Very high temps under load (I5 3570K)

nitroxc

Honorable
Oct 7, 2013
32
0
10,530
So I've had my I5 3570K for a while now and I am looking to overclock it but I am encountering a big problem.

Currently my case has pretty good airflow with 2 120mm intake fans in the front and a 120mm exhaust fan in the rear. Cooling the cpu is the hyper 212 EVO using the stock thermal paste (was about 2 months old when I applied it)

With all this running intel burn test my temps are all over the place. The lowest it goes is around 60c and it is always jumping up to 85 degrees. I know for a fact 80c on a stock clock with a hyper 212 is way too hot.

Idle it sits around low-mid 40c's which I also know is really high because with the same case/cooler I was able to get around 25c on my old fx 8120.

So my question is, what should I do? Shall I buy some AS5 and see if replacing the thermal paste and reseating it helps? I am looking to overclock to 4.5ghz and I am interested in buying a 240mm AIO. Any suggestions?
 
I know how to do it, not really sure if I want to risk it though

Do you think it could be an issue with the thermal paste? When I applied the heatsink the thermal paste I was using (that came with the heatsink) had been opened once before, shut and then left for like 2 months before I used it again.
 


You're about 10C above on both idle and load. I would reseat it.
And the evo should be good enough for a 4.5 gig overclock on that cpu, just get some quality thermal paste.
 
Really? I've heard it's only really okay for slight overclocks? Also I think I will get a 240mm AIO anyways with some noctua fans because the stock fan on the hyper 212 is a bit too loud for my liking when its running at 1500rpm
 


Well 4.5 is not a high overclock on that CPU. Some of those chips are able to do 4.4 on stock voltage, witch doesn't produce that much more heat than stock.
 
Credentials: i5 3570K owner

Ivy Bridge and Haswell CPUs use TIM instead of solder between the IHS and the die. This isn't so much a problem, but when Intel mounts the headspreaders, it usually doesn't seat completely onto the die, leaving a small gap between the two which is bridged only by paste, which causes high temperatures. This can be "fixed" by removing the headspreader, scraping off the glue that holds the IHS up and prevents it from sitting flush with the die, and then putting it back on.

80c is, however, safe and normal for this CPU. It will happily run at 90c+ for years, especially at stock speeds. IBT also puts an unusual and unrealistic load on the CPU, which you are not likely to see in any real world situation. I'll agree that this seems a bit high for a Hyper 212 at stock clocks, but it's very possible that regardless of what cooler you use, you'll not see significantly lower temperatures unless you remove the lid from your CPU (as I did). I would not bother with an AiO watercooler (these CPUs don't actually make that much heat), it's pretty much delid or leave it along.

IMG_20120929_235041_zpssb3hqpm4.jpg


 


I disagree. 4.5-4.6ghz is pretty much the limit on Ivy Bridge. Some Sandy Bridge CPUs were able to go higher, but I would not expect more than 4.4-4.5 on a 3570K without delidding.

Although many don't need to increase voltage offset with a 4.4ghz OC, voltage still goes up because of Intel's voltage table, which scales with clockspeed.
 
I was messing around with the voltages a while ago, and I managed to get it stable at 1.8v at 4.2ghz however I took it off because 93c really werent safe. Also I know IBT puts alot more stress than usual on the cpu but I know for a fact at stock voltages with a hyper 212 pretty much no cpu should hit 80c. Hell my 8core amd fx cpus (some of the hottest running cpus) with the same setup was barely hitting 60c with a 4ghz overclock.

I'll order some new thermal paste tonight and apply that, if it reduces the temps then good. If it doesn't, I might look at delidding it.
 
1.8v?

Voltage should be in the range of 1.0-1.2v for those clocks, there's no way 1.8v is correct.

Food for thought: You can't compare temperatures on an AMD CPU to those of an Intel CPU, because AMD's temperature sensor is not placed as close to the hottest parts of the CPU. "90c" as reported on an Intel CPU may actually be cooler than "60c" on an FX.
 
not 1.8v sorry, 1.08v. Although I did just test it and i put boost to 4.2ghz and It wouldn't boot until 1.2v which is strange because I remember running it at 4.2ghz and 1.08v when i first got it.
 
The minimum stable voltages for my CPU at various clockspeeds:

3400mhz - 0.984v
3600mhz - 1.016v
4000mhz - 1.088v
4100mhz - 1.120v
4200mhz - 1.144v
4300mhz - 1.200v
4400mhz - 1.248v
4500mhz - 1.320v
4600mhz - 1.380v
4700mhz - 1.450v (approximately)
4800mhz - not stable with any voltage

OGdTJ0F.png


Your mileage may vary.
 


Exactly.

A modern chip can handle in excess of 100 degrees.

It isn't recommended but 80 wouldn't bother me on a burn test.

As said above, reseating it is a good idea and that costs you nothing (if you have a bit of paste lying around)

 
Yes I know 80c is fine for the cpu but the thing is it shouldn't be nowhere near that high on stock with custom air cooler. I'm gonna order some good thermal paste and try reseating it, if it doesn't drop the temps then ill look elsewhere.
 
Right my MX-2 finally arrived today and I reseated the cooler with new thermal paste. Before I reseated I did an IBT can got 85c. After I reseated I did the test twice, first time getting 80c (5c impovement) and second time I got 82c.

So it seems that I have simply lost the silicon lottery and got a hot cpu 🙁

I'm gonna go ahead and assume that the best thing I can do now is delidding it? I am tempted to do this but first have a question or two. I know that coolaboratory's liquid metal TIM is one of the best for delidding but I'm wondering is it best to just take off the IHS and put new TIM on then put the IHS back on or is it best to just do a direct contact? Also, if I do direct contact on the cpu die with the cooler, would I need some new fitting hardware due to the height difference of the cpu's surface? Or is it a case of just tightening the cooler down a little bit more?
 


Because its still really hot, I just ran an IBT today right, its a bit hotter than normal outside (still only 23c) and I had to stop it after 5 seconds because on the stock speeds one of my cores spiked to 102c and the other 3 ranged from 92-98.