram usage.

    One problem has occurred though and it has me worried. My GFX card was showing some glitches in Witcher 3 on ultra and am concerned I damaged my card after overclocking the FSB to 210 without setting pcie manually to 100. I thought auto would be fine.

    Anyway after the 17 runs of intel burn all my screen icons where moved to screen one like the gfx driver failed. This has happened twice and winding if it is sign of instability or NVIDIA drivers not liking cpu overclock software???
  11. I doubt PCIe frequency can damage GPU permanently, voltage maybe but frequency can make problems only temporarily. Try replacing drivers anyway.
  12. So here is the thing. I am playing witcher 3 on ultra and my temps are sitting at 31c... do you reckon I could probably push this thing a little bit more maybe managed to stabilise 4.5ghz? I was hitting 65-67c under load at 4.5 originally... didn't like that at all. Point is if my daily "load" temps are just 31c I could push it a bit more.
  13. Temperature vise you are sitting pretty and MB is good for OC, I'm surprised you couldn't hit 4.5 but what's your CPU cooler ?
  14. Eve 212. I have managed to get the overclock stable at 4.3 @1.375v and temps at 59c under IBT.

    Ram is 1600 9-9-9-24 1.5v
    Cpu/NB 1.225v (OC failed with this @1.20v)
    DRAM 1.5v manually set.
    Turning T probe on instead of extreme I get 2c lower temps and that is while currently testing 4.4ghz at the same voltages. The RAM stability has been causing me head aches and now it's stable at 1600 I am pushing to 4.4ghz. Not changing NB from 2200 and HT 2600. Those 2 changes brought some stability to my OC so just changing multi and turning off one cooling setting I had left on auto and going from Extreme to T probe for a 4.4ghz test. Even with 11-20c idle temps my cores jump fast and steadily increase to 61-62 with extreme instead of Tprobe. Just trying anything to squeeze a bit more out at 1.375v. Have been reading most people got 4.5 at 1.375v with mid 50s temps. I thinks this chip is a bit of a power pig so the temps are going up higher.
  15. 4.4ghz is about the comfort zone for a 212 mate in all honesty.

    Regarding your ram , what's your stick configuration ?? 4 sticks nearly always causes issues on am3+ boards when it comes to speeds past 1600mhz.
  16. 2x8 of 1866. 9-9-9-27. Under OCCT load I get 55-56c at 4.3ghz. On IBT I get 59 solid. Under gaming loads like witcher3 ultra settings I am getting 31c.... just setting 4.4ghz at this voltage 1'more time to see if it was my ram settings and not voltage. If I am right and my instability was due to ram settings then it should hold 4.4ghz @1.375v. This was always stable for 2 hours in prime but then small ft but then failed blend so tried IBT and OCCT.

    Thanks for the head up Matt. I kinda thought that anyway. I think this chip will stay at 4.3 as I have a feeling it's gonna fail 4.4.

    Just glad it's finally stable after 20 IBT runs.
  17. Try the ram at 1866 10-10-10-28 , it'll likely run at that.

    Thing is no there isn't really a performance difference between 1600 cas 9 & 1866 cas 10 (especially on an am3 build)
  18. Fell over at 4.4ghz and I am not happy to push to 1.38v with the temps I am getting. Will be building a 7700k with a 750D case in a few months. That will be fun!
  19. 4.4 can be a tipping point mate , if you look at my rigs in my sig I hit 4.3ghz at 1.332.
    4.4 took 1.3725, anything above close or past 1.4v.
    Just one of those things , every chip is different.
  20. In my experience, 1.4v is most CM 212Evo can handle and that translates to no more than 4.5GHz, 4.4 more likely. I needed much larger cooler, Mugen 3 or CM 616 to hit 4.7 at a bit less than 1,5v and CM Nepton 140XL for 4.8v @1.52v but every processor sample may be different. In any case, your MB is not a factor, I have seen comfortable 5GHz on those boards.
    It all boils down to voltage needed for particular frequency you are gunning for because that's what produces heat with VRM to be able to handle loads.
  21. Well just completed 10 runs of IBT at the smallest setting to test heat on 1.38v and I never broke 57c. Notched it up to the 4000mb test and temps hit 59c so stopped half way and am now testing 13000mb on a 20 run test. Temps still holding at 59c. Did some gaming beforehand and clocked 44c max. That seems pretty good to me if it holds. I won't make 4.5 on this cooler due to its need for power BUT 4.4 has been my target for a week now and it's finally getting there and very stable thus far.

    For the 7700k build I was thinking the large Noctua triple fan set up or a OI cooler with a 280mm rad. That should allow a lot more cooling but have heard the Noctua coolers are almost on par with the OIs. Comes down to noise for me though. Probably go with Noctua:
    http://www.ebuyer.com/663815-noctua-nh-d15-dual-radiator-quiet-cpu-cooler-with-two-nh-a15-fans-nh-d15
  22. Kuniyosh said:
    Well just completed 10 runs of IBT at the smallest setting to test heat on 1.38v and I never broke 57c. Notched it up to the 4000mb test and temps hit 59c so stopped half way and am now testing 13000mb on a 20 run test. Temps still holding at 59c. Did some gaming beforehand and clocked 44c max. That seems pretty good to me if it holds. I won't make 4.5 on this cooler due to its need for power BUT 4.4 has been my target for a week now and it's finally getting there and very stable thus far.

    For the 7700k build I was thinking the large Noctua triple fan set up or a OI cooler with a 280mm rad. That should allow a lot more cooling but have heard the Noctua coolers are almost on par with the OIs. Comes down to noise for me though. Probably go with Noctua:
    http://www.ebuyer.com/663815-noctua-nh-d15-dual-radiator-quiet-cpu-cooler-with-two-nh-a15-fans-nh-d15


    Noctua coolers = incredible quality & performance - that bloody colour scheme though !! I just dont get it ,does anyone own any case & components that fit in with brown & tan fans ???
  23. I think you can buy black ones, aftermarket. Good sales strategy. Make the worlds best cpu air cooler and give it fugly fans so people buy these after purchase:
    http://noctua.at/en/nf-f12-industrialppc-2000-pwm
  24. Kuniyosh said:
    I think you can buy black ones, aftermarket. Good sales strategy. Make the worlds best cpu air cooler and give it fugly fans so people buy these after purchase:
    http://noctua.at/en/nf-f12-industrialppc-2000-pwm




    I prefer the black & grey redux still.
    Thats the issue though,youre talking a £70-80 cooler & replacing the fans is another £30 - granted you can drop the included ones as front intakes where you cant see them but that is an awful lot of money.

    For £80 though you can get the cm airmaker 8 & the deepcool gamer assassin 2 which perform so close to the noctau as to make it negligible & both will look far better doing it.
  25. 4.4 stable under very high IBT 20 runs. Should I do 20 maximum just to be sure? Some say no need.
  26. I'm one of those who'd say no need at all mate.
  27. 4.4ghz @ 1.3875 61c full load 20 runs IBT STABLE!
  28. 8 runs IBT custom 12000mb of 16000mb and test failed. Voltage needed would be 1.39-1.40v and that is just to much heat for the 212 eve. So 4.3ghz it is but it shows. The difference in witcher 3 is 5-10fps. I cannot wait for my 7700k build. Getting the maximum hero of the next one up. May go for the vega when released as I need to do video editing but we will see. Won't do a custom loop but am tempted to get an IO but heard they are noisy. You guys recommend any?
Ask a new question

Read More

Overclocking RAM CPUs