Pea sized blob method for Q6600?

Hi wingman ;)

After some experimatation this is what I found gives the best results:

After tinting both mating surfaces, a pea-sized blob, a small garden pea rather than a big marrowfat one, applied to the centre of the cpu. Using the TIM syringe I then gently spread it into a little square blob about 15mm square. When attaching the heatsink I too found the tendancy for it to begin sliding even if it was very level, I lower it into position then rotate about 20 degrees in either direction before screwing down. When screwing down I do both sides up hand tight then alternate with a quarter turn on either side to ensure even spreading of the TIM as the heatsink pressure on the cpu is increased.

Tinting the surfaces prior to the actual TIM application gave me a noticable temperature drop, this is something I reckon many people overlook as I don't read about it much on here & you can clearly see that the imperfections have been filled, no matter how hard you try to scrape it all back off some remains emedded.

Regarding the drying out of TIM I also found this with MX-2:

mx2onheatsinkjpg.jpg


After about 2 weeks I began getting a steady rise in load temps, I suspect the MX series of thermal pastes gives good results out of the box but degrades as it cures, wheras AS5 improves as it cures & since I went back to it I've had excellent results:

http://forums.overclockers.co.uk/showthread.php?p=18124491&highlight=temp+creep#post18124491
 
Last edited:
Yes, did exactly what it instructed on the Arctic Silver website & although I found the line method for quads wasn't as effective for coverage as a square shaped blob, tinting most definitely does give me better results than non-tinted surfaces.
 
Horizontal line is the (tested) best method for the Q6600. You should never manually spread the thermal paste anyhow. Some heatsinks i.e. those with direct contact heat pipes will require a slightly different technique and more paste.
 
There were some nice videos posted up not too long back that showed a selection of TIM application methods and why some might be better than others. Wouldn't know what to search on to try to find the thread now though.
 
C64 said:
Also those are very low volts for 3.4ghz could be why it isn't stable.

Every cpu and every heatsink is different though some concave etc you just have to find the best method for yours.

I tried spread method and lines and x's all sorts but found blob in middle size of a black pepper ball best for mine.
But that's using MX3.
Click to expand...

WingZero30 said:
Here are my results for 51 passes of Intel Burn Test.

[email protected] (VID 1.2625v )
Gigabyte GA-EP45-UD3LR
4GB (2x2GB) DDR2 Kingston HyperX PC2 8500 1066MHz (running@1008MHz) dual channel config

Cooling solution: IFX-14 with 2x Akasa Viper fans and MX-4 thermal paste

Bios Vcore: 1.4250V
Idle Voltage: 1.376V
Load Voltage: 1.312-1.296V

Chose 'Free' ram available from Windows task manager. Actually sligtly less value of 2400MB for more accurate testing hence higher value of over 40GFlops.

q660034ghz51ibtpasses.png


q660034ghz51ibtpassessu.png



There is always a difference of 7-8C between core0,1 and core2,3 no matter what thermal paste I apply. I use the horizonal line method as recommended for core 2 quads.
Click to expand...

the above are the results with the horizontal line method some days ago.

I have now enabled load line calibration and bios vcore is 1.35v giving idle and load voltage of 1.312v. So it should be able to handle 3.4Ghz fine considering it can go even lower to 1.296v and still pass 51 iterations of IBT.

I think I am definitely being held back by thermal compound application/temperature.

Thanks bainbridge. Interesting findings about mx-2 and as5. Will look into it further
 
51 runs with a max temp of 68c is great tbh don't know why you are worrying so much.

As long as gaming load is under say 65c and prime about the same it's fine.

It's definitely not temperature causing any fails if anything that's vcore being too low or northbridge.

Mines hit 90c and passed.
 
Last edited:
C64 said:
51 runs with a max temp of 68c is great tbh don't know why you are worrying so much.

As long as gaming load is under say 65c and prime about the same it's fine.

It's definitely not temperature causing any fails if anything that's vcore being too low or northbridge.

Mines hit 90c and passed.
Click to expand...

Yeah the northbridge gets very hot. I made a thread about it and there were interesting posts. Some people modded the heatsink for northbridge. The thing is the northbridge heatsink gets very hot and you can only keep your finger on it for 1/2 sec most. It's like you are touching a metal that is scorching.
 
Rroff said:
Horizontal line is the (tested) best method for the Q6600.
Click to expand...

Not for me it isn't, I tried lots of different methods & settled for what works best for me.

Rroff said:
You should never manually spread the thermal paste anyhow.
Click to expand...

Shaping a pea-sized blob into a 15mm square gives me excellent coverage once screwed down, it spreads itself right to the edges of the cpu without splurging out of the sides.
 
Ok ran 20 passes of IBT this morning which it passed. The windows were wide open but not much air was coming into the room as I believe the wind direction was opposite to the window airflow intake direction.

For [email protected], 3.4 x 16 = 54.4GFlops is the theoretical maximum value

I have been getting over 43GFlops which is roughly close to 80% of the theoretical maximum and this is the maximum I get which is still high enough. The other day I ran IBT and I was getting 28GFlops for the same speed lol as I didn't choose right ram settings and temps were definitely lower.

Here are the results:

q6600ricegrainmethod.png


There is still temp difference of 7C between core0,1 and core2,3. However overall temps seem to have increased by a whopping 7C aswell :eek: with rice grain method compared to the results with horizontal line method which I posted previously.
The heatsink was tightly bound to the cpu surface.

Hmm it seems that rice grain method isn't very good for Q6600. I haven't tried the pea sized blob method. I am assuming it may give same results. I may well have to go back to horizontal line method.

Your thoughts guys?
 
tinted first here the a peppercorn size blop which is really just for good measure as the tinting process is pretty as much as you need, all the crevaces are filled with the tinting process
 
Just a quick example short run but prime runs don't go over 65c.
I am using blob in middle method prime stays under 65c but iBT can get very hot anywhere from 75-85c.
MX3 under a tuniq tower 120.
To be under 70c IBT are great temps I hover around 80-85c on a long run.

All that matters really are gaming loads as long as those don't go much over 60c I think it's fine.
Capture.jpg
 
Last edited:
Ok applied the pea-sized blob method in the middle and temps are more or less similar to horizontal line method. After 3-5 IBT runs, the room became quite warm as if the radiator has been running in the room.

Seems like the rice grain method wasn't as efficient in my case.

Load line calibration was enabled.

Bios Vcore: 1.35v
Idle Voltage: 1.312v
Load Voltage: 1.312v

Considering that VID is 1.2625v, so (1.35-1.2625)/1.2625 = 6.93%
So for 6.93% increase in cpu voltage, 1Ghz stable overclock is impressive :) plus it requires modestly low load voltage at 3.4Ghz :).

However as stated before that no matter what thermal paste application method I use, core0,1 always run 6-8C hotter than core2,3. I guess that core0,1 chip is different from core2,3 chip. Just wished that core0,1 ran atleast 5C cooler. Hmm never mind :).


q6600peasizedblobmethod.png



q6600peasizedblobmethod.png



The time duration for IBT is 86.571 mins or 1 hour 26 mins which I think is long enough for such a stressful program.

Now thats out of the way I think the next step is:

Prime Blend test.
 
@C64

Yeah I have been monitoring my temps while heavy gaming and temps stay under 60C:).

I am surprised that your load voltage is 1.4v for 3.5Ghz even though your VID is only 2/3 notches higher than mine.
 
WingZero30 said:
@C64

Yeah I have been monitoring my temps while heavy gaming and temps stay under 60C:).

I am surprised that your load voltage is 1.4v for 3.5Ghz even though your VID is only 2/3 notches higher than mine.
Click to expand...

Try 390x9 at those voltages ? you might be surprised and find that you will also need 1.4 volts + to get 3.5ghz IBT stable.
You also need to set it to 4 threads from auto.

Not even sure I am running the correct settings when it comes to IBT you set the free memory you have right ?
How long are you meant to wait in windows for it to cache everything though etc ?

Yea I think most of these chips require around 1.38-1.42 volts for 3.5 and 3.6 + is anywhere from 1.43 - 1.5 volts.
Mine probably will do 3.4 ghz on same voltages as yours but I haven't tested that as i like to run at 3.5ghz +

But again at what settings on IBT is it considered watertight ? it says to set it to just under whatever free memory you have mines around 2000 (at the moment) maybe i'll wait until win7 caches everything first.
It also says to disable speedstep which I haven't done yet shall try some clocks and settings today with speedstep disabled.

See how many volts you require for 390x9 I am using custom setting 2000mb 4 threads on IBT mine requires 1.41 set in bios for 390x9.
And I think will do 394x9 at that voltage.1.30 northbridge and cpu pll set to 1.52.

Just to be safe I'd get it IBT passed and still bung one or two notches up on the vcore above where it's stable at.

I need to get a thermalright archon + 2x 140mm fans.
 
Last edited:
C64 said:
Try 390x9 at those voltages ? you might be surprised and find that you will also need 1.4 volts + to get 3.5ghz IBT stable.
You also need to set it to 4 threads from auto.

Not even sure I am running the correct settings when it comes to IBT you set the free memory you have right ?
How long are you meant to wait in windows for it to cache everything though etc ?

Yea I think most of these chips require around 1.38-1.42 volts for 3.5 and 3.6 + is anywhere from 1.43 - 1.5 volts.
Mine probably will do 3.4 ghz on same voltages as yours but I haven't tested that as i like to run at 3.5ghz +

But again at what settings on IBT is it considered watertight ? it says to set it to just under whatever free memory you have mines around 2000 (at the moment) maybe i'll wait until win7 caches everything first.
It also says to disable speedstep which I haven't done yet shall try some clocks and settings today with speedstep disabled.

See how many volts you require for 390x9 I am using custom setting 2000mb 4 threads on IBT mine requires 1.41 set in bios for 390x9.
And I think will do 394x9 at that voltage.1.30 northbridge and cpu pll set to 1.52.

Just to be safe I'd get it IBT passed and still bung one or two notches up on the vcore above where it's stable at.

I need to get a thermalright archon + 2x 140mm fans.
Click to expand...

Yeah I tried going to 3.6GHz and bios vcore was 1.5v+ and the system still didn't stabilise and crashed lol hence I have enabled LLC in bios for 3.4GHz. I think the amount of vdrop/vdroop that we experience depends a lot on the mobo voltage regulator coupled with northbridge chipset design i.e. P35,P45,X48 etc and the chipset cooling.

So 4 phase voltage regulator won't be as efficient as 6 phase or 8 phase in distributing the required voltage to cpu so overclocking stability will be more challenging. On the other hand 6 or 8 phase shouldn't require as much bios vcore and vdrop/vdroop should be less aswell imo.

Coming back to IBT; here is again the thread and the actual article on the subject of linpack (Linx,IBT) for clarity:

http://forums.overclockers.co.uk/showthread.php?t=18206940

When I start windows I turn off antivirus momentarily and immediately turn on cpu-z, coretemp and IBT(not starting). I turn on windows task manager and for 4GB (2x2GB) ram in dual channel mode, I usually get 2734MB something as 'Free' ram. So I choose 'custom' stress level and input 2500MB to ensure accurate results.
What I have noticed is that if I don't do any activity then 'free' ram decreases quite a lot to 1980MB something in about 10 minutes. So I always ensure I quickly start IBT. Ofcourse the amount of ram also depends on the background processes that windows is running and as the article suggests minimising them. I usually just leave them as I don't want to mess up the whole process. I think 2000MB of free ram is still good enough as loading will be sustained for some time during each run.

Choosing 'free' physical ram just ensures that the calculation data is readily available to cpu and that it doesn't have to wait for it thereby reducing the loading and lowering temps which occur if virtual ram is also part of stress testing such as during 'maximum' level going by the article. So free ram should also mean higher GFlop values.

Judging by the comments in article the auther notes 70% of the theoretical maximum GFlops as acceptable as some people have been getting about that much.So in your case:

[email protected] : 3.510 x 16 = 56.16GFlops (Theoretical mximum value)

So during IBT testing you should be looking at around 39.312GFlops as the minimum acceptable value for passing the IBT test. Anything higher is a bonus.
Correct me if I am wrong but isn't GFlop the measure of computer speed at which it executes/processes the software instructions? So higher the Gflops mean that cpu will execute the instructions faster thus generating more heat.

The amount of 'free' ram is directly proportional to loading time it takes the cpu to complete each run. So larger the value, longer each run will take to complete for same clock speed of cpu and thus cpu is being stressed longer. So in my case if I am getting around 83s for 2500MB of ram then theoretically for 5000MB the testing time for each run should be doubled to 156s as the amount of calculations to be processed has doubled.

I did some preliminary testing with threads set to 4 and I seem to be getting more or less same results as posted above. However I will be posting my results soon:).
 
Last edited:
My tuniq tower cannot cope just doesn't cut it when it comes to 3.6ghz and IBT it will hit 90c.
And i think you are meant to set free ram not available ?
 
You must log in or register to reply here.
Back
Top Bottom