Need help OCing fx4100

[Ajamma]

Hey guys, new member here and new to the OC world.. So last week I got some computer upgrades from the local shop, including an:
-AMD FX4100 zambezi
-Asus M5A97 LE R2.0
-Gigabyte 2048MB GDDR5 Geforce GTX 660
-OCZ900MXS PSU

I know its not the most impressive upgrades but unfortunately it was the best this shop had, and i didn't know any better. Main reason why I wanted/ needed upgrade cuz i had a power outage that fried my old PSU and decided to upgrade other components since i bought the original comp over 3 years ago. That plus I started playing Planetside2 which im learning is a VERY Cpu intensive game

Anyways back to the OC, so I been just trying to OC myself using what info i could gather from this site and others on OCing the fx4100 (most info from 2012) and not having much luck. Aiming for 4-4.2Ghz, which iv read is doable on stock cooling

So I started fresh today by clearing CMOS, setting up my current UEFI BIOS as such:
EZmode System Performance and "Normal"
Under Ai Tweaker:
Ai Overclock Tuner=manual
CPU Ratio=20.5
AMD turbo technology=disabled
CPU bus freq=200
PCIE freq=100
Memory freq=DDR-1333MHz
CPU/NB freq=auto
HT Link Speed=auto
CPU Spread Spectrum=auto
PCIe spread spectrum=auto
EPU powersavings mode= disabled
CPU Load Line Callbration=auto
CPU/NB Load Line calibration=auto
CPU & NB Voltage=offset mode
offset mode sign= +
CPU Offset Vlotage= 1.320V(0.043750)
CPU/NB offset mode Sing= +
CPU/NB offset voltage=1.225V(auto)
CPU VDDA Voltage=2.510V(auto)
DRAM voltage=1.500v(auto)
NB voltage=1.100v(auto)
NB ht voltage=1.223V(auto)
NB 1.8v voltage=1.800(auto)
SB voltage=1.100v(auto)

And everything under "CPU configuration" Disabled. (CnQ, c1E, svm, core c6 state, hpc mode, apm master mode)

With these settings I was able to do a Prime95 Small FTTs torture test for 30ish minutes with my CPU temp maxing at 59C and CPU cores maxed at 53C, which seemed pretty good. Iv read that i should be aiming CPU temp to stay below 65C and Core temps around 55C

Then i tried the Prime95 "blend" test and after a minute or 2 my CPU core #4 had an ILLEGAL SUMOUT

So Iv just went back into UEFI BIOs and bumped my CPU voltage 1 to ).050000 hopfully giving me enough voltage to get thru the blend test (20-30min ish), while making sure temp's dont get out of hand, just to get an idea if im stable enough for the night for some gaming. And if i am, was planing on doing the Prime95 "blend" test over night(6-7hours)

Anyways I know its alot of info but i just wanted to give as much as possible so hopefully you guys can let me know if Im on the right track, doing something completely wrong, or just some pointers.

Im using HWMonitor, CPU-Z, and Core Temp to monitor temps and what not, So if screenshots of anything will help thats not a problem.
Thanks!

 

[Ajamma]

Hey guys, back with an update. After bumping CPU voltage to .050000 I ran the blend test again and samething. ILLEGAL SUMOUT after 1min. So went back into Bios and bumped it up to 0.056250, started up the blend test again and was going great. CPU temp hit a max of 61C and the CPU cores maxed at 53C again but then I got another ILLEGAL SUMOUT at 18min. Since my temps are not changing to much i figure i am safe to bump up CPU voltage once more and hope that does the trick. would u guys agree?

Also just a couple questions on OC
-Should I be using Blend or Small FTTs test on prime 95. Iv been told use the small FTTs test for quicker stability tests for a quick idea on temps and stability. Thats why i started with the Small FTTs and with no errors or warrnings, So i wanted to see how it would do on the Blend test which has lead me to some cpu voltage increases in hopes of stability and dec temps
-Should I turn back any power saving options i disabled in the bios after i decide im happy with the OC or will that just mess it up.
Thanks again for your time.

 

[ShrimpBrime]

4.1ghz is a good clock at this temperature. You might be able to squeeze another .5 multiplier, but it may also take more voltage to pass the stress tests......

 

[Johan45]

to OCF Ajamma
How about you post up some info so we can make a more informed decision. We like to see the CPU-z main tab, SPD tab and memory tab. Then run prime95 with HWmonitor open and take a snp of that. This will give us alot of vital info. To help us help you

 

[Ajamma]

Ok Johan45, I will work on getting that for you right now.

 

[Ajamma]

Ok reporting back after 25min Blend test on prime95 with CPU voltage bumped up to 0.06250(no errors or warnings), and here is a link to screenshot of all tabs off CPUz, Hwmonitor and Coretemps open!

https://imgur.com/Lz2W73G

 

[Mandrake4565]

This is easier for us lazy people to read.

 

[Ajamma]

yea sorry i couldnt figure out how to do that, be sure to next time.

 

[ShrimpBrime]

yea sorry i couldnt figure out how to do that, be sure to next time.

Ya you have to about 70c being the max. You can push it a little more. try a little reference clock (200) and click 2 at a time ie: 202. Do a quick test each time you raise it up. Try leaving the voltage where it's at, you should have no issues. Everything stock like how you have it "should" be good for 210 bus or lower without changing the ram voltage or timings divider ect.

 

[Ajamma]

Ok, so is that the CpU bus freq your talking about right? I will try tweaking that when I get home. Also am I ok with any of my other options, like CPU/cpunb load line calibration or anything need or should be changed? Or just leave it if it seems to be working.

maybe worth mentioning that last night after my last stress test I wanted to see how the OC worked with Planetside2. I had cpuZ, hwmonitor, and core temps open to monitor while playing, temps seemed fine but I noticed Core #1 was taking most if the Load. Is this normal?

 

[RGone]

but I noticed Core #1 was taking most if the Load. Is this normal? = How many cores of a cpu are in heavy use depends on how the game software itself is written.

 

[Mandrake4565]

Ajamma, you seem to be heading in the right direction. We usually recommend 2 hours Prime Blend for stability. Though stability is different for everyone, if your pc doesn't crash for what you use it for, then it's stable. As posted above, max temps we consider acceptable for general usage are 70 on the Cpu and 62 on the Package/core. If you're looking to see how far you can push just keep upping the multiplier and add voltage when you have a prime blend failure. At some point you may want to run Prime to hours to see if it is truly stable. I find that 20 minutes of prime is good to establish temporary stability and see if my temps are in check, though when your are getting to the your limits you will need to add more Cpu V in order to pass 2 hours. This is where you will have to compromise between your OC speed and you temperature. At some point you may also want to start raising the FSB and see if you can get a bit further using it to OC, but this is where it will get tricky. Raising the FSB will also change the Memory speed, Ht Link and NB Frequency.

Furthermore, I have learned that taking my time and only changing 1 parameter at a time makes overclocking a lot less frustrating. Don't try to push to far too fast, when you get ahead of yourself it can become a truly experience.

 

[Ajamma]

but I noticed Core #1 was taking most if the Load. Is this normal? = How many cores of a cpu are in heavy use depends on how the game software itself is written.

ok cool, thanks for the info. Anyways im home now, was doing some more research and i'm thinking ill turn off CPU/PCIe spread spectrum from auto to off to see how that goes, then slowly up my Bus speed.

Update: So after reading that you guys say its ok to take the CPU temp to <70 and the cores the below <62 i decided i was going to up the voltage some more as i still have some breathing room on temps. With bus speed back to 200 i increased multiplier to 21 for 4200ghz, ran blend test and at 5min in core#3 get not a warning but a fatal error. Was thinking about dropping multiplier back and start increasing the bus speed. Good idea? ill report back my findings

 

[RGone]

Furthermore, I have learned that taking my time and only changing 1 parameter at a time makes overclocking a lot less frustrating. Don't try to push to far too fast, when you get ahead of yourself it can become a truly experience.

You can say that about two more times at least. I only have to think back about 6 mos and I got in a hurry beating up on MY own FX-8350 on the FatLady mobo and completely lost my way. Thought I had dorked my cpu and the entire schmeer. I have done this long enough to know freeken better and yet I screwed my ownself and had to change the mobo out to find my way again.

Was talking with fellow member C_D on the phone last evening and he asked a question after my question and I got up and went and got mY LOG of Vcores for each cpu speed and answered the question. Well maintained logs are a MUST as I see it. And especially so for a brain faht like me.
Bobert....

 

[Ajamma]

Hey guys, after a couple hours tweaking Iv come up with 4214ghz at a X21 cpu multiplier and a 200 bus speed. up'd the cpu voltage i bit since i last posted. Iv left all other voltages on auto(stock?), and the temps still seem to be reasonable.

I've attached a screenshot of my current setup after a 22min Prime95 stress test to get give you guys an idea of how its looking.

Maybe a few more tweaks and a longer stress test is still in order(thinking 2-3hours) before i can declare it "stable". how does everything look so far?

 

[Mandrake4565]

Ajamma, temps looks good but you're getting near the top. I would suggest trying for a 2 hour prime run to see if it is indeed "stable". If you fail, I would back down on the multi .5. since you're near the top temp wise and additional Cpu V will probably send you over.

As Rgone mentioned above take notes of your settings, it makes it easier to get back on track when you lose your way. In addition, I wouldn't run core temp and Hwmonitor at the same time, it can sometimes lead to wonky temperature readings.

 

[Ajamma]

Ajamma, temps looks good but you're getting near the top. I would suggest trying for a 2 hour prime run to see if it is indeed "stable". If you fail, I would back down on the multi .5. since you're near the top temp wise and additional Cpu V will probably send you over.

As Rgone mentioned above take notes of your settings, it makes it easier to get back on track when you lose your way. In addition, I wouldn't run core temp and Hwmonitor at the same time, it can sometimes lead to wonky temperature readings.

ok sounds good and thanks for the feedback. I will be running a 2-3hour stress test later this evening.