Help with overclocking fx 4100

[OverStock]

Alright so first time overclocking and I just finished my first build 2 weeks ago. Going for 4.6ghz by the way since that's the highest stable 24/7 OC I've heard anyone using really. Here's my set-up.

AMD fx 4100
Coolermaster Hyper 212 Evo
Asus M5A99FX Pro 2.0
SeaSonic M12II 620w Bronze
G.Skill Ripjaws X Series 8GB (2 x 4) CAS 8 1.5v
Nvidia GTX 660
NZXT Phantom Case (2x bottom side 120mm intakes, 1x back top 120mm exaust, 1x top 200mm exaust) Looking to get a 140mm for the front and maybe another 200mm for the top.

Now to show you where I'm at first of all, I've been reading countless tutorials on the web for over clocking the fx 4100 and last night installed my Cooler master Hyper evo and started off upping the multiplier last night to see how far i could go with it. Got up to 4.2ghz stable at 35*C idle earlier today and 47* after an hour and a half of prime. After that I played Battlefield 4 for a match (15-20 minutes) to test out my fps and when i got done the temps were at 44* (low - medium settings) and then 45* after another match with high settings. Thought those temps were kind of high considering I was getting 47* on prime and i thought that would be a lot more intensive than a game, no? So anyway, then I went for 4.4 (still 35* idle) and got my first BSOD after 1 minute of prime. Guessing i need to start upping the voltage now right? Next I went down to 4.3 to see how stable that was curious to find the max stable multiplier OC for this set-up. 35* idle again and stable at 48* after an hour of prime. (Will post screen shots underneath)

So I'm trying to figure out what to do now. I've heard to just do multiplier and voltage on here and many other places, but the guide you guys have on here for OC'ing Bulldozers says to do a FSB. That's one thing i haven't researched too much into since I've been seeing so many guides on just multiplier and voltage so I was wondering if that's the path i should go down (im guessing my motherboard could handle it well right?) or if i should just go for the multi + voltage way. And either way I'd really love to have a little bit of help if anyone has time, especially to calm down my OCD with all of this lol.
Thank you in advance!

Here's the pictures at 4.3ghz after an hour of prime. By the way, should i submit screenshots at idle too?

Edit: Oh yeah by the way I have all the power saving features and all that stuff turned off (as far as i know).

 

[RGone]

For termps we need to know the cpu temp that you do show but you should have pulled on the bottom edge of HWMonitor and made it larger so we could see the more important Package/Core temp that is hidden. Don't just scroll that small HWMonitor since that then covers the voltages. Enlarge HWMonitor to just so the Package/Core temp shows up with everything else already there. Nope don't need video card just Voltages thru the Package/Core temp.

Recommend Max of 70c on CPU_Socket temp.
Recommend Max of 62c on Package/Core temp.

3.6Ghz is stock speed for the FX-4100 so at 4.3Ghz you are already 700Mhz over stock speed. You need to add Vcore by at least two increments to even believe 4.4Ghz will be stable for 2 hours of P95 Blend and it may take a little more than 2 upward increments of CPU Voltage When most any cpus get to X point the next bump up in Vcore is not just a minor one but often pretty hefty. Give it some electron go juice.
RGone...

 

[OverStock]

Thanks for the quick reply first of all! Appriciate it. And yeah I thought i should probably include the rest of hw monitor but thought of it after i already stopped running prime, but next time I'll remember that. Also I was running prime in small FFT test after hearing someone on youtube say that was the one to go with to stress test the cpu with, but would blend give me more accurate or meaningful results?

So should i just up the voltage by 2 increments or 3 and then try 4.4 or should i go for the method i noticed in another thread that you liked to do for minimal crashes with going to the highest vcore you would want to use and then dialing the vcore back down with the multiplier set at well say 4.6 in my position, untill you crash? And if so, what would you recommend for the vcore? 1.4 ok? Didn't want to go too high.

 

[RGone]

P95 Blend is the defacto 'standard' we use here in the AMD Cpu Forum section.

2 increments. 3 increments. What is a few increments between friends. Mainly increment enough to pass P95 Blend mode at the next cpu speed you wish to test for stability.

Yes I do go ahead and dial in the "most" Vcore I plan to use and then continue to up the multiplier until I do get a crash. Now since I am at the MAX Vcore I ever plan to use, only one thing to do and that is back down the overclock a little until P95 Blend runs for at least 2 hours and temps stay inside of the range already stated above. Shortcuts all the little wienie steps that may BSOD my windows install and is just as sensible as I can reason. If I am at MAX Vcore then the issue is not Cpu voltage but something else I need to slow so I can pass P95 Blend mode. And since we overclocking only with the multiplier we are NOT screwing with any of the other three busses and only have to lower the multiplier by most likely 0.5 and repeat P95 Blend and we are just about job done. Nothing to do but tweak up the other busses and see if the ram will overclock and job done. I can take a never before seen FX processor and doing as described, have it within just a few points of Max Stable in a few hours not days.
RGone...

 

[OverStock]

Alright so i went for 4.6ghz tonight and gave it 1.4v to work with. Was at 36* idle and after 2 hours and a half of prime (blend this time) it was only at 52*C Socket and 50* on the package. (Screen shots below) I'm going to try to go for 4.7 now on the same voltage and see how that does. Probably wont keep it at that stable (possibly) at least for now, but stoked its at 4.6 at 1.4v stable as I've seen some people have to have higher voltage for even lower clocks. After i try 4.7 I'll try to dial down the voltage until it crashes and post the results. Any suggestions as of now?

By the way, on the core temp screen shot it shows 57* as the maximum, but it only showed that as the temperature for a split second when i opened up core temp a little bit after i was already running prime. Altough it does say 58* in hw monitor and didn't notice when it went up to that, but the whole time it was mainly around 52* going up to 53 a few times around the first hour mark of prime. Also didn't know if i needed to include the memory and SPD tabs of CPU-Z at this point, but i did anyway.

Edit: Forgot to put this in, but core #3 wasn't doing much most of the time. Now looking back into the prime window worker #3 says "ERROR ILLEGAL SUMOUT" after about 20 minutes of testing. (Screenshot below) Whats wrong with this core? ahem "thread" (im guessing core #3 would be a thread right?)

 

[trents]

The pics are still valid even if you're done running Prime. We mainly want to see the max temps and they will still be there if you don't shutdown HWMonitor.

 

[trents]

Temps look good but you should consider tightening your RAM timings. Compare what you are in currently running them at ("Memory" tab of CPU) with what the manufacturer recommends in the XMP-1600 column of the "SPD" tab. With the exception of the CR timing you could snug down the others quite a bit as long as you loosen the CR from 1T to 2T. I would also look at bumping up the HT Link and NB frequencies to about 2400 mhz. That may require a small voltage increase in the NB.

 

[OverStock]

Ok, I'm about to go to 4.7 now and ill bump up the HT Link and NB up to 2400mhz and see how that goes.. I'd like to tweak the memory like you said but I have no idea how to do that at this point so I'll have to do some research into that. I'm guessing mem test would be a good thing to run in terms of tweaking the memory though? That's about as far as I've really learned about the memory aspect.

 

[RGone]

Ok, I'm about to go to 4.7 now .

You are not yet stable at 4.6Ghz because you had an error of illegal sumout. Until you fix that at 4.6Ghz no sense in trying to go any faster. It surely will error there as well.
RGone...

 

[SubPar]

I have the same chip with a stable 4.4 o/c running 24/7 and the good people here helped my immensely, whatever they are telling you treat it like the voice of a higher power my boards ****tier and i managed a decent overclock, im sure you'll figure it out with their help. Like Rgone said dont try pushing till you are stable, that could lead to melty melty :

 

[OverStock]

Ok yeah i totally forgot that worker 3 failed at 4600mhz and just went up to 4700 increasing the NB and HT Link to 2400mhz as trent mentioned. Ran prime for 30 minutes and as you guys may have guessed worker #3 AND 4 failed this time. 3 failed right away (compared to after 24 minutes last time) and 4 failed about 10 minutes in. I'll try an increase in voltage at 4600 for now i guess? Anything else I could change? Trent mentioned the small increase in NB votage but not sure if thats CPU/NB Voltage or just NB Voltage in my bios.

 

[OverStock]

I have the same chip with a stable 4.4 o/c running 24/7 and the good people here helped my immensely, whatever they are telling you treat it like the voice of a higher power my boards ****tier and i managed a decent overclock, im sure you'll figure it out with their help. Like Rgone said dont try pushing till you are stable, that could lead to melty melty :

True, I will take your (their) advice. I'm definitely not trying to melt my first build lol!

 

[RGone]

Okay "OverStock" you were not stable at 4.6Ghz with your chosen 1.4V Vcore so as I wrote in Post #4, IF you are using what is going to be your Max Vcore and P95 fails then you must reduce the overclock until your chosen Vcore is enough for P95 Blend mode to work without error for at least 2 hours.

I think truthfully you picked an artificially LOW Vcore to play with since the whole idea of overclocking and by my give it the Maximum Vcore I will ever use for 24/7 daily use, that Vcore would have been more like 1.5Vcore and likely 4.6Ghz would have passed P95 Blend mode for 2 hours.

If you are going to cramp yourself in with a Vcore that most of us know will be too low to go up the speed scale then you might as well just hop-scotch up the Vcore and Cpu speed scale just a little step at a time and test for stability at each level of cpu speed increase. So start back over at some cpu speed like maybe 4.2Ghz and test for at least 20 mins of P95 Blend mode and if it does not fail P95 then move to 4.3Ghz and test and rinse and repeat until P95 Blend fails in some way and then up the cpu voltage and rinse and repeat until you climb up the speed ladder until you get to where the volts are more than you will use and/or the temps get out of what is considered to be safe with that being <70c on CPU/Socket temp and <62c on the Package/Core temp as shown in HWMonitor (Free) when it is open on desktop logging min/max temps and voltages AT the same time as P95 Blend mode is in operation.

If you want to report results to us we need to see the three (3) screen captures of CPUz > CPU Tab > Memory Tab > SPD Tab and additionally a capture of HWMonitor that was open and logging Min/Max temps while P95B was running.
RGone...

 

[OverStock]

Alright i ran 4.6 at 1.45v because I'm a little worried about going up to 1.5 right away and it had hit 65* on package for a second and it was staying at 62/63 package. So i turned it off right away when i saw it was at 63* without taking a screen shot. So im going to go down to 4.5 since im guessing giving it more voltage would make it get even hotter? Am I right on that? And is the NB voltage that trent was talking about raising the CPU/NB Voltage or just straight NB voltage in the bios? And should i raise that at this point with the NB and HT Link at 2400mhz? Would that help with temps? Or just stability?

 

[RGone]

ANY Voltage increase will ADD some amount of heat. Increasing speed will add some small amount of heat. So if it takes 1.45V on cpu to be stable but you do not like the HEAT then lower the Cpu Speed.

Most of us have cooling that allows for 1.25V on CPU_NB and that is n0t NB.
RGone...

 

[OverStock]

Ok thanks for confirming that. I just did 4.5 at 1.45v with the same 2400mhz on NB and HT link and after 18 minutes of prime it went up to 64* (package) for a second so i took screen shots and turned it off. By the time i got done taking the screen shots it was back down between 60-62 though. So any suggestions of what i should do now? Should i take the NB and Ht link down in voltage to get better temps at 4.5 or should i just go down to 4.4? What other ways are there to get temps down at this point? Would it help a lot installing a 140mm fan in the front of the case to go to 4.6ghz? Or another 200mm exhaust at the top?

Screenshots at 20 minutes of prime. Core temp min temp was high because i opened core temp after starting prime. Took the HW screenshot right when it was hitting 64.

 

[MongGrel]

65, bah, put that baby on 1.5 and play around.

get worried when ya see 90+ in ITB, he he.

 

[OverStock]

65, bah, put that baby on 1.5 and play around.

get worried when ya see 90+ in ITB, he he.

Well just to point out, that was at package not socket temps. They said the max recommended was 62 so wasn't trying to really go over that, but does that mean I shouldn't EVER go over it or just not to STAY over that? I'm still on the former side of things, but correct me if I'm wrong. Also should I do a burn test before i do a 12 - 24 hour prime?

 

[Mandrake4565]

Overstock just out of curiosity why did you jump from 1.4 to 1.45 Cpu V? You were losing the #3 worker when running prime at 4.6, which usually means you're close in voltage. I would set it back to 4.6 at 1.40625 and run prime again if it fails give it another bump in voltage. You may be able to get 4.6 and still stay under the suggested temps.

Also should I do a burn test before i do a 12 - 24 hour prime?

I find that if my rig can pass 2 hours of prime blend it will take anything that I throw at it without a hitch. That's what I use to as my stability benchmark.

 

[OverStock]

Overstock just out of curiosity why did you jump from 1.4 to 1.45 Cpu V? You were losing the #3 worker when running prime at 4.6, which usually means you're close in voltage. I would set it back to 4.6 at 1.40625 and run prime again if it fails give it another bump in voltage. You may be able to get 4.6 and still stay under the suggested temps.

I went to 1.45 because I didn't know how close i was to passing and i saw a lot of people had to go to almost 1.45 to get to 4.6 stable
and also i wanted to keep trying to go to the top voltage id feel comfortable with using and go down from there. But after a couple tests at 4500mhz bringing down the voltage to bring down the high temps i brought it down to 1.43125 and brought the NB and Ht Link down to 2200mhz and I'm still stable running prime after 2 hours and a half at 61* socket 60 - 62* package with a max of 62 socket and 64 package so I'd say that's pretty decent right? Think I'll drop the voltage down one more step as to try to not get a max of more than 62* and see if its still stable. Or maybe i should drop down the NB and the HT Link back to 2000? or Auto? Not really sure what to do. I'll probably try both.

I find that if my rig can pass 2 hours of prime blend it will take anything that I throw at it without a hitch. That's what I use to as my stability benchmark.

Okay thanks that's good to know.