Temps w/ i7-4790k and H100i GTX

Maxxtraxx · June 10, 2015

Hi all,

I've been an avid LTT forum surfer and channel watcher for about 10 months now, and I've been enjoying the forums and the channel greatly.

I wanted to share about my recent cooler... modification? On my 4790k and h100i gtx for any that may benefit from what I've experienced.

Approximately 3 months ago I upgraded from an i5-2500k with hyper 212(non evo) on a gigabyte z68xp-ud3 mobo to an i7-4790k with a corsair h100i GTX and an Asus z97-a mobo. I was excited to see the differences in temperature and overcooking ability, the 2500k performed well and with a simple Intel xtu tune still runs stable at 4.5ghz all the time with temperatures in the 60deg C range.

The 4790k after initial setup and a slight bump to 4.5 GHz with core voltage simply left on adaptive was running under stress testing with Intel xtu at temperatures hovering around 90deg C. I tried running stress tests with occt and prime 95 to test whether those tests would prove any different... to no avail, they actually pushed the temperature up to the 95-99deg C range. The temps I was getting lead me to try reapplying the thermal paste several times (noctua NT-h1) with no change in temperatures. The temperatures I was getting were not consistent with what Linus or any other reviews I had found on the 4790k. I had heard that the Haskell processors averaged 10deg C warmer than the sandy/ivy bridge processors but this was pushing 30deg or more and with a far superior cooler(no disrespect for the hyper 212). Alas the cooler was pumping, the fans were running, under normal gaming loads temps were more than reasonable and I had spent 2 days trying different things with no changes, so I left it be and moved on until new information... and desire to try again surfaced.

Today I decided to give it another go, I had noticed upon installing the back brace for the cooler that there was some play in it even with the retaining screws fully tightened and I had also noticed that the pump block nuts seemed to bottom out on the retaining screws relatively early(imo). So after reading a forum article(forget where) in my vain search for ANY information, I found a posting detailing a similar setup where rubber washers were used to take up the slack in the back brace and help increase the mounting pressure(Linus's videos referencing the current high mounting pressures used on modern processors also reinforced the idea).

I decided to go over to lowes and look in the hard to find parts bins for a washer that would work well, I found a culprit I though would work, a rubber grommet 11/32 by 1/8 ($1.27 per grommet). The grommet are stiff with some give when squeezed and seemed to work well placed between the motherboard and the cooler back plate. The pump retention nuts now pushed much harder on the pump block brace and I was a bit hesitant to turn them all the way down, I had reapplied thermal paste and now fired everything up for some idle temperatures.

Idle temperatures dropped immediately from 40-43deg C(core 1)

To a 30-33deg C(core 1) and 24-28deg C(cores 2, 3, 4)

for reference my 2500k with 212 idles at 29-31deg C(even across cores)

Load temperatures which were 90deg C with XTU prior

Dropped to a 60-63deg C with XTU after

There's my results, they stay consistent after 30 minutes of CPU stress testing with XTU. If there are any others that have experienced high temps with their corsair h100i gtx's, this may be a DIY modification that could give better results. I've gone from a bit disappointed in the processor and the cooler because I was assuming I had done very poorly in the silicon lottery, to now being extremely happy and hopeful that I can get some additional clock speed out of my investment.

Lays · June 10, 2015

Do you have adaptive voltage set? If so, that's another issue. Turn it off and use "override" or "manual"

You could try that as well, it may get temps down even more.

Maxxtraxx · June 10, 2015

Do you have adaptive voltage set? If so, that's another issue. Turn it off and use "override" or "manual"

You could try that as well, it may get temps down even more.

Thanks for the tip! Will turning off the adaptive core voltage mode force the CPU to operate at that voltage at all times? If no then that is appealing to me, if yes then I think I would prefer a higher voltage (up to 1.3v) under heavy load then have adaptive scale it back when full voltage is not needed to help with silicon longevity.

Lays · June 10, 2015

Thanks for the tip! Will turning off the adaptive core voltage mode force the CPU to operate at that voltage at all times? If no then that is appealing to me, if yes then I think I would prefer a higher voltage (up to 1.3v) under heavy load then have adaptive scale it back when full voltage is not needed to help with silicon longevity.

Yes it will use the voltage you've set 24/7. If you're only using 1.2-1.3volts, it's fine to run at that 24/7, and a lot safer than adaptive imo.

Silicon longevity is irrelevant imo. Look at how many oem PC's that were in terrible airflow environments that ran for 5+ years in dust filled cases that were on the brink of tjmaxing the entire time that still work perfectly.

The only way you're going to kill a cpu before it's time to replace it with an upgrade is if it's faulty in the first place, or you run it at 1.6 volts 24/7 at 100% load for months on end.