Intel Core i7 4790K reaching 90C, When overclocks. Need Help!!!

[The N]

Hey Guys,

I have been observing some abnormal behavior on i7 4790K prcessor. It was all good a couple of days ago, but since yesterday, this processor producing an abnormal behavior with no clue why.

Let me get straight to the subject. When I overclock i7 4790K at 4.5GHz with 1.12vcore in bios, and put it on torture test of Prime 95 25.11v, it jump straight to 85C-90C within 3 minutes. On the other hand, it touches 70C at stock clocks. which also seems a bit higher to the previous 62C when it was behaving normal. I asked a buddy, he says, Chip problem. But I am sure if that's completely right.

Also, I tried to reseated the Cooler, Noctua NH-U14S, a couple of time, and re-apply thermal paste twice with dot and spread technique. Even, I tried a different Air cooler, too, but nothing has been changed.

Some screenshot of load temperatures:
i7 4790K @ 4.5GHz

i7 4790K @ 4.0GHz (stock)

My BIOS settings for overclocking are:
Multiplier x45 as Sync All Cores
Voltage 1.25v
DRAM 2400mhz (default voltage)
Turbo enabled (It get auto enable when increase the multiplier from stock)
EIST disabled (by default it is Enable)

All other options have been untouched.

Previously, when processor was doing normal, I overclocked i7 4790K at 4.5GHz and settings used were same as above. But now, it producing issue, no idea, why.

Here is teh temp screenshot for past overclocking being maxed out to 75C. (same system config)

System Config:
ASUS Maximus VI HERO
Core i7 4790K 4.0GHz (subject)
Gskill 8GB 2400mhz
Enermax REvolution XT 650W GOLd
Noctua NH-U14S Cooler
Noctua Nt-H1 Thermal Compound

Please help me identifying the issue, whether it is mobo or cpu, creating problem.


Regards,
The N

 

[Finners]

What time scale are we talking about between the old over clocked screenshot and the current one?

You could be using a different version of prime 95 that uses the avx extension to stress the CPU. This adds additional voltage to the chip so would explain the increased temps.

Also not related to your temps but your memory is only running at 1600mhz

 

[Folterknecht]

90°C with 1.25V under load with that CPU and cooler is normal (if not not delidded).

Edit: checking the screenshots on the first your running 1024FFTs, the second shows 8K - which makes a big difference in temperatures for every CPU

 

[The N]

@Finners
Chip was thermally showing extremely normal behavior, until yesterday, when I experienced this issue. The last screenshot on temp. is the proof of it, taken on past thursday.
All the tests, has been run conducted on Prime95 version 25.11 and blend mode used across the testing. I also tried Realbench Stress test, same results there, too.

@Folterknecht
No, not normal, bro. Please check the last screenshot being taken on past thursday during prime95 torture test, same blend being used. Also, the blend test increase the size as it progress.

 

[jboydgolfer]

it reaching high temps on a 500mhz OC running prime isnt too surprising imo. when i run P95 on my 4790 it reaches pretty high, but your on air. it looks like you have a single core reaching higher temps (which isnt uncommon). It could be due to whatever ihs compound intel used on those? But that one cores temp is what i would consider as too much higher.

 

[cucker tarlson]

Either replace the cooler with a push-pull dual tower or get a 5775c. Faster and waaay cooler for games.

sorry, 5775c not an option for z87.

my suggestion is simply not to run prime95. I haven't touched prime on my last two cpus (4790k,5775c) and they ran rock stable overclocks for over 3 years combined now. game stable is enough, to check the stablility and performance improvement I suggest dropping the games to lower resolution and detail to really stress the cpu and see how much returns the oc actually brings.

also, I noticed huge temp amplitude between cores. way too high. I experienced sth similar in the past and the issue was the thermal paste not spread evenly.


your 4790K 1.26v only reaching 70-ish degrees in p95 on your cooler makes me think it was either power throttling or it was some lighter test in p95. power throttling can and will occur even if clocks are reported correctly. you may see 4600mhz in cpu-z but the cpu is throttling, temps don't skyrocket as much and performance suffers a lot.

 

[rtwjunkie]

No, not normal, bro. Please check the last screenshot being taken on past thursday during prime95 torture test, same blend being used. Also, the blend test increase the size as it progress.

I think what @Folterknecht is saying is not normal, but that is is within safe limits.

If your CPU is not throttling, then it is still operating within its design parameters while running an extreme test.

 

[Toothless]

It runs hot, it's normal. I'd stop using prime95 and just use it as it is. I'm running 4.6 on 1.25v and as long as I'm not dumb and stress test it all the time I won't have thermal issues.

It's like there is a trend here.

Either replace the cooler with a push-pull dual tower or get a 5775c. Faster and waaay cooler for games.

sorry, 5775c not an option for z87.

my suggestion is simply not to run prime95. I haven't touched prime on my last two cpus (4790k,5775c) and they ran rock stable overclocks for over 3 years combined now. game stable is enough, to check the stablility and performance improvement I suggest dropping the games to lower resolution and detail to really stress the cpu and see how much returns the oc actually brings.

also, I noticed huge temp amplitude between cores. way too high. I experienced sth similar in the past and the issue was the thermal paste not spread evenly.


your 4790K 1.26v only reaching 70-ish degrees in p95 on your cooler makes me think it was either power throttling or it was some lighter test in p95. power throttling can and will occur even if clocks are reported correctly. you may see 4600mhz in cpu-z but the cpu is throttling, temps don't skyrocket as much and performance suffers a lot.

You need to chill with the 5775c advertising. It's not easily available and is too expensive of a sidegrade for a thermal "fix" to a 4790k.

 

[hat]

Maybe the thermal compound between the CPU die and heatspreader is wearing out. Might be time for a delid, if you're comfortable doing so.

 

[Tom_]

Prime 95 is oberheating Haswell CPUs even without overclocking.
Later CPUs are lowering the CPU-Clock, when AVX-Instructions are used.
On Haswell CPUs the SSE/AVX-Unit is running at full Speed.
I know that, because I also have the 4790K.

 

[GorbazTheDragon]

Ok, same cooler as me, similar voltage/clocks.

Power wise you are getting around 90w overclocked, for reference what I get at 4.6/1.25 is around 95-105w using XTU. That would lead me to believe the processor is behaving normally in terms of its power throughput.

At those power levels I usually have the package temp hovering in the mid 80s and the core temps around 70-75c with one of the cores ~80c, above the others. I do however have the delid (which dropped my temps by around 7-8c and my system (heatsink included) was recently completely cleaned of dust, so the cooling might be slightly better than what you are getting.

Unfortunately the older screenshot doesn't show the power draw or the p95 version, so it's difficult to tell whether the processor is suddenly drawing more power (due to newer p95 version possibly) or is just cooling worse for one reason or another.

I would suggest looking at whether you need to clean the dust out of your system, that may be the culprit, but honestly the power/temp values look like more or less what I'd expect from these chips... They just run hot.

Edit: Just did some testing with p95 25.11 and tuning my OC settings, 4.6/1.245 gives me ~115w power on the VRM output, package temp sits steady at 88c, core temps are 86 on highest and 75 on lowest. Ambient is 21c. I'd guess this is a little cooler than what you are getting, but not by much. Enough to make me believe the delid and ambient temps can make the difference. Also, might be a chance that the TIM is degrading on your chip. It might be wise to compare your VRM output current/power too, as this directly determines the temperatures given all else constant.

 

[The N]

Ok, same cooler as me, similar voltage/clocks.

Power wise you are getting around 90w overclocked, for reference what I get at 4.6/1.25 is around 95-105w using XTU. That would lead me to believe the processor is behaving normally in terms of its power throughput.

At those power levels I usually have the package temp hovering in the mid 80s and the core temps around 70-75c with one of the cores ~80c, above the others. I do however have the delid (which dropped my temps by around 7-8c and my system (heatsink included) was recently completely cleaned of dust, so the cooling might be slightly better than what you are getting.

Unfortunately the older screenshot doesn't show the power draw or the p95 version, so it's difficult to tell whether the processor is suddenly drawing more power (due to newer p95 version possibly) or is just cooling worse for one reason or another.

I would suggest looking at whether you need to clean the dust out of your system, that may be the culprit, but honestly the power/temp values look like more or less what I'd expect from these chips... They just run hot.

Edit: Just did some testing with p95 25.11 and tuning my OC settings, 4.6/1.245 gives me ~115w power on the VRM output, package temp sits steady at 88c, core temps are 86 on highest and 75 on lowest. Ambient is 21c. I'd guess this is a little cooler than what you are getting, but not by much. Enough to make me believe the delid and ambient temps can make the difference. Also, might be a chance that the TIM is degrading on your chip. It might be wise to compare your VRM output current/power too, as this directly determines the temperatures given all else constant.

Hello, thanks for your detailed and handful of information.

I have been using Prim95 25.11v thoughout the testing. So previous testing and latest testing all are same, as far as the program version goes.
I don't think Noctua NH-U14S is low for i7 4790K, especially at those clocks. And, I already had tried dual tower SilverStone Heligon HE01 for comparison, but nothing changed. Secondly, I think TIM is fine as wo different TIM units have been used of same Noctua NT-H1.

FYI, I am testing this CPU in an open environment as test bench; it rules out any dust/cleaning factor.

Looking at your temperatures, I can say, my i7 4790K's thermal results are not that bad. On the other hand, I have ran AIDA64 'Stress test', not FPU. The max temperature reached at 72C, which leads me to believe the processor is actually good at gaming load. But, when it would come to torturing through Prime95 and Realbench, this processor will not be able to maintain an ideal temperatures.

Could motherboard be giving in more voltages despite manual settings?

Also, One thing I did notice while overclocking i7 4790K is, when I increases the multiplier in BIOS, let's say x45, the Turbo mode under 'CPU Power Management' switched to Enabled. While At 4.0GHz it is Disabled. Maybe if that is messing into something.

Here is teh BIOS screenshot
https://imgur.com/a/wWsHA

 

[cucker tarlson]

Like I said, your temps now and before suggest there was power throttling before and now there isn't. That's good cause power throttling can impede your performance in games and other stuff. My CPU at 4.2GHz OC has lower framerate than it has at stock 3.6GHz until I turn off the power limit. It shows 4.2GHz and great temps but the performance just isn't there. Once the limit is removed it is 12% faster from what I measured in CPU intensive games.

 

[GorbazTheDragon]

I wouldn't run my CPU into the 90s under regular load, even high 80s is stretching it, but like you say, there's a big difference between synthetic and gaming loads, in which I usually get to around 45-55c. And even when I am running encoding tasks I barely see my setup hit 70c.

I'm not really too fond of the BIOS overclocking for everyday stuff, at least on my board it always ends up with the power saving voltage modes off, so the chip is always running at the OC voltage even when idle and downclocked. I am currently using XTU to force the adaptive voltages and manage the clock speeds after I boot up under "safe" settings (4.4@1.2v).

This turbo mode you mention appears to be related to Intel Speedstep, which manages the downclocking under low loads, here's a thread I found on it: https://rog.asus.com/forum/archive/index.php/t-54999.html

If I had to guess, it's an extension of Speedstep so that it functions when the clocks are above the stock boost levels, so it's something you want on as long as you are not doing bench runs.

 

[cucker tarlson]

Yup, I was a die-hard BIOS overclocker, but I was recently converted to using XTU. So much easier. You just go between profiles with two clicks. Default,overclocked,undervolted, so easy to set up an then switch instantly. I had some audio crackling with XTU on win 8.1 but I don't get any on win 10, so it's perfect now.

 

[eidairaman1]

Make sure a plastic cover is not on the bottom of the hsf.

 

[Kursah]

Is the processor under warranty or did you purchase the performance tuning plan? If so, you might consider RMA. Otherwise you might consider delidding if you want to run cooler.

What else are you using beyond Prime to test?

My good 4790K even benefited by a reduction in temps between 12-15C...now where it used to run the mid-70s on my U14S + NT-H1, running at my daily OC of 4.8 @ 1.26v, it now runs right around 60 or so under load. I know someone said that 4790K's run up to 90C under a U14S, but that hasn't been my experience. 70s-80s have been though. It is a very good cooler, single or dual fan.

My not as good 4790K also benefited equally from delidding. Running stock in my server under a CM212EVO + ACMX4, now loads up to the mid 60's where before it would reach 80C. It has always ran hot, had a lot less TIM than my better CPU under the lid...was also a lot easier and quicker to clean. I also re-glued the lids of both CPU's down with a thin layer of RTV sealant, worked great.

I did the razor blade + heat gun removal method, got a tube of Cool Laboratory Liquid Ultra for pretty cheap, was all easy enough to do with patience and heat. Heat the CPU really good to soften the glue holding the IHS on, warm the tube of CLU to make it easier to manipulate. If in doubt of that method, consider more effective cooling...which may or may not help if there's another issue going on.

But that being said, I've seen plenty of Haswell and DC chips at stock and OC'd configs running up to the throttle limit of 100C for years, still going strong today. They throttle very effectively and try to maintain the highest clocks within the thermal threshold. It is actually works really well, most folks never see it because they keep their temps down. I, like many others prefer to have my CPU's run cooler, but I'm also not going to worry too much if I tag 80 or 90C either. These CPU's seem to handle it quite well, and considering I have seen more hit those temps at totally stock settings with stock coolers, Intel must've felt pretty confident in that too.

Though as said above, gaming loads don't usually drive CPU's as-hard-as a dedicated stress test. If you're running cool in your daily and dedicated uses for that system I wouldn't worry so much. We all have preferences on the cooling front, hence why I run a U14S, EVO, and have delidded (finally) my CPU's.

 

[GorbazTheDragon]

@Kursah What kind of power figures are you getting for those temps? Makes it easier to compare the cooling performance.

I'm just using MX-4 paste under the IHS and I still have them loose. I might mess around at some point with re-seating it because I'm running 2 cores at ~12c lower than the others. The lower cores are in mid 70s.

Lucky you though, I can't get 4.7 at 1.27v any more on my chip. It needs 1.33 for 4.8 at which point I can't run artificial loads any more because of the temps.

Agree with the rest and whatnot.

 

[OneMoar]

its pretty common knowledge not to run prime on haswell and up, use linpack non-avx with a medium dataset

NH-U14S isn't particularly spectacular especially with a single fan
nothing surprising here

 

[Kursah]

@Kursah What kind of power figures are you getting for those temps? Makes it easier to compare the cooling performance.

I'm just using MX-4 paste under the IHS and I still have them loose. I might mess around at some point with re-seating it because I'm running 2 cores at ~12c lower than the others. The lower cores are in mid 70s.

Lucky you though, I can't get 4.7 at 1.27v any more on my chip. It needs 1.33 for 4.8 at which point I can't run artificial loads any more because of the temps.

Agree with the rest and whatnot.

In heavier games like Ashes, Crysis 3, I may see around 370W consumption from the wall as per my APC UPS...though I may have to double check that figure as the PC isn't the only device on the battery circuit. But around 400W consumption for an OC'd 4790K and a stock 980Ti, which I believe hits around 250W under heavy loads on its own...plus all the other overhead, I could see that figure being about right.

HWiNFO64 reports that max load was at 121.571W for the CPU.

I was lucky with this CPU, my first 4790K I bought used for a steal. It never OC'd worth a damn and ran hot at stock. Made a great home server CPU for the past couple of years now. I have been tempted to try OC-ing it again to see if I can get any more out of it...but honestly I've been so content with my good one that I haven't bothered. I also haven't tried to push further than 4.8 since delidding.

 

[GorbazTheDragon]

OK, yeah was asking about the VRM/CPU reported load, I measured around 120W out of the VRM on my board for those temps. I guess I'll have to try a reseat in that case, might shave a few degrees off those high cores, but it doesn't seem outrageously bad as it is.

Like I mentioned, daily load sits under 60c so I don't see it being problematic in any way temperature wise, I just don't want to push more volts through the chip because I need it to last right now...

 

[EarthDog]

Get a 5775c?

Dont run prime??

Wont run this cpu in the 80s???




Come on tpu....

 

[The N]

Is the processor under warranty or did you purchase the performance tuning plan? If so, you might consider RMA. Otherwise you might consider delidding if you want to run cooler.

What else are you using beyond Prime to test?

My good 4790K even benefited by a reduction in temps between 12-15C...now where it used to run the mid-70s on my U14S + NT-H1, running at my daily OC of 4.8 @ 1.26v, it now runs right around 60 or so under load. I know someone said that 4790K's run up to 90C under a U14S, but that hasn't been my experience. 70s-80s have been though. It is a very good cooler, single or dual fan.

My not as good 4790K also benefited equally from delidding. Running stock in my server under a CM212EVO + ACMX4, now loads up to the mid 60's where before it would reach 80C. It has always ran hot, had a lot less TIM than my better CPU under the lid...was also a lot easier and quicker to clean. I also re-glued the lids of both CPU's down with a thin layer of RTV sealant, worked great.

I did the razor blade + heat gun removal method, got a tube of Cool Laboratory Liquid Ultra for pretty cheap, was all easy enough to do with patience and heat. Heat the CPU really good to soften the glue holding the IHS on, warm the tube of CLU to make it easier to manipulate. If in doubt of that method, consider more effective cooling...which may or may not help if there's another issue going on.

But that being said, I've seen plenty of Haswell and DC chips at stock and OC'd configs running up to the throttle limit of 100C for years, still going strong today. They throttle very effectively and try to maintain the highest clocks within the thermal threshold. It is actually works really well, most folks never see it because they keep their temps down. I, like many others prefer to have my CPU's run cooler, but I'm also not going to worry too much if I tag 80 or 90C either. These CPU's seem to handle it quite well, and considering I have seen more hit those temps at totally stock settings with stock coolers, Intel must've felt pretty confident in that too.

Though as said above, gaming loads don't usually drive CPU's as-hard-as a dedicated stress test. If you're running cool in your daily and dedicated uses for that system I wouldn't worry so much. We all have preferences on the cooling front, hence why I run a U14S, EVO, and have delidded (finally) my CPU's.

Beyond prime? I considered FPU, but it takes real hard on cores. Almost 100C touch upon initiations.
So AIDA64's FPU is out of context here.

For everyday use, at x45 multiplier, it hardly touches 70c-75c. But at dedicated stress tests like Prime95 and FPU, temperatures goes beyond 85C.

I wouldn't have cared about stress temps but since I am doing testing for review, I would need to settle down this issue to come with final results. Delidding is not an option for me.
With Noctua NH-U14S, CPU Overclocked 4.5GHz @1.250v, it reaches up to 85c - 88C very quickly, hardly completes 10 minutes test. But reseating do some variations as well.

The variation is like this:
Noctua NH-U14S sometimes give 62c, but sometimes with reseated, it reaches up to 71c-72C; core0 always 8-10c higher than other cores.

Maybe I am not very lucky with the chip, as the behavior is deteriorative atm.

 

[Toothless]

I'd offer to test it for you since I have a couple of boards that OC pretty well and a couple coolers that are pretty good but you're a tad far away.

 

[R00kie]

Try and lie the PC down on its side, and see if that improves it a bit, if it does, youve got some problems with your cooler having bad contact with the CPU.