Poor performance from H105 with stock 4790k

mpj88 · 20 Sep 2014 at 17:21

Hi,

I've built a new rig:

Asus Gene VII
i7 4790k (@ stock)
SuperFlower 750W PSU
Kingston 8GB 2400MHz

Now I originally built it with a corsair H80i. I saw a cheapish H105 come up on the forums and thought I'd try it as assumed it would be better performing and be able to be quieter when not stress testing.

Its currently idling at 23-27 degrees which I'm happy with. I set the fans to maximum (approx 2500 rpm) and run small ffts on prime. The temperature soars to 80 degrees reported in coretemp and has shut itself down once or twice when doing this.

The pump is run off a fan controller and is running at max. Fan controller reporting at 1900rpm.

I've checked the TIM and that seems fine.

I know stress testing is above normal use however this is normally what I would do for overclocks. This is leaving no room to fiddle when I've paid for an unlocked proccessor!

Did I set my hopes for a dual rad performance too high!?

Make · 20 Sep 2014 at 18:03

Sounds a bit high for stock. Try reseating the cooler.

Prime95 temps are normally wayyy out of anything you'd see in real use, but that seems a bit high for stock. What temps do you have on a normal load, like gaming?

malcolm · 21 Sep 2014 at 13:00

mpj88 said:
Hi,

The pump is run off a fan controller and is running at max. Fan controller reporting at 1900rpm.

Connect the pump direct to a fan header, a fan controller will never give the full 12v so the pump will not run at the correct speed.

setter · 21 Sep 2014 at 13:08

80c is actually not too bad, mid 90's here. Tbh you shouldn't be running p95 on theese chips. They're not designed/built to handle such high loads. Google delidding and you'll see why. Run Asus real bench ti get an idea of more realistic loads.

mpj88 · 21 Sep 2014 at 19:32

malcolm said:
Connect the pump direct to a fan header, a fan controller will never give the full 12v so the pump will not run at the correct speed.

Its a powerful fan controller. I tried it on a mobo fan header and it reported the same rpm.

setter said:
80c is actually not too bad, mid 90's here. Tbh you shouldn't be running p95 on theese chips. They're not designed/built to handle such high loads. Google delidding and you'll see why. Run Asus real bench ti get an idea of more realistic loads.

Hmm ok. Just was hoping to try some overclocking but not very confident if I cant stress test!

I have researched de lidding but not overly keen on doing this.

setter · 21 Sep 2014 at 19:35

Just stress it with normal tasks, asus real bench ive mentioned above is a good quick realistic test. My max temp in it at 4.7ghz, (1.300 vcore) is 81c. At stock using the latest version of p95 it's 94c in seconds. At 4.7ghz in games i get anything from 59 to 70c on the hottest core, but thats with two gpu's chucking a lot of heat into the case.

Threepwood · 21 Sep 2014 at 19:41

Sure, but some things will only be stable if the system is passing a hard core test.

For me, if the system is stable enough to run IBT then I know ARMA will also be stable.

If not IBT stable then ARMA is guaranteed to crash. Even if it is running cooler.

Just an example

setter · 21 Sep 2014 at 19:45

Threepwood said:
Sure, but some things will only be stable if the system is passing a hard core test.

For me, if the system is stable enough to run IBT then I know ARMA will also be stable.

If not IBT stable then ARMA is guaranteed to crash. Even if it is running cooler.

Just an example

Yep, i understand what you mean and can agree to a point. But if the op overclocks and runs stress tests such as p95/Ibt/LinX, the system will in all likelyhood hit throttle limits and shutdown. Therefore you may never know if it's fully stable or not. Also on previous gen cpu's, ive ran the above tests as long as 24 hours in some cases, only to load up a game and have it crash in no time.

BriT · 21 Sep 2014 at 19:49

Surely that's far too high for stock? my gets to 62c max on stock when running prime.

setter · 21 Sep 2014 at 19:56

Which version of p95? The latest 28.5 using small fft is insane for temps. Seen the mid 90's on two 4790k's at stock. My old 4770k was a bit cooler but that was under an older version of prime. Intel said that devils canyon would run cooler due to better thermal paste, improved ihs fitting and better power delivery. Well that turned out to be a bit of a porkie, in most cases ive seen the DC chips run hotter than the older series.

BriT · 21 Sep 2014 at 20:08

Ah im using the older version, 25.11.

FredFlint · 21 Sep 2014 at 20:26

I have the h105 pump wire connected to a 4pin to molex to the PSU so it just runs at full speed all the time.

setter · 21 Sep 2014 at 20:29

H105 is a decent enough cooler, as is any current high end air heatsink. For normal use you wont see the sort of temps stress tests generate.

mpj88 · 21 Sep 2014 at 21:30

Threepwood said:
Sure, but some things will only be stable if the system is passing a hard core test.

For me, if the system is stable enough to run IBT then I know ARMA will also be stable.

If not IBT stable then ARMA is guaranteed to crash. Even if it is running cooler.

Just an example

this was my thinking!

setter said:
Which version of p95? The latest 28.5 using small fft is insane for temps. Seen the mid 90's on two 4790k's at stock. My old 4770k was a bit cooler but that was under an older version of prime. Intel said that devils canyon would run cooler due to better thermal paste, improved ihs fitting and better power delivery. Well that turned out to be a bit of a porkie, in most cases ive seen the DC chips run hotter than the older series.

Yes, using 28.5 build 2. Ok, I don't feel so bad now then!

During a bit of gaming last night. I didn't see it go above 60 with the fans turned down so I'm happy for now!

setter · 21 Sep 2014 at 21:45

You can still oc and get good temps in your everyday tasks, and if/when you get an error you tweak from there. For bsods you can google the error codes to help diagnose what may be causing the issue.