Re-apply thermal paste yet again?

[JimmiG]

Every time I apply thermal paste I feel that I mess up somehow. The first time I tried the "pea" method but because some paste kept coming out of the syringe after I stopped pressing, it left a little line trailing to the right of the dot, and it and caused the round dot to become slightly stretched/elongated to the right. Also I failed to align the heatsink properly on the first try, so I had to lift it slightly after I had pushed it down onto the CPU, which may have created air bubbles.

So after a few days I decided to clean and re-apply the thermal paste.and try the "X" method this time. This time I messed up a little and the center of the X ended up off-center (slightly too far down), which is obviously not good since the CPU die is in the center.

Temps seem to be about the same, maybe 1-2C higher now, but it's hard to tell because the weather keeps changing from day to day. Should I give it another try or just leave it as is? Some say they gain about 5-10C by getting a really good TIM footprint...which would certainly come in handy (pushing ~85C in Prime95 Small FFT's at 4.3 GHz/1.2V). It's just so hard compared to the videos where they just dab some onto the, install the heatsink and when they take it off they have a perfect, smooth layer. Wonder how many takes they do before it looks like that...

 

[Araxie]

at most you gain 1-2 degree with a perfect TIM apply.. wich in most case its a very thin and even layer of TIM, and really depend on the TIM and cooler.. but 10C for just a TIM apply method? no way thats possible.. what cooler are you using? what TIM?. because I see that temps on 4770k@ 4.7ghz 1.3V

in general rule to down the temperature of IB/HW chips the best its just delid the chip..

 

[mhenley]

I have seen 10C before, but they went from one of those melting thermal pads to a high quality non-curing paste AND lapped both the sink and processor at the same time. Results in those numbers are not common. What heatsink are you using? What paste? Your methods sound right, (either pea or X) and it being slightly off wouldn't necessarily harm things considering how much pressure there is.

Also, 85C at full load with those volts sounds very reasonable.

 

[Nenu]

Op you worry too much.
The shape of the glob being altered by a tiny trail of paste wont matter one bit.

If its still too hot, its either down to high ambient, bad case airflow/re-circulating warm air, poor cooling or badly fitted cooler.

 

[JimmiG]

because I see that temps on 4770k@ 4.7ghz 1.3V

That's why I'm wondering if something is wrong. I'm using this cooler and Arctic Cooling MX-4.
http://www.techpowerup.com/reviews/Phanteks/PH-TC14PE/6.html

Temps are much lower with Blend or In-Place large FFT, it's only "Small FFT's" (8 - 16K AVX) that sends the temp into the mid-80's range. Idle is about 28C. Something seems to be wrong, but on the other hand, I read stuff like this:
http://forums.extremeoverclocking.com/showthread.php?p=4095062

I can boot and run cinebench at 4.6HT with 1.25V, but when it comes to prime small ffts, 4.3HT 1.15V hits 90C. That's with a ZT-10D heatsink with two NF-F12s full speed. 21-22C ambient.

That's far worse than my temps. So maybe it's just Haswell that doesn't like Small FFT's...

If its still too hot, its either down to high ambient, bad case airflow/re-circulating warm air, poor cooling or badly fitted cooler.

Ambient is about 27C (summer, no AC). I have five 140mm case fans in a Fractal Design XL R2 full tower case. Cooler is fitted correctly - it's not exactly rocket science, you basically just screw it on. The only thing that didn't go perfectly, like I said, is the TIM application...

 

[Araxie]

That's far worse than my temps. So maybe it's just Haswell that doesn't like Small FFT's...

Thats totally right.. in fact if you read [H] 4770k Review they put this:

Tests are Tests, Right?
ASUS warned us against using synthetic stress tests like are shown above.

Unvalidated stress tests are not advised ( such as Prime 95 or LinX or OCCT, Intel Burn Test or other comparable applications ). For high grade CPU/IMC and System Bus testing Aida64 is recommended along with general applications usage like PC Mark 7. Aida has an advantage as it is stability test has been designed for the Haswell architecture and test specific functions like AES, AVX and other instruction sets that prime and like synthetics do not touch. As such not only does it load the CPU 100% but will also test other parts of CPU not used under applications like Prime95. Other applications to consider are SiSoft 2013 or Passmark BurnIn. Additionally this generation has a more specialized point of consideration for synthetic stress tests. When using an adaptive vid voltage control will be automatically controlled by the iVR when a complex concurrent AVX load is initialized from Applications like Prime95 or Aida Or LinX more voltage will be supplied than has been defined/requested.

so yes, Haswell do not like Prime95...

try with aida.. make couple of test..

1)Only CPU Stress Test (Typical load on regular games)
2)Only FPU Stress Test(Insane amount of heat generated btw in just FPU)
3)CPU+GPU Stress Test (Typical load under CPU intensive games..)

 

[JimmiG]

Here are the results in Aida64 with my "mis-aligned X" method of TIM application at 4.3 GHz / 1.2V

FPU-only is about the same as P95: 85C
FPU+CPU = 77C
CPU only: 61C
CPU+FPU+GPU: 80C
CPU+GPU: 62C

All temps are the max recorded for the hottest core after temps have stabilized.

Also, Prime95 with the middle option "In-Place Large FFT's" results in just 68C, even though it sais "maximum heat and power consumption". Seems like the pure FPU tests are the worst. Maybe the CPU+FPU test bottlenecks the CPU, giving the FPU's some breathing space.. either that or my poor TIM application skills are only affecting the FPU somehow, which doesn't make sense considering how small the Haswell die is.

What I don't understand is how people are getting their CPUs to work with 1.3V and claiming temps in the 70C to low 80C range. I'm pretty sure my 4770K would do 4.5 GHz @ 1.3V like everyone else's, but the not while staying below 100C...