I5-4670k instantly insanely high temps C90+ with prime95 at only 4.0Ghz. Hyper 212 Evo.

[kosanovskiy]

Hello so I'm starting to get worried a little about my temps and my CPU in general. So I just build my new rig I5-4670k, MSI GTX 970 Gaming 4G, 16GB 2133 G.Skill Ram (1.60V). So I started over clocking using the Asus Dual Intelligent Processors 5 and on that is says my temps are 35C while Real Temp GT 3.70 shows they are much higher being at 45-50C. My PC does a max over clock at 4.4GHz and the prime95 torture test puts this at 100C (which is kind of sad) and when I over clock ti no matter what ratio I choose once I get past 3.8GHz my temps go past 85C. As you can see even 4.0GHZ under load and at 2.5%, 18% and 99.8% load temps are weirdly extremely high. I'm running a air cooled system consisting of 5 noctua NF-P14 FLX fans and the CPU cooler being Cooler Master Hyper 212 Evo. Lookign at everyone else temps mine are horrible and I dont know why. I already atempted to re-seat it and re attach the whole assembly completely appliyign new termal past that came with the CM CPU Cooler and nearly all out of it after so many re-dos an and still am unable to find what is wrong and why are my temps so high.

I did jack it up little by little from 3.4 to 4.6 and my stable max is 4.5 then a crash at 4.6, but i never dare to go above 4.4 since the temps are all the same after 3.8GHz. 3.8 is where I get below 85C at 100% load with prime95. My voltage is 1.195 for anything from 3.7-4.3, just trying to run 4.0 since it seems like a good in between point. I do play games GTA V Maxed everything and temps around 75-80C on my 2560x1440 monitor. Which is why I'm wondering is the torture test is legitimate because I can't think of a more complex thing to do on this machine but protein folding.

I have 2 in the front as intake, and then 2 top and 1 back as exhaust. Very good setup personally.

Did anyone else ever run into such stuff? Or do I just have a defective cooler AND a shirty CPU?

Here are the pics.


Asus DIP5 info vs RTGT 3.7

Real Temp GT 3.70 99.8% Load

Real Temp GT 3.70 18% Load

Real Temp GT 3.70 2.5% Load

My Specs:
CPU: i5-4670k
GPU: 2x MSI GTX 970 Gaming 4G (SLI)
Mobo: Asus Z97-A
RAM: 16Gb Ripjaws 2133 Cas9 1.6V
CPU Cooler: Hyper 212 EVO
PSU: EVGA Supernova G2 750 80+ Gold
Case Fans: 5x Noctua NF-P14 FLX
SSD: 256Gb 850 Pro, 500Gb 850 Evo
Case: Corsair 300R

 

[MarkW]

Not all CPU's can overclock as high as others. Personally, anytime that temp gets over 90c, I power off. IMMEDIATELY!

It appears that you jacked up the power, and are surprised that the chip got hot. More Power = More Heat.

You may very well have a weak CPU when it comes to overclocking. Every chip is different. But one thing you flat out do NOT do when overclocking is jack it up to what you saw someone achieve for an overclock all in one step. NEVER EVER DO THAT! You literally do baby steps. Move up a little, stress temp it some, and if there are no problems, go up a little higher, and stress test again. When you finally reach the point that there is a problem, step that overclock back down at least 1 step, and I actually go back 2 steps, and then stress test it for hours. If that passes, then play some games for an evening, and if all of that goes well, and the CPU temps are still below 85c under full load, then you know where that chip can go.

Lower that overclock some to get those temps back down to reasonable numbers.

And what in the world are you doing with 5 Noctua fans? 1 or 2 go in the front to pull cool air in. One goes in the rear to pull the warm air out. And then what are you doing with the other 2 or 3 fans?

 

[i7Baby]

"once I get past 3.8GHz my temps go past 85C." - then don't go past 3.8GHz.

RealTemp shows cpu core temps - which is 5C hotter than CPU (case) temps.

Not all same CPUs OC the same.

Only use Prime 95 v26.6 (or OCCT) for Haswell CPUs Other versions are too hard on the CPU and ram.

 

[kosanovskiy]

Not all CPU's can overclock as high as others. Personally, anytime that temp gets over 90c, I power off. IMMEDIATELY!

It appears that you jacked up the power, and are surprised that the chip got hot. More Power = More Heat.

You may very well have a weak CPU when it comes to overclocking. Every chip is different. But one thing you flat out do NOT do when overclocking is jack it up to what you saw someone achieve for an overclock all in one step. NEVER EVER DO THAT! You literally do baby steps. Move up a little, stress temp it some, and if there are no problems, go up a little higher, and stress test again. When you finally reach the point that there is a problem, step that overclock back down at least 1 step, and I actually go back 2 steps, and then stress test it for hours. If that passes, then play some games for an evening, and if all of that goes well, and the CPU temps are still below 85c under full load, then you know where that chip can go.

Lower that overclock some to get those temps back down to reasonable numbers.

And what in the world are you doing with 5 Noctua fans? 1 or 2 go in the front to pull cool air in. One goes in the rear to pull the warm air out. And then what are you doing with the other 2 or 3 fans? My guess is that you are literally destroying your airflow with them...

I did jack it up little by little and my max is 4.5 then a crash at 4.6, but i never dare to go above 4.4 since the temps are all the same after 3.8GHz. 3.8 is where I get below 85C at 100% load with prime95. My voltage is 1.195 for anything from 3.7-4.3, just trying to run 4.0 since it seems like a good in between point. I do play games GTA V Maxed everything and temps around 75-80C on my 2560x1440 monitor. Which is why I'm wondering is the torture test is legitimate because I can't think of a more complex thing to do on this machine but protein folding.

I have 2 in the front as intake, and then 2 top and 1 back as exhaust. Very good setup personally.

 

[kosanovskiy]

"once I get past 3.8GHz my temps go past 85C." - then don't go past 3.8GHz.

RealTemp shows cpu core temps - which is 5C hotter than CPU (case) temps.

Not all same CPUs OC the same.

Only use Prime 95 v26.6 (or OCCT) for Haswell CPUs Other versions are too hard on the CPU and ram.

Wow, so that means I got a really really * CPU huh?

I was runing prime97v285. Shoudl I'll try again with 26.6.

 

[i7Baby]

Could be. Try 26.6 or OCCT.

 

[MarkW]

Well, you left all that out of your original post.

Maybe. You see, air flows like a river. It is intended to flow from front to back. And if you have the top fans changing that front to back airflow, you could very well be leaving pockets of warm/hot air around your CPU and video card. Run Prime95 for an hour with your current setup and monitor the CPU temps during that hour. Then shut down, disconnect the 2 top fans, and run the same test again. See if the CPU temps change at all.

 

[kosanovskiy]

Could be. Try 26.6 or OCCT.

WOW, with 26.6 I'm at 67C on 4.4GHz. WTF why is it so different from the one I used initially? Also I run the Small FFTs torture test.

 

[kosanovskiy]

Well, you left all that out of your original post.

Maybe. You see, air flows like a river. It is intended to flow from front to back. And if you have the top fans changing that front to back airflow, you could very well be leaving pockets of warm/hot air around your CPU and video card. Run Prime95 for an hour with your current setup and monitor the CPU temps during that hour. Then shut down, disconnect the 2 top fans, and run the same test again. See if the CPU temps change at all.

Sorry I'll edit that in. But it appears the Prime95v 28.5 shits on CPU temps. While 26.6 is only at 67C for me. I don't even know what to trust anymore. But yeah I perfected the airflow and tested that already and get lowest temps for GPU's this way since they are both OCd at 1505.

 

[photonboy]

Hi,

1) Go back to default settings for the CPU and retest. If in doubt restore the BIOS to default settings. Updating your BIOS if newer exists should do the same thing.

2) Now test again using "CORE TEMP" which should be pretty accurate. I personally use Prime 95, set to "BLEND" and make sure all four cores are used (can monitor with Task Manager-> Performance with it showing all graphs not just one.)

3) For assurance also run the CPU diagnostic from Intel.

4) If temps are still high we can investigate if you have a problem with your CPU cooler setup (paste, mounting or fan control). CASE cooling is also important but generally you don't need that much for a gaming PC. I have a front case fan for intake and one for exhaust and both run 700RPM which provides plenty of cooling for me (also GTX680 card).

*It will vary a depending on room temp etc, but I'm guessing no more than 70degC using Prime95 with Blend and that cooler and DEFAULT settings (which should be 3.9GHz on Turbo I believe). For GAMES I'd expect below 60degC.

**Running at 4GHz won't make much difference but does raise the temperature a lot if you mess with the voltage. Default Turbo is 3.8GHz and the "quick" overclock in the Asus BIOS would likely give you 4GHz anyway without messing with voltage so I'd advize doing that if that's not the method you already did.

One quick method is:
a) all back to default, then
b) choose "XMP", then
c) manually raise the core multiplier for each core to 40 (then test with MEMTEST and with the CPU tests to ensure stability). If unstable drop to 39. If stable and temps are okay try 41. You could go higher with voltage tweaks etc but I just don't think it's needed and you'd probably want a better cooler to start doing that anyway.

Links:
Prime95: http://files.extremeoverclocking.com/file.php?f=205

Intel Diagnostic: https://downloadcenter.intel.com/download/19792/Intel-Processor-Diagnostic-Tool-64-bit-

Core Temp (use individual core temps. should show four): http://files.extremeoverclocking.com/file.php?f=205

Memtest: www.memtest.org

Update:
The max temp for a core shown in Core Temp was 71degC for me. 23degC room temp. That's a different setup slightly with:
- Prime95 blend (after five minutes)
- Noctua NH-D14 (fan at 40% max)
- i7-3770K @ 4.2GHz (no voltage changes. only change was multiplier for each core raised slightly)

 

[kosanovskiy]

Hi,

1) Go back to default settings for the CPU and retest. If in doubt restore the BIOS to default settings. Updating your BIOS if newer exists should do the same thing.

2) Now test again using "CORE TEMP" which should be pretty accurate. I personally use Prime 95, set to "BLEND" and make sure all four cores are used (can monitor with Task Manager-> Performance with it showing all graphs not just one.)

3) For assurance also run the CPU diagnostic from Intel.

4) If temps are still high we can investigate if you have a problem with your CPU cooler setup (paste, mounting or fan control).

*It will vary a depending on room temp etc, but I'm guessing no more than 70degC using Prime95 with Blend and that cooler and DEFAULT settings (which should be 3.9GHz on Turbo I believe). For GAMES I'd expect below 60degC.

**Running at 4GHz won't make much difference but does raise the temperature a lot if you mess with the voltage. Default Turbo is 3.8GHz and the "quick" overclock in the Asus BIOS would likely give you 4GHz anyway without messing with voltage so I'd advize doing that if that's not the method you already did.

Links:
Prime95: http://files.extremeoverclocking.com/file.php?f=205

Intel Diagnostic: https://downloadcenter.intel.com/download/19792/Intel-Processor-Diagnostic-Tool-64-bit-

Core Temp (use individual core temps. should show four): http://files.extremeoverclocking.com/file.php?f=205

Update:
The max temp for a core shown in Core Temp was 71degC for me. That's a different setup slightly with:
- Prime95 blend (after five minutes)
- Noctua NH-D14 (fan at 40% max)
- i7-3770K @ 4.2GHz (no voltage changes. only change was multiplier for each core raised slightly)

Bios up to date, and already did retest at which my CPU is at 27C at 3.4GHz.

What version of Prime95 do you use? Because different versions give me diff temps. 26.6 imat 67 for 4.4GHz which 28.5 gives me 95-100C at the same 4.4GHz.

Runing the test right as I type this.

*Will update as soon as finished.

That is the method I use, and it automatically puts me at 4.2GHz max being 4.5 but 4.4 is most stable (excluding the temps). According to Prime95v28.5


Update: Everything Passed with the intel diagnostics.

 

[photonboy]

Update:

**First of all, your temperatures MAY BE CORRECT. We mainly need it at stock 3.8GHz Turbo settings and to know the room temp to give a more accurate guess.

Just so you're aware there are THREE ways to report CPU temperature (not counting incorrect values) which can account for differences between software though I know for a fact some are just incorrect (Speedfan is notorious for incorrect values):

#1 - sensor next to each one of the physical cores in the CPU (preferred method)

#2 - sensor on CPU circuit board

#3 - sensor on MOTHERBOARD next to CPU

Core Temp should show the individual core temps at the BOTTOM, though it looks like HWMonitor is correct from THIS post.
http://www.tomshardware.com/answers/id-2204524/4670k-hyper-212-evo-temperatures-high.html

You should stress the CPU then use the HIGHEST core temp reported once the stress program has been running a few minutes.

*The i5-4670K is known to run hot however which is why the Haswell Refresh was created (i5-4690K, i7-4790K..). If you're really running over 85degC it's not a problem though.

I noticed you talked about overclocking "from 3.4GHz" though with Turbo on (by default) you actually would be getting 3.8GHz automatically as needed. So again, manually going to 4GHz is in reality only a 5% overclock which as mentioned above could be done without screwing with voltage.

Voltage is the largest reason why temps raise. When I went from 4.2GHz to 4.4GHz I was required to raise the voltage slightly but that made over a 10degC difference. Going from 3.9GHz to 4.2Ghz without changing voltage made no noticeable difference.

 

[kosanovskiy]

Update:

**First of all, your temperatures MAY BE CORRECT. We mainly need it at stock 3.8GHz Turbo settings and to know the room temp to give a more accurate guess.

Just so you're aware there are THREE ways to report CPU temperature (not counting incorrect values) which can account for differences between software though I know for a fact some are just incorrect (Speedfan is notorious for incorrect values):

#1 - sensor next to each one of the physical cores in the CPU (preferred method)

#2 - sensor on CPU circuit board

#3 - sensor on MOTHERBOARD next to CPU

Core Temp should show the individual core temps at the BOTTOM, though it looks like HWMonitor is correct from THIS post.
http://www.tomshardware.com/answers/id-2204524/4670k-hyper-212-evo-temperatures-high.html

You should stress the CPU then use the HIGHEST core temp reported once the stress program has been running a few minutes.

*The i5-4670K is known to run hot however which is why the Haswell Refresh was created (i5-4690K, i7-4790K..). If you're really running over 85degC it's not a problem though.

I noticed you talked about overclocking "from 3.4GHz" though with Turbo on (by default) you actually would be getting 3.8GHz automatically as needed. So again, manually going to 4GHz is in reality only a 5% overclock which as mentioned above could be done without screwing with voltage.

Voltage is the largest reason why temps raise. When I went from 4.2GHz to 4.4GHz I was required to raise the voltage slightly but that made over a 10degC difference. Going from 3.9GHz to 4.2Ghz without changing voltage made no noticeable difference.

So are the real temps should be based on Prime95 version 28.5 then? But yeah 3.4 is the minimum I started manually after using the Asus Overclocking Utility thing just to see the changes.

Well that good to know. Pretty much take everything with a grain of salt. Running CPUID my temps are only 3C lower than Real Temp GT 3.70.