High Stock Temps - What am I Doing Wrong?

[fuzzykeys]

HWMonitor is showing temps from the high 70's to low/mid 80's running prime95 small FFT with my CPU at stock settings. It seems like I can't really overclock at all from here. Any ideas?

Here's my hardware setup...
4770k
GTX 700
Corsair h80i
Corsair Carbide 500R
Samsung 840 Pro

 

[househead]

Prime95 and Intel Burn test are BAD TOOLS to test Haswell if you believe what you read.

Prime has my 4770k hitting 90C at stock speeds. I believe it to be overvolting the CPU. Saying that, 70-80 is not a worry and is what I would expect to see during a stress-test. get worried when it's hitting 90-95.

Try Intel Extreme tuning Utility which has a stress-test and benchmark mode. It's actually a half decent piece of software.

 

[fuzzykeys]

I don't have a Windows installation. Can you recommend something for OSX?

 

[househead]

I don't have a Windows installation. Can you recommend something for OSX?

Sorry I don't know of any well-known tools to stress a CPU in OSX. I'd be very interested what others might be able to suggest.

Much of the overclocking community appear to be gamers which probably explains why much of the good advice around overclocking suggests windows tools for stressing. I'm sure there are OS X methods ... one could run a Handbrake encode I suppose, that should certainly get your CPU working hard!

 

[neilhart]

HWMonitor is showing temps from the high 70's to low/mid 80's running prime95 small FFT with my CPU at stock settings. It seems like I can't really overclock at all from here. Any ideas?

Here's my hardware setup...
4770k
GTX 700
Corsair h80i
Corsair Carbide 500R
Samsung 840 Pro

Well I don't agree with househead. Prime95 torture test and Geekbench stress tests are good utilities to load up the CPUs and check the load temps.

That said, there have been extream high temps seen with some Ivy Bridge and some Haswell CPUs. There is a group of overclockers that take the heat spreader off of the Intel packages, clean the epoxy off the lid and board and clean off the Intel heat sink compound. They then use their favorite heat sink compound and replace the lid. Some report 15 to 25 degrees C improvement in loaded CPU temps.

Use Google to learn more on the subject.

Good modding,
neil

 

[macnb]

Prime95 and Intel Burn test are BAD TOOLS to test Haswell if you believe what you read.

I have NOT read that anywhere. What facts are you basing your comments on ?

HWMonitor is showing temps from the high 70's to low/mid 80's running prime95 small FFT with my CPU at stock settings. It seems like I can't really overclock at all from here. Any ideas?

Low to mid 80's is not too bad with Prime95. Though a little high.
I have an i7-3770K cooled via H100i. I used to hit 98+ DegC with Prime95 OC @ 4.7 Ghz (same with Win8).
I found couple of things I did wrong:
1. The backplate that Corsair provide was incorrectly mounted on the back of the motherboard. This backplate is NOT symmetrical on both axis. It has too notches to clear the existing screws on the motherboard. I had mounted it 90 deg out. I am sure you have checked this with your H80i but just in case you haven't....worth checking.
2. My pump block was slightly snagged on a capacitor next to the socket so that it was not fully flush with the CPU's IHS. This is a common problem with Gigabyte boards. Worth checking that your pump block is not resting on anything other that the CPU's IHS.

I fixed those two issues and my temps dropped by 8-10 DegC. Still not brilliant.

I have subsequently, de-lidded my i7 and completely cleaned off Intel thermal interface and the black rubber seal. I applied Coollaboratory Liquid Ultra (CLU) onto the CPU die on the back of the IHS. On top of the IHS, I used Artic MX-4 thermal paste.
I can now OC to 4.9Ghz @ 1.416v with max temp of 87 DegC for 15 hours.
I have managed 5.0 Ghz @ 1.500v with max temp of 95 DegC for an hour but for it's too hot for the extra gain in performance (apart from the 5G bragging rights ).
Oh, btw, I run my stock Corsair fans on the H100i on a custom curve...max 1500 rpm (fairly quiet).

Geekbench 3 score @ 4.9Ghz = 18447
Geekbench 3 score @ 5.0Ghz = 19148

So for me, delidding made a massive difference in temp, voltage and fan noise.
I gained 200Mhz in perf and a temp reduction of about 15 DegC.

 

[oswaldm]

@macnb

I wonder how long the CLU is going to last under the lid. I used to follow the topic on overclock.net and people over there reported that after some period of time the temps wen't high again. I can't provide actual links to the discussions - just a memory from that time.
I am yet to delid my 3570K because I'm hitting 71-75*C on 4.3GHz at 1.25V


@fuzzykeys

Reset the cooler. Use a good thermal compound such as Prolimatech PK series. You can also provide info on you ambient room temps, where did you place the radiator for the H80i etc.

 

[macnb]

@macnb

I wonder how long the CLU is going to last under the lid. I used to follow the topic on overclock.net and people over there reported that after some period of time the temps wen't high again. I can't provide actual links to the discussions - just a memory from that time.
I am yet to delid my 3570K because I'm hitting 71-75*C on 4.3GHz at 1.25V

I think there's a lot of misinformation about CLU. A lot people have reported no issues with CLU after 9-12 months. They checked under the IHS and the CLU was still in liquid form and completely malleable. CLP (Liquid Pro) is the old version that does solidify after a while - that's why they brought out the CLU.
So I am not worried at all about the CLU. I do check overclock.net on regular basis.
Your temps don't seem to bad - though 1.25v seems high for 4.3Ghz. I can get 4.6Ghz @ 1.25v
De-lidding your CPU should get you down to low-mid 60's DegC.

 

[oswaldm]

I think there's a lot of misinformation about CLU. A lot people have reported no issues with CLU after 9-12 months. They checked under the IHS and the CLU was still in liquid form and completely malleable. CLP (Liquid Pro) is the old version that does solidify after a while - that's why they brought out the CLU.
So I am not worried at all about the CLU. I do check overclock.net on regular basis.
Your temps don't seem to bad - though 1.25v seems high for 4.3Ghz. I can get 4.6Ghz @ 1.25v
De-lidding your CPU should get you down to low-mid 60's DegC.

Thanks for the clarification. How about the issue of the IHS sitting tiny-bitty lower to the core (because of lack of the adhesive)? I remember that someone damaged/cracked their core due to too much pressure from the heatsink.

My 3570K is a turd... But my mobo won't let me test it properly due to lack of advanced OC settings in the BIOS. I can only operate on offsets and that was very iritating from the start. I managed to get it decent on 4.3GHz and I left it there without further investigation. Not being able to set your vcore manually is a bitch.

 

[macnb]

Thanks for the clarification. How about the issue of the IHS sitting tiny-bitty lower to the core (because of lack of the adhesive)? I remember that someone damaged/cracked their core due to too much pressure from the heatsink.

To crack the die is very very hard (due to the gap left behind after removing the seal). Whoever did that clearly did something terribly wrong.
The CPU socket has spring loaded pins and when the CPU is placed on it, the springs will take/make up for the gap and then there will be a slight (very minute) flex of the CPU PCB as well when the cooler is pressed down onto the IHS. So nothing to worry about.
Some people actually lap (grind) the bottom ridge (that used have the seal) to INCREASE the gap.
There are people who have removed the CPU mounting bracket completely have mounted the cooler DIRECTLY onto the die (without the IHS). That has a high chance of cracking the die if you don't have proper (finely measured) stops in place.

Regarding OC, I prefer to use OFFSET voltage (DVID setting on Gigabyte boards) instead of FIXED Vcore as I am not OC'ing 24/7.
The advantage of using offset is that most of the time, the Vcore will be lower thus running cooler and quiet and hence prolonging the CPU life. The Vcore only increases when demand is placed on the CPU. I am into photography and video and not processing video/photos all the time so do not need high Vcore all the time.
The disadvantage of offset voltage is that when there is demand on the CPU, Vcore will up slightly higher than really needed (around 0.01v to 0.03v) but that depends on the how good the VRM implementation is on the board. Gigabyte VRM is rock solid. E.g. I can get stable run (24 hours+) of Prime95 at fixed 1.400v for 4.9Ghz but if I use offset by setting DVID = 0.175v, Vcore would rise to 1.416 (sometimes to 1.428v) for the same Prime95 24 hour run.

So, if you want cooler system for general computing, use Offset voltage.