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

My Journey - Overclocking My 8350 and 7950.

Discussion in 'Overclocking & Cooling' started by jed, Dec 11, 2012.

  1. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Just got a new system built a few days ago and I'm beginning the fun and frustrating process of overclocking it to it's limits. I would be happy to get input from anyone w/ experience OCing the 8350 or 7950! Also post your stable specs, benchmarks, etc. if you'd like.

    I'm going to keep a running updated 'chart' in this first post just to help keep track of progress. Benchmarking with 3d Mark 2011 on the medium mode (can't buy that stuff right now!) and using Prime95 to quick check CPU/RAM stability, usually for 5-10 mins at a time.

    I am using the ATI Catalyst tool to simply OC the GPU and Memory clock settings and manually control the fan. Looking for a good program to start tweaking GPU voltages, haven't done this for a couple years. Don't remember what I used to use...

    CPU is being monitored by CPU-Z, GPU by GPU-Z and system temps by HWMonitor. CPU is being tweaked in the BIOS. All "Test X" are 3dMark 2011 tests at medium mode. Other tests are specified...

    Test 1 (8.9 drivers) - P6898 - G 7214 P 7417 Comb 5679

    Test 2 (9.0 drivers) - P6965 - G 7103 P 6888 Comb 6173

    Begin GPU OC - DEFAULT 925 / 1250 @ .950 V

    Test 1 940/1270 - Max 61C
    7230 - 7454 / 6932 / 6229

    Test 2 965/1300 - 62C
    7353 - 7631 / 6927 / 6230

    Test 3 990/1325 - 62C
    7488 - 7824 / 6924 / 6244

    Test 4 1010/1350 - 63C
    7598 - 7987 / 6916 / 6245

    Test 5
    1075/1375 - 63
    7716 - 8155 / 6924 / 6263

    Test ATI Tool - 3:30 sec, stable 56C w/ manual 80% fan

    Test GPU Tool - stabilized ~62 C w/ 80% fan

    Test 6 1100/1400 - 64C
    7874 - 8489 / 6736 / 6107

    Test 7 - 1110/1410
    Screen lockup or monitor shutoff

    Assuming gonna have to start tweaking volts now. Pretty good boost without doing it yet...

    Default CPU/RAM:
    Clock ratio: 4000MHz
    Multiplyer: 20
    Clock: 200
    Mem Clock: 6.66
    Memory: 1333
    NB: 2200
    CPB (boost) Ratio: 4200
    HT: 2600

    Ran Prime95 - Stable @ 53 C for 20 minutes. Small FFTs

    BEGIN CPU OC

    Test 8 - 1095/1395 w/ 20x205=4100MHz, 1366 RAM
    8032 - 8597 / 7024 / 6288

    CPU Test 1 : x20 x 205, 1366 RAM, Prime95 5 min @ 48 degrees

    Test 9 - 20x210=4200MHz, 1399 RAM
    8062 - 8597 / 7078 / 6409

    Test 10 - 20x220=4400MHz, 1466
    8138 - 8580 / 7385 / 6603

    Test 11 - 220x227=4540 MHz, 1512
    3dMark BSOD at last test, code 124 (voltage)

    Test 12
    - Same as above, increased voltage from 1.425 to 1.4375
    8189 - 8563 / 7585 / 6779

    Test 13 - 20x232 = 4640 (CPUz reads 4800) - 1546 RAM, 1.4625 V
    8220 - 8541 / 7712 /6955

    CPU Test 2 : x20 x 232, 1546 RAM, Prime95 5 min @ ~58 degrees, max 61

    CPU Test 3: x20 x 232, 1546 RAM, Prime95 Blend started at 3AM
    Blend ended 12 PM. Monitor unresponsive - showed only black screen with mouse pointer that didn't move. Forced reboot.
    Errors that are shown in results file upon reboot:

    [Tue Dec 11 03:04:51 2012]
    FATAL ERROR: Rounding was 0.484375, expected less than 0.4
    Hardware failure detected, consult stress.txt file.
    FATAL ERROR: Resulting sum was 6.900125323877905e+018, expected: 6.900125323877972e+018
    Hardware failure detected, consult stress.txt file.
    FATAL ERROR: Rounding was 0.4992675781, expected less than 0.4
    Hardware failure detected, consult stress.txt file.
    [Tue Dec 11 03:11:53 2012]
    FATAL ERROR: Rounding was 0.5, expected less than 0.4
    Hardware failure detected, consult stress.txt file.
    FATAL ERROR: Final result was 00000000, expected: 50F74AF0.
    Hardware failure detected, consult stress.txt file.

    CPU Test 4: same clocks, Started at 12:30 PM, instant BSOD. Codes: 101, 019, 000, 04, FFFFF880009B3180. Tweak CPU voltage...

    CPU Test 5: reverted back to 20x227, 1512, p95 5 min @ 55 degrees... 1 error on worker #8.
    Error is shown:
    FATAL ERROR: Rounding was 0.5, expected less than 0.4
    Hardware failure detected, consult stress.txt file."

    CPU Test 6: 22x205, 1.425, 1366 RAM. x9 HT and x10 NB. Prime95 5 min - 0 errors

    CPU Test 7: 23x200, 1.450, 1333 RAM. Same HT/NB. Prime95 hits 5 min @ 53 C - 0 errors

    CPU Test 8: 23.5x200, 1.45, 1333. Prime95 hits 5 min @ 53 C. 1 error, 100 warnings...
    Errors read:
    FATAL ERROR: Rounding was 0.5, expected less than 0.4
    Hardware failure detected, consult stress.txt file.
    ERROR: ILLEGAL SUMOUT
    Possible hardware failure, consult readme.txt file, restarting test. (repeat this error 100 times)

    CPU Test 9: Same clocks/freqs... Prime95 hits 5 min @ 54, still running. Max temp 62C. One error on thread four, test stopped after 4 tests. One error on thread 7, stopped after 24 tests.
    Errors read:
    FATAL ERROR: Rounding was 0.4975585938, expected less than 0.4
    Hardware failure detected, consult stress.txt file.
    FATAL ERROR: Rounding was 0.4985351563, expected less than 0.4
    Hardware failure detected, consult stress.txt file.

    Add .020 DRAM Volts

    CPU Test 10: 23.5x200, 1.45 V, 1333, DRAM @ 1.520 V - Prime95 hits 5 min @ 52C, no errors. 4 successful runs.

    Going to begin some BF3 @ 7:30 see how these clocks hold. Going on 40 minutes with 54 Max CPU temp and 59 max GPU temp.

    Ran BF3 for 2.5 hours on all ultra, highest temps 57 CPU, 60 GPU.

    CPU Test 11: Same clocks. Prime95 Started at 10:07, running until 12:50. Two errors in core 4 and five, same errors as usual.
    Errors read:
    FATAL ERROR: Final result was 00000000, expected: E91EEB5A.
    Hardware failure detected, consult stress.txt file.
    [Tue Dec 11 22:18:49 2012]
    FATAL ERROR: Rounding was 0.4924316406, expected less than 0.4
    Hardware failure detected, consult stress.txt file.

    Let PC sit overnight w/ monitor shut off. Turn monitor back on and "NO SIGNAL" error w/ black screen. PC components running fine inside of my case... Have to force reboot. GPU OC settings were disabled overnight. Went into monitor options and disabled DDC/CI for troubleshoot #1.

    ADD 12.11 BETA DRIVERS

    Test 14: 1095/1395, CPU @ 23.5x200 (4700), 1.45v, 1333 RAM @ 1.520v
    8696 - 9718 / 7151 / 5938

    Beta drivers added huge increase to graphics score. From 8541 to 9718... Now I need to figure out why my CPU scores are bringing it down, and scoring even lower than my lower clocks from before!!

    Revert to lower CPU clocks but higher RAM freq...

    * Test 15: 1095/1395, CPU @ 20x232, 1.4625v. 1546 RAM @ 1.51v...
    8965 - 9671 / 7860 / 6708 - best test yet by far.

    Only way I'm getting RAM to OC is to raise the FSB Freq and lower the multiplyer (to be able to stay around 4.7 Ghz). The only way the RAM boots successfully is by sticking to the default multiplyer (6.66) and upping the FSB freq.

    CPU Test 12: Starting Prime95 at 1:05 AM. Clocks are 20x232 @ 1.4625v and 1546 RAm @ 1.51v. Ran BF3 for about 3 hours with no issues prior to. 3 cores already shit out after about 10 minutes... Same old BS errors. Test stopped at 1:22 AM... Temp got up to 65 as well which isn't chilly by any means.

    CPU Test 13: Have the CPU at stock clocks/settings as well as RAM. Going to reboot the OC process. No trouble running games, etc. with previous clocks but terrible stability with prime. Prime started around 11:00 and still running good at 3:30 AM. Shutting down prime error-free at 12:30 PM. Highest CPU temp was 51 C over a 13.5 hour stretch.

    CPU Test 14: 21x200 (4200), 1.36v. P95 Blend - 12:35 to 1:35 PM max temp 47C. Intel BT on High - 5 passes, max 46 C/61.5W. Intel Burn 10 passes at Very High - 47 C, 62.2 W.

    CPU Test 15: 21.5x200 (4300), 1.36v. Intel BT on Very High: 10 passes 2432 seconds, 44 C/ 62.6 W.

    CPU Test 16: 22x200 (4400), 1.36v. Intel BT on VHigh: Instant BSOD. Error 4A.

    Add volts from 1.3625 to 1.3875

    CPU Test 17: 22x200 (4400), 1.3875v. Intel BT on VHigh: BSOD after a couple passes, error 1e: "ADD MORE VCORE" according to the BSOD thread.

    Increase volts.

    CPU Test 18: 22x200 (4400), 1.425v. Intel BT on VHigh: 10 passes, max 50 C, 62.1 W, 2402 seconds.

    CPU Test 19: 22.5x200 (4500), 1.425v. Intel BT on VHigh: 10 passes, max 54 C, 62.1 W, 2433 seconds. P95 40 FFT tests for 1 hour: 0 errors, max 54 C, 60.8w.

    CPU Test 20: 23x200 (4600), 1.425v. IBT on VHigh: 3 passes, code B3 BSOD. "Add vcore".

    Add V-Core - 1.425 to 1.4375

    CPU Test 21: 23x200 (4600), 1.4375v. IBT on VHigh: 3 passes in 717 seconds, auto shutoff. 3.07, 3.07, 3.94. 50 C, 61.4 W.

    Increase V-Core 1.4375 to 1.4625

    Game Test: GPU OC'd and CPU = 23x200, 1.4625v. Ran BF3 for about 1 hour. Max 55 C CPU, 60 C GPU.

    CPU Test 22: 23x200 (4600), 1.4625v. IBT on VHigh: Failed after 6 passes. 3.07 to 1.61. 56c/62.2w.

    Add volts - 1.4625 to 1.4875

    CPU Test 23: 23x200 (4600), 1.4875v. IBT on VH: 10 passes, 2375 seconds, 63C. P95 Blend: Ran from 2:00AM to 12:30 PM - high 65 C. Core 5 1 error after 26 tests. Core 4 stopped after 399 tests. Core 4 made it until 10 minutes before I ended...

    Game Test 2: Ran Far Cry 3 on all Ultra, 2x AA for literally 6 straight hours. No issues. High temps were: 63 CPU, 71 GPU.

    :toast:
     
    Last edited: Dec 19, 2012
    Super XP and nt300 say thanks.
  2. Jack1n

    Joined:
    Oct 8, 2012
    Messages:
    1,020 (1.44/day)
    Thanks Received:
    229
    Location:
    Central Israel
    It seems like my sapphire 7950 dosent like a high memory OC either,which card exactly do you have?

    Edit:never mind i saw which one you have on a diffrent thread.
     
  3. MxPhenom 216

    MxPhenom 216 Corsair Fanboy

    Joined:
    Aug 31, 2010
    Messages:
    9,965 (6.74/day)
    Thanks Received:
    2,223
    Location:
    Seattle, WA
    For GPU stress testing. Just play BF3 for a few hours. Dont use tools.

    Possibly try bumping Vcore a bit for the CPU.

    And use MSI after burner for GPU overclocking so you can tweak voltages. Assuming the 7950 you have doesnt have locked voltage.
     
    jed says thanks.
  4. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    How much is a bit... I've already bumped from 1.425 to 1.4625 for just a small clock increase. How much can these 8350's take? 1.5 would be getting scary with my x6 so that's what I'm used to. Will DL Afterburner.
     
  5. Norton

    Norton WCG-TPU Team Captain

    Joined:
    Dec 21, 2011
    Messages:
    8,601 (8.58/day)
    Thanks Received:
    19,447
    Here's my 8350 (4.4Ghz (200x22), Turbo off, Voltage- Auto (appr.1.35v avg), memory@ 1600Mhz)

    [​IMG]

    Temps are great using a Xig Night Hawk (2x 1500rpm fans) :toast:

    I don't bench but it has been crunching 24/7 @100% load since I installed it a few weeks ago w/no errors. :cool:
     
    jed says thanks.
    Crunching for Team TPU More than 25k PPD
  6. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Hey Norton, thanks for sharing. Maybe i'll try leaving the clock speed and upping the multiplyer instead. Is the "Turbo" perhaps the "Core Performance Boost" I'm seeing in my BIOS? If not, where did you turn yours off? Also, how do you get your HWMonitor to view each core's temp? Mine only shows the whole package.
     
  7. Norton

    Norton WCG-TPU Team Captain

    Joined:
    Dec 21, 2011
    Messages:
    8,601 (8.58/day)
    Thanks Received:
    19,447
    It sounds like it but I'm not sure how the BIOS is setup on your board- your best bet would be to check the manual to make sure.

    The HW Monitor is the latest version- not sure of your version but was always able to read the core temps regardless of version. Your issue may be the version or the board you're using (M5A99X Evo here)... remember that the core readings aren't entirely accurate but they are a decent reference when running under load.
     
    Crunching for Team TPU More than 25k PPD
  8. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Thanks for the info man.

    Curious, should I be leaving the NB and HT freqs/multipliers at auto, default or should I be scaling them up as I bring up the overall clock speed? I read a long time ago to 1) not let the HT exceed the NB and 2) try to keep them each under 2 GHz. This was a while ago.
     
  9. Norton

    Norton WCG-TPU Team Captain

    Joined:
    Dec 21, 2011
    Messages:
    8,601 (8.58/day)
    Thanks Received:
    19,447
    iirc default/auto NB is 2200 and HT is 2600 on my FX. I have no trouble with those settings but am not an extreme overclocker :eek:

    I do remember hearing that FX is more sensitive to NB and HT changes so you may want to do some more research.
     
    jed says thanks.
    Crunching for Team TPU More than 25k PPD
  10. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Thanks mate.

    Ran a few more tests, updated in the main post. These "Rounding" errors are really bugging me.

    Edit

    Supposedly the rounding errors are due to lack of RAM voltages (in most cases). Bumping from 1.5 to 1.520. Although my ram is currently stock speed...
     
    Last edited: Dec 12, 2012
  11. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Keep losing connection to my monitor after I let my PC run overnight... This ever happen to anyone else? Works fine when you go to bed, shut the monitor off, wake up and turn it on and get "no signal". Have to force reboot. My GPU OC settings are disabled at this time so I'd imagine that's not it...
     
  12. Radical_Edward

    Radical_Edward

    Joined:
    Jan 24, 2010
    Messages:
    3,587 (2.11/day)
    Thanks Received:
    1,927
    Location:
    Oregon, USA
    What video driver are you using? Some of the ones that shipped with the cards did that. (I use the 12.11 beta 4 drivers currently.)
     
    jed says thanks.
    Crunching for Team TPU
  13. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Hey man, using the 12.10 version currently. It did this with my old card as well and then stopped... Maybe it was due to a cable change or something? It's got VGA, DVI and HDMI. Only one of the three worked last time (I think DVI, or maybe it was VGA....)
     
  14. brandonwh64

    brandonwh64 Addicted to Bacon and StarCrunches!!!

    Joined:
    Sep 6, 2009
    Messages:
    18,583 (10.12/day)
    Thanks Received:
    6,090
    Location:
    Chatsworth, GA
    Install the newest driver. Hell the 12.11 is a performance driver and will give you a decent noticeable boost!

    http://www.techpowerup.com/forums/showthread.php?t=173865
     
    jed says thanks.
    Crunching for Team TPU
  15. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Why does my Catalyst say it's up to date and not inform me when new updates are available? >_< Maybe since it's a beta I guess... I'll DL the drivers though.

    How do you guys install it? Usually I DL the whole suite. Uninstall the whole current one and reinstall the whole new suite. This one's just a driver.

    Just uninstall the driver from device manager?
     
  16. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Okay it came with a whole suite so it is installed. Huge bump in 3dMark score. Went from 8220 Overall to 8696, and 8541 to 9718 in just the graphics score! Wow.

    Does anyone know why I get lower physics/combined scores with my current CPU clocks (23.5x200=4700, 1.45 V, 1333, DRAM @ 1.520 V): 7151 Phys, 5938 Combined...

    Than I did earlier with lower clocks (20x232= 4640 - 1546 RAM, 1.4625v CPU): 7712 Physics/6955 Comb...

    Does that small bump in RAM really make those scores jump by almost 1,000 each?
     
    Last edited: Dec 12, 2012
  17. eidairaman1

    eidairaman1

    Joined:
    Jul 2, 2007
    Messages:
    12,393 (4.70/day)
    Thanks Received:
    1,456
    yes it can, up that ram to what you had before. Heat is another issue
     
    jed says thanks.
  18. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Thanks for the info. Should I just let my RAM scale up as I increase clocks and let it get as high as I can? Whenever I try to actually clock it to 1600 or something it won't POST and resets RAM to 1333... But when I let it gradually increase (like I got to 1546 before) it would boot up and obviously make a big deal in 3dMark... OR am I missing something with OCing the Ram? Tried increasing volts to 1.52 from 1.5 and still wouldn't POST after trying to get it to 1600.
     
  19. Jstn7477

    Jstn7477

    Joined:
    Aug 30, 2009
    Messages:
    3,833 (2.08/day)
    Thanks Received:
    1,528
    Location:
    Sarasota, Florida, USA
    Are you setting it to 1600 when the base clock is at 200? I don't remember from your other thread but what speed of RAM did you get? Check the timings too and make sure it's at 9-9-9-28 @ 1600.
     
    Crunching for Team TPU More than 25k PPD
  20. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    I try 1600 by raising that DRAM multiplier to 8.0 instead of 6.66. That changes it from 1333 to 1600. The timings are 9-9-9 for sure. Yes, I've tried both when the clock is around 225 and also 200. The base freq of the RAM is 1333. It should at least be able to hit 1600, I just need to figure out how to go about doing it.

    Do I need to raise NB clock freq beyond the automatic raises it gets when changing the CPU overall ratio? Or a certain type of voltage needs to be raised other than DRAM volts? The last bootable RAM increase was: 20x232 CPU ratio, 1.4625 V - 1546 RAM w/ no voltage boost. This errored out (small errors) in Prime95 but ran games for 3-4 hours w/o a crash.

    Edit:
    Just tried a slew of different things. Upped NB freq, didn't boot w/ 1600. Tried it with clock @ 200 and clock at others like 235 etc., didn't boot. I've now got it to boot at the current settings: 20x232 4.64 GHz, RAM is @ 1546 MHz @ 1.51 volts.

    Only way I'm getting RAM to OC is to raise the FSB Freq and lower the multiplyer (to be able to stay around 4.7 Ghz). The only way the RAM boots successfully is by sticking to the default multiplyer (6.66) and changing the FSB freq.
     
    Last edited: Dec 13, 2012
  21. MxPhenom 216

    MxPhenom 216 Corsair Fanboy

    Joined:
    Aug 31, 2010
    Messages:
    9,965 (6.74/day)
    Thanks Received:
    2,223
    Location:
    Seattle, WA
    You can put 1.65v though that kit im pretty sure. Now you need to run Prime95 blend.

    Hate to say it, but should have got a 1600 or 1866 kit. Would be have been a bit easier haha.
     
  22. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Blend still gives tons of errors that I have no idea what they mean. I can run BF3 all night but blend lasts 10 mins before it starts shutting threads down with either "hardware failure" (thanks for being specific!) or "expected .05, got .04" or something.

    I agree I should have, didn't know I was upgrading the whole system at the time I ordered though (-_-) For now I'm not too concerned because it still blasts BF3 @ ultra without touching the stock settings on the CPU. But you know how we all like to blast these f***ers until they want to melt.
     
  23. MxPhenom 216

    MxPhenom 216 Corsair Fanboy

    Joined:
    Aug 31, 2010
    Messages:
    9,965 (6.74/day)
    Thanks Received:
    2,223
    Location:
    Seattle, WA
    Instability man, and frankly if you cant run Prime95 for 8+ hours, your system isnt stable

    Maybe drop the clock to 4.4 and go from there. I think it might be memory though. Maybe losen the timings or put more voltage through it.
     
  24. jed

    jed

    Joined:
    May 14, 2011
    Messages:
    285 (0.23/day)
    Thanks Received:
    45
    Should I just bust it up to 1.57-1.58 then? The stock is 1.5 but if it can supposedly take 1.65...
     
  25. MxPhenom 216

    MxPhenom 216 Corsair Fanboy

    Joined:
    Aug 31, 2010
    Messages:
    9,965 (6.74/day)
    Thanks Received:
    2,223
    Location:
    Seattle, WA
    LOL your memory is $38 on the egg. Thats the same price as the 1600mhz kit. Haha, side from that,

    I dont see much of an issue running up to 1.65v on that kit. I ran 1.64 on Corsair Vengeance kit that was rated for 1.5v. Id like the hear from Dave in this regard though.
     

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

Share This Page