4770k + Maximus VI Formula - Prime95v28.5 Temps & OCing battle

S

Syphon Filter

2[H]4U
Joined
Dec 19, 2003
Messages
2,597
Guys,

It's been a while since I've posted so I am hoping someone can help me out here. My system is as listed in my sig. Namely the following:

4770k + H100i
Maximus VI Forumla
2x 8GB Kingston 2400 CL11
AX1200i PSU

I've just tried to run prime95v28.5 at STOCK settings (system in sig) using the Small FTTs setting. My temps pretty much immediately shot up to 100/95/94/89 (cores 0-3 respectively). These temps were recorded with RealTemp 3.70. What are your thoughts on this? Is this normal for a 4770k running at stock? It seems (really) excessive to me.

Running the IETU Stress Test at STOCK sees temps hitting the low 70s at their highest so I am alarmed by the Prime95 temperatures. With cores hitting 100 the system will surely throttle/shutdown no?

Secondly, it seems I have a below average overclocker.

At 1.35Vcore/1.9Vin/1.25Vcache I can boot into Windows at 4.6GHz but not get stable. Same goes for 4.5GHz. This includes trying various LLC settings up to Level 4 (50%). The system either hard-locks or BSODs. I am guessing this chip just won't do it and I lost the lottery.

I can get 4.3GHz out of the chip quite nicely at 1.275Vcore/1.9Vin/1.25Vcache and a touch of LLC (Level 1 or 2). The system seems stable and completed a an overnight run (8 hours) of the IETU Stress Test. At these settings I see temps in the high 80s according to RealTemp. I have NOT tried running Prime95v28.5 though because of the above concerns about temperatures.

This is as far as I have got and I am unsure what to do next. As such, I would really appreciate some of your thoughts/help on all of the above. I look forward to the reponses.

Cheers,

sf.

EDIT:

Even running a Blend test in P95 at stock soon sees temps hitting the 100 degree mark :( WTF is going on???
 
Last edited:
Ahhhh...

Is this the issue with running Prime95 with voltages on "Auto" which I believe means "Adaptive"? Is Prime causing my voltage to sky-rocket resulting in excessive heat???
 
Shoddy TIM process. You sure your pump is on high rpm? Mine is at 62 from running prime for an hour on the same board you have with an h105. Ive been talking to a guy on YouTube and his is just like yours. His runs about 97 with an h105. Amazing how these are all over the place like this.
 
I have the RPM monitor from the H100i plugged into my mobo and it's reporting 2300rpm or thereabouts.

Does that sound right to you? I don't run the Corsair Link software (because it's rubbish) so I only have my mobo monitor to go on. The power is plugged straight into the PSU.
 
Thats horrible temperature for a 4770K at stock settings with a h100i... The pump its good mine its at ~2350RPM.. But i would check the backplate first, see, the backplate have a unique mount position... And its like this: (http://cdn.overclock.net/a/a7/350x700px-LL-a71b1057_backplate.png) (sorry cant attach in a propper way bc im on cellphone right now).. You should be sure its aligned properly with the CPU retention bracket bolts at the back of the motherboard... Second haswell do not like primer95 in none of their version even in the newer one, use aida64, for max stress use only FPU for stability use CPU+FPU... It will load at better levels (more similary to heavy games like crysis 3 or far cry 3)..

For haswell its never recommended to use auto voltage or adaptive.. Use manual voltage and set windows power management to performance and change the minimum processor state to 1%. If you have the motherboard app (dual intelligent processor 4 for your board) as mentioned in this post: (http://hardforum.com/showpost.php?p=1040793519&postcount=7) even using fixed/manual voltage.. Windows will be able to underclock and undervolt under fixed voltage, but what its most important the voltage will be capped and you will avoid those insane temps.. For stock settings you can try a fixed voltage of 1.000v or 1.050v. You should be fine with that..

And for last thing.. Delid it.. Probably you got one of those with crappy TIM and bad mounted processor heatsink..
 
Syphon Filter said:
Ahhhh...

Is this the issue with running Prime95 with voltages on "Auto" which I believe means "Adaptive"? Is Prime causing my voltage to sky-rocket resulting in excessive heat???
Click to expand...

yes, only use manual.

Edit: whoops Araxie beat me to it. You don't want to know what kinda voltage you were running. I'm hoping they get this issue right with the z97 boards but it's probably a problem with the chips. Not sure, really.
 
since your going to probably have to redo the past and block why not try stock intel fan just to check and see if maybe your h105 is bad?
 
Drew1966 said:
since your going to probably have to redo the past and block why not try stock intel fan just to check and see if maybe your h105 is bad?
Click to expand...

Ah shit, good call. I literally just cleaned, repasted and reseated my H100i.

Araxie:

Are we saying that Prime95 is a waste of time? Even with the new TIM (Antec F7) I am seeing temps hitting 100C using P95.

Removing P95 from the equation and sticking with IETU and AIDA I'm seeing maximum temps of 80C. That is a max though, generally they're sitting around 65-70.

Any thoughts?
 
I could have sworn I posted a reply to this thread earlier today but I must not have hit the "Post" button!

I have my system running at 4.3GHz with 1.28Vcore (ish) set manually with a bit of LLC set, I think about 25%.

I left it running AIDA64 overnight (8 hours) and saw maximum temps of about 83-87*C across the cores (I would have to rerun/check again to be exact seeing as I fluffed my earlier post).

During that time it seems that Vcore was ROCK solid locked at 1.28V so I feel we can be sure that there's no Adaptive Volts shenanigans going on.

What I want to know is why is Prime95 causing such ridiculously high readings??? Plenty of people seem to be running at reasonable temps...

Is my only remaining option to delid and sort out the TIM under the IHS?

EDIT:

Check out this post I found -

http://forums.anandtech.com/showpost.php?p=35104980&postcount=36

If the RAM speed makes such a difference to the temps then that could be why I am seeing such high readings. My RAM is at 2400MHz!!!

I am going to knock back to JEDEC 1333MHz and see what my temps are like then.
 
Last edited:
So...

I've just got home to test out the RAM speed vs. thermals lead from the above link.

For reference I was running my RAM at 2400MHz/1.65V and under stress (IETU/AIDA64) seeing temps typically hovering in the 60s and 70s with a maximum of 86*C.

I have just dropped my RAM to 1300MHz/1.5V and under the same loads I am seeing temps ranging from 54*C at the low end and maximum of 81*C.

I am going to do some more testing (and probably another overnight run).
 
That sounds about right to be honest. v28.1 p95 does crazy things to a 4770k and I'm guessing you have hyper threading on? Can turn that off fir testing purposes in my opinion. It won't help the stability testing in terms of validity and just adds heat maybe 3-5 degrees. I don't think you have a problem anymore and probably don't need to delid either unless you just feel like putting less stress on the system. Don't think you'll see anymore than 65c or so doing valley or playing games.
 
Yeah HT is on.

My question is, is Prime95 actually a valid tool for this generation of Intel CPUs? I see quite a bit of conflicting advice about this and it's tricky to know for sure.

Stability is one thing but if Prime95 is presenting situations that cannot be considered legitimate (for want of a better phrase) then is it sensible to use it as a (the) benchmark of stability?
 
I think it is absolutely a valid tool. Very few things adequately test AVX2. P95 28.1 for Haswell finds things at low FFT size that other stress tests will not find and it will cause low level stability issues that have nothing to do with heat in the real world. The issue I see with these errors on low FFT size is that they are usually northbridge related and dropping ram speed isn't really the answer so much as messing with the I/O offsets and maybe System Agent. And then what I'm trying to get across is that the errors are usually very quick to manifest once prime gets into low FFT ranges, which really push the heat and the IMC. You can just fast forward to these low fft ranges by using custom settings in prime instead of blend. I do not think "overnight" stability tests are necessary, although I'm all for a couple hours or whatever makes the end user feel confident.
 
Hey Siphon,

I have pretty much the exact setup as you minus the ram (I have 4x 8gb of some 2400 kingston hyperx) and I'm using the water 3.0 extreme AIO cooler.

Well I did some manually o/c as well since the onboard asus level up couldn't even load windows at 4.2ghz. The highest stable I have gotten so far is only 4.4. I have gotten 4.5 to load windows and do a lot at 1.39v/1.9v but within a minute of a CPU benchmark I get BSOD. I have not been able to get 4.5 stable so I think I have a lemon as well.

In reply to your topic, I cannot really compare apples to apples as I have not ran any good stress testers yet. I was using the Asus RealBench 2 on the encoding test just to do quick cpu stability tests. During this benchmark the hottest my 4 cores got to at the 4.4ghz was 81c. When I was attempting the 4.5 I did see temps go to 83c. Again - I'm sure this test is nothing compared to what you were doing, but your temps still sound a little high to me. I am very inexperienced in overclocking so don't take my opinion too far.
 
I'm getting similar results, with my setup systems in the sig. I have a thread going with pic if you wanna compare with your setting.
 
I just put my system together over the weekend and I'm running into the same heat issues you are. I'm adding a 360mm radiator to the system also.

If I run the Intel Stress Test or Prime95 it maxes out the CPU at 99-100. Otherwise it stays in the 50-70 range with everything else.

I ordered the "Naked Ivy" mounting set for my EK waterblock if I get the guts the delid the processor ;)
 
For reference, my chip only reaches those temps on Prime.

Both AIDA and Intel Extreme Tuning thingy are considerably lower than that for the majority of the time, perhaps with the odd peak.

However, Prime will see temps of 100*C almost instantly.
 
I figured out part of my problem, I have my Waterblock on the wrong way! I need to rotate it 180 degrees..the barbs are pointing at the video card instead of the memory! I'll pull it when I get my new Radiator in a few days.
 
glad im not the only one experiencing high temps. im running a single closed loop on a 4770k, MF6 as well. 4.2 ghz @ 1.3 vcore seems to be the only way I can get p95 stable for 8 hours. temperatures plateau at 80-85C on small FFT's. im running the latest version of p95.
 
I got my second radiator in and fixed the positioning of my waterblock last night...so far I've only maxed it out to 90C, so thats a nice 9-10c drop in temps I was getting before. I'll have to do more testing tonight.
 
You must log in or register to reply here.
Back
Top