4690K high temps

Toggle sidebar Toggle sidebar
S

Shifty_Effect

Honorable
May 4, 2017
2
0
10,510
System specs:
OS: Windows 10 64-bit
CPU: i5-4690k (stock settings as best I can)
Motherboard: MSI G45 Gaming
GPU: MSI 290x 8GB
RAM: 16GB Corsair 1866
Cooler: Corsair H115i (Corsair H110i GTX before that)
PSU: EVGA SuperNOVA 1000 G1
Case: Corsair 750D
Fans: All 140mm, 2 front intake, 2 top intake (cooler pull config) and one exhaust.
BIOS: Updated to latest from MSI
Tools used for monitoring: HWMonitor, Intel ETU, and Corsair LINK.


I'll try to make a long story short, and I'll start with my old liquid cooler the Corsair H110i GTX. Installing the cooler was a hassle, the worst of it was how poorly the waterblock would sit on the CPU. The back plate was always loose and no matter how I screwed everything down the waterblock would wobble including the back plate itself. I read from Corsair that this was normal so it let it go. After initially completing my build my CPU temps didn't impressed me, always in the high 30s to mid 40s no matter how fast my fans or cooler pump was working. Never tried to overclock after running small stress tests as they would hit the 80s causing me to immediately stopping the test. Temps while gaming varied anywhere from 50°C - 70°C+. About a month ago I got tired of the temps as I really wanted to start learning to overclock and get better performance, so I reapplied thermal paste with Arctic MX-4. Reseated the cooler but this time I used rubber washers between the back plate and the motherboard which made the waterblock seal tight to the CPU without wobbling. I thought my temps would be normal but I was wrong. Temps were the same but the cooler fans would ramp constantly and the only way the temps would be decent is if I flicked the radiator tubes. I figured that the cooler was bad so I RMA'd it. Corsair "upgraded" me to the H115i cooler I am currently using. With the H115i I didn't need to use the rubber washers, waterblock seemed to seat just fine. My idle temps were the same as before but my stress test temps never went past 55°C. Progress was made, but that night I noticed one of the two fans I installed on the cooler was orientated wrong, I goofed and fixed it the next morning. I read online that I may not be giving the pump enough power so I got another SATA cable solely for the pump only. Everything went fine but now my temps are slightly higher during normal usage and the stress test show high temps.

Temps from stress testing are as follows:

Intel ETU Stress Test 1hr
CPU temp: 63°C
H115i temp: 32°C
Fans: 660 - 780rpm

X264 Stress Test 1hr
CPU temp: 64°C
H115i temp: 33°C

Prime95 Blend Test 1hr
CPU temp: 71°C but hit 95°C at some point
H115i temp: 33°C
Fan: 600 - 1860rpm

Not completely sure what happened with the Prime95 test as when I checked in on the test it was at around 71°C. Hopefully no damage was done. I ran the test again just before posting this discussion but stayed the whole time to monitor it, barely 5 minutes into the test it almost climbed to the 90°C again before I hit stop, here are the results:

Prime95 Blend Test
CPU temp: 87°C
H115i temp: 30°C
Fan: 440 - 1800rpm

Also I should include that I changed all CPU settings to what ETU considered my chip's "default". The only settings that would not take hold after rebooting the system are as follows:

Turbo Power Max
Default: 88w
Shown: 512w

Turbo Boost Short Power Max
Default: 110w
Shown: Unlimited

Turbo Boost Power Time Window
Default: 8sec
Shown: 16sec

Processor Current Time
Default: 105A
Shown: 256A

Also for reference my system's overall temps are completely fine, high 20s or very low 30s during idle and normal usage, airflow seems to not be an issue. Corsair LINK shows a sensor reading at 127°C constantly on the motherboard, which I assume is just faulty and the CPU fan header that the cooler plugged into shows that the fans are running at 1440rpm, which is incorrect.

I am at a loss on if it's again the cooler, my motherboard, the CPU itself or a combination. I can't see how I can physically make the waterblock any tighter to the CPU, it seems to have a good seal to me. I also use the dot or pea method of applying thermal paste, never too much or too little. Thank for reading this, post any help would be appreciated, I'll try to answer whatever questions I can.
 
Solution
Karadjgne
The theory works, but it will have certain adjustments. So far you've been pushing the cpu mainly, and with an aio after a half hour test (needs that long for liquid temps to equalize) your case temps will start to rise somewhat as you are dumping all that heat into the case. With a full tower, you have tons of volume to fill, so that process could take even longer, but eventually you will raise case temps significantly. During gaming, you'll be using a gpu that's also dumping heat into the case. Again, with a full towers larger interior volume, this'll take a while to change, but will happen. When using both cpu and gpu, you are dumping massive amounts of heat into the case, with only a single exhaust. Case temps are going to climb...
Sort by date Sort by votes
Karadjgne

Karadjgne

Titan
Ambassador
Dec 26, 2012
30,742
4,559
142,790
Fans below @1200rpm work better as pull config, fans above @1500rpm work better as push config, speeds in between are basically the same. You'll really want the rad fans as top exhaust, any extended usage of cpu and/or gpu is going to make case temps skyrocket with just a single exhaust, that's effectively being blocked by the downdraft of the rad fans. Pretty much think about being the same thing as a convection oven.

Intel cpu's are highly sensitive to temp variations, being checked 2x a second, so any core usage will make the fans ramp up quickly. This is entirely normal behavior for Intel cpu's, especially when getting bombarded by windows updates. The best way to work around that is change the way the cpu cooler responds and use liquid temps vrs cpu temps to govern fan speeds. The cores might bounce all over the place, but liquids have a much slower rate of change, having a far greater capacity to absorb energy than the cpu lid. Then adjust fan curves accordingly. Your aio would then behave more like a custom loop and less like an air cooler.

P95 is great for temps, but not the blend test, which does not use constant 100% cpu loads, it'll vary loads with ram usage. Use instead the small fft test. Also, don't use the newest versions, they use extremely high amounts of AVX instructions among others, which are totally unrealistic and will drive temps far beyond any reasonable expectations. Use instead version 26.6 as it's the most stable and accurate to date.

There's no reliable software built to accommodate every board and cpu. Even SpeedFan gives me 1 sensor at 255°C and another at -125°C, both of which are physically impossible, the sensors aren't broke, they just are not sensors at that address the software is looking for. So the Corsair link is about as accurate as any other software.

The h115i fans will run 1000-2000rpm, depending on fan curve and/or performance settings so will probably show better temps as push in exhaust config rather than pull
 
Upvote 0 Downvote
S

Shifty_Effect

Honorable
May 4, 2017
2
0
10,510


My CPU at 40°C give liquid temp of 30°C. CPU at 70°C-80°C give liquid temps of 33°C. A 3°C in difference would be extremely hard to set a fan curve.

My radiator is in a pull config as an intake now. I did have the radiator as an exhaust pushing the air out but there was no difference in temps at all, even system temps, at least nothing to get excited over. Having it draw air into the case gives me positive air pressure so I don't have to deal with dust as much. I don't mind either way if it fixed the temps but I've tested it and it didn't.

Thank you for the additional information, I will surely use it going forward as I'm figuring out a solution.

 
Upvote 0 Downvote
Karadjgne

Karadjgne

Titan
Ambassador
Dec 26, 2012
30,742
4,559
142,790
The theory works, but it will have certain adjustments. So far you've been pushing the cpu mainly, and with an aio after a half hour test (needs that long for liquid temps to equalize) your case temps will start to rise somewhat as you are dumping all that heat into the case. With a full tower, you have tons of volume to fill, so that process could take even longer, but eventually you will raise case temps significantly. During gaming, you'll be using a gpu that's also dumping heat into the case. Again, with a full towers larger interior volume, this'll take a while to change, but will happen. When using both cpu and gpu, you are dumping massive amounts of heat into the case, with only a single exhaust. Case temps are going to climb, that's unavoidable, and with each degree that case temps go up, the gpu is going to see reciprocal amounts of increased inefficiency, the warmer the air shoved across the heatsink, the less effective it becomes, higher the fans spin, hotter the exhaust. Which adds a cyclic process, further raising case temps, further affecting the gpu and round and round. A good, heavy, couple of hours gaming and gpu temps will be climbing through the roof.

You won't see that unless you run gpu torture tests for an hour or so.
 
Upvote 0 Downvote
Solution
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom