Acceptable CPU Temps & Stress Tests

nomido_27 · 28 Feb 2014 at 08:09

Hi all,

Hoping you can advise on CPU temps & stress tests.

My new build PC was running handbrake and crashed out after 10mins (it was at 100% CPU). I then ran Prime95 and the system crashed again after 10 mins giving this time the windows blue screen error.

Theres a lot of mixed opinions out there so looking for some advice on the following:

When fitting the H100i (which comes with paste already on it) I'm not sure whether I was supposed to have cleaned the CPU first (as it was stressed tested before receiving it)?

Also I had to re-adjust it a couple times so not sure if that caused the paste to be uneven.

QUESTIONS
1. Should the CPU be capable of running at 100% without crashing?
2. What should the ideal idle CPU Temperature be?
3. What should the max CPU temp be under load (100%)?
4. Am I better off to leave my H100i cooler as is or clean it and replace with new thermal paste? (Bit of a n00b so don't want to make it worse - see further info below)?

FURTHER INFO - SYSTEM
ASUS Maximus VI Hero
i7-4770K (OC'd to 4.4 by overclockers.co.uk)
16GB TeamAir RAM
Corsair H100i cooling
Corsair Carbide 500R case
Corsair AX760 PSU
AMD Radeon HD7700

FURTHER INFO - TEMPS
CPU Idle: 40 degrees celsius
CPU 100%: 62-76

OTHER CONSIDERATIONS
1. How much reduction (roughly degrees Celsius) could changing the paste make?
2. I've read references to Voltage, can anyone give me the laymens on what it is and what to look out for?
3. Any tips to help diagnose what's crashing it?

Many thanks

gjc10212 · 28 Feb 2014 at 08:19

Can you post a screen shot of your BIOS so we can see the volts? Failing that write them down....

hawkhench · 28 Feb 2014 at 08:52

Those temps look about standard for that chip, and shouldn't be causing your shutdowns. As gjc10212 said your voltages seem more likely to be at fault.

nomido_27 · 28 Feb 2014 at 09:25

Thanks for the replies.

Is the voltage something that's set on the CPU or the PSU?

How do I control it and decide what to set it to?

I'll post up a screenshot/the details asap this evening :-)

DoneADougalOnSofa · 28 Feb 2014 at 09:33

I had crashes at 4.4 with my 4770K.
Setting 'System agent voltage' to 1.250v (or a +0.250v offset, mobo dependent...) worked for me. No crashes since.
This is apparently more important if you have all four Ram slots occupied rather than just two.
Maybe worth a pop?

EDIT Do you have the BSOD error code?

nomido_27 · 28 Feb 2014 at 09:43

Hi, thanks for the post; ill look into the System Agent Voltage/offsetting.

I'm currently using 2 of the 4 RAM slots. The BSOD was not the standard "old" style, it was one of those new windows 'friendly' blue screen which said an error had occurred, when I went to details it mentioned a memory dump. Does that help or are more specific details available/required?

mmj_uk · 28 Feb 2014 at 12:45

nomido_27 said:
1. Should the CPU be capable of running at 100% without crashing?

Yes, some people will argue otherwise but they're just trying to justify unstable overclocks. AMD have a tendency to throttle under 100% load but that's more down to their utterly insane power draw/temperatures which they've had to push to stay competitive. AMD have decided that to get more speed they would place artificial limits on how much can be done based on their opinion of what is normal.

All Intel CPU's will run maxed out at stock with no problems, if you choose to run it at a speed which is not 100% stable that's up to you...

nomido_27 · 28 Feb 2014 at 15:22

Thanks mmj_uk, I spoke to Tech Support and they believe its the vCore setting (as it passed their stress test fine) and of course I'm using a different PSU.

Any advice on how to know what level to set it to?

Purgatory · 28 Feb 2014 at 15:48

mmj_uk said:
Yes, some people will argue otherwise but they're just trying to justify unstable overclocks.

100% agree, an unstable overclock is useless if you can't run your PC 24/7 at 100% load and you are just asking for trouble if your overclock is unstable, it can cause things like corruption of your hard drives and the loss of all your data in time.

DoneADougalOnSofa · 28 Feb 2014 at 15:52

Best way is to up it by 0.005 or 0.01v and try again- so if your vcore reads 1.2v, set it for 1.21v and have another go. Or you could try a positive offset (of 0.01v) instead, either way really.
If that doesn't do it, try 1.215v or 1.22v, and so on.
I think (brain fade) mine runs at 1.24v or thereabouts. Above numbers are for illustrative purposes only!

nomido_27 · 28 Feb 2014 at 16:28

That's really helpful dougal1331; ill give that a try and report back this evening.

nomido_27 · 28 Feb 2014 at 16:30

One follow up question...

If the vCore/Offset change allows Prime95 to run the CPU at 100%; is it safe to let that run overnight/24hrs even of the temp is 65-75?

DoneADougalOnSofa · 28 Feb 2014 at 17:05

Yes that is fine. Haswells run pretty warm anyway, you're still in the safe zone even at 75 so I don't think temps are an issue in your case. If they were mid-eighties, then that would be pushing it a bit...
Bear in mind that Prime95 and other benchmarkers stress your rig more than general usage- you might never get a BSOD during 'normal' use. (But I accept that that is not the point of being a Overclocker!).
If I may add: usually considered best to go for direct settings, rather than offsets, where possible. My last mobo was el cheapo, and offset was my only option...

Sayso · 28 Feb 2014 at 17:06

yes, that temp is fine.

nomido_27 · 28 Feb 2014 at 17:47

Much appreciated; ill factor all your comments in when making the changes.

joeyjojo · 28 Feb 2014 at 18:15

I'd suggest using Intel Burn Test (or Linx). Setting the memory to maximum is usually best, and do 3-5 loops for a quick test, and 10-100 for a long test.

It usually finds errors faster than Prime in my experience.

nomido_27 · 28 Feb 2014 at 20:00

hi joeyjojo, that sounds good i'll try that - plus i wont have to mess around with ISOs as with using MemTest86.

All, as recommended by gjc10212, please find the below link showing my current voltages.

http://postimg.org/image/nih7bjup1/

It'd much appreciated if you could confirm:

1) Everything looks ok?
2) Which voltage it is that I need to increase to resolve the issues mentioned above?

Many thanks.

alex_123_fra · 28 Feb 2014 at 20:25

I wouldn't use prime or Intel burn test for any significant period of time. I use cinebench, asus rog realbench and normal gaming. You can even do some handbrake encoding if that is the type of use you anticipate.

gjc10212 · 28 Feb 2014 at 20:35

nomido_27 said:
hi joeyjojo, that sounds good i'll try that - plus i wont have to mess around with ISOs as with using MemTest86.

All, as recommended by gjc10212, please find the below link showing my current voltages.

http://postimg.org/image/nih7bjup1/

It'd much appreciated if you could confirm:

1) Everything looks ok?
2) Which voltage it is that I need to increase to resolve the issues mentioned above?

Many thanks.

In Extreme Tweaker change the CPU Core Voltage Overide to increase the voltage. Monitor your temps with each increase, and only increase in small increments at a time.

Can you post a screen shot similar to mine?

nomido_27 · 28 Feb 2014 at 20:48

Alex_123_FRA - like the idea of running handbrake as an additional test (seeing that's one of the main functions I'll be using it for) & pushes the CPU to 100%.

Cheers for the post gjc10212, ill give that a shot and report back...