[GFORCE100]

Hi all,

So here's the story, same Abit IP35 PRO motherboard, same 16B1 BIOS, same Zalman 9700NT cooler but 2 different Q6600's, both G0 steppings, different Vcores, same 3.6GHz frequency (9x400).

Q6600 #1. actual load Vcore 1.41V
Under Prime95 25.5 Core Temp 0.95.4 reports load temps as 70/70/71/71

Q6600 #2. actual load Vcore 1.33V
Under Prime95 25.5 Core Temp 0.95.4 reports load temps as 74/74/69/69

Now my questions:

Question #1 Why are the temps so high on Q6600 #2 if its running on such a low Vcore at 3.6GHz? I've reseated the 9700NT with three different patstes (AS5, Artic Ceramique and Zalman STG-1). Temps remain higher than with Q6600 #1.

Question #2 What would you guys do knowing the 9700NT can cool better and Q6600 #1 is proof of that (i.e. higher Vcore but same temps hence an indicator that something is wrong with Q6600 #2)?

Are you all going to jump in the same direction by saying lap Q6600 #2?

I put a business card on Q6600 #2 IHS and it seems flat to me, very slightly covex with the sides lower than the middle (minor light shows through at the sides).

It's a 1.2500 VID CPU btw but it seems to be very stable when overclocked. I know I can get temps down to around 60C with the 9700NT with this low Vcore.

Thanks,

[hUMANbEATbOX]

All CPU's are different. That's pretty much it.

[CGR]

It could be that #2 just has a poorer contact between the core and IHS than #1.

[GFORCE100]

Here's a screenshot of the latest temps on Q6600 #2 at 1.33V load @ 3.6GHz.

[dipspit]

I'd have to reiterate that not all cpus are created equally. You have to keep in mind the tiny measurements involved and how even the slightest variance could effect the contact points. Also, the imperfections in the first chip may have better contact with the zalman than the second chip. Instead of lapping the chips, try lapping the cooler and re-test.

[hUMANbEATbOX]

Another thing to consider is that the temps are all estimations, calculated by software. By no means are the readings perfect.

[bing]

Not sure if your mobo can do under voltage, click my sig on "Intel UVOC" to verify if the HSF is working as expected.

[xilix]

The only thing I can tell ya is that if you have a G0, those are safe temps, and that if you want lower temps, you're probably gonna have to go with a better cooler.

[GFORCE100]

Well the Zalman 9700NT is just about as flat as they come, and nickel plated too.

It's either poor contact between cores and IHS, or poor contact between Zalman 9700NT and IHS.

I need to find a razor blade to attest just how flat the IHS is.

Sadly the IP35 PRO doesn't allow undervolting, the lowest it will go is whatever the VID is, so in this case 1.2500.

So far it's a case of amazing low Vcore CPU, I mean how many G0's to 3.6GHz at 1.33V actual? but poor temps.

[graysky]

Two major factors come to mind for me:

1) how flat are the surfaces (i.e. lap them). My thread on lapping my q6600 and my thread on lapping my u-120 ex
2) vocres that high will produce heat levels exponentially higher since heat is a function of the square of voltage.

Please have a look at my overclocking sticky. Do a keyword search for "missing intel" to read about that formula and then, scroll to the bottom and read my section on thermal management.

...what are your temps @ 9x333 and what vcore are you using for 9x333?

[GFORCE100]

Originally Posted by graysky

Two major factors come to mind for me:

1) how flat are the surfaces (i.e. lap them).
2) what are your temps @ 9x333 and what vcore are you using for 9x333?

1) Using a razor blade, Q6600 #2 surface is flat I would say with some uneveness here or there but rather micro than macro, nothing some thermal paste couldn't fill easily (~0.3mm). The 9700NT surface is almost 100% ideal, at most gaps around ~0.1mm.

2) At 9x333 I set the lowest voltage possible in the BIOS, which is the VID of the CPU, Q6600 #2. I set it to 1.2500V. Under Prime95 25.5 Small FFT same as before, the actual load Vcore is 1.18V and my temps, according to Core Temp 0.95.4 are 55/55/51/51. The PWM CPU fan doesn't spin 100% due to the fact the CPU temp isn't that high, according to Abit's uGuru it's only 41C. PWM temp is also very cool at 46C (these under Prime95 25.5).

The 9700NT feels cool, there is a steady breeze through it from the 110nm fan.

[majinwar]

Sorry, I'm not trying to jack this thread, I just have a question that is probably a dumb one, but I'm going to ask it anyways since I've never built a computer from scratch myself, but soon will!

The green board around the intel chips, do you remove the cpu from it before putting it into your mobo? Or does it stay attached even when you put it into the mobo?

[hUMANbEATbOX]

Originally Posted by majinwar

The green board around the intel chips, do you remove the cpu from it before putting it into your mobo? Or does it stay attached even when you put it into the mobo?

You mean the PCB (printed circuit board) around the core itself? You certainly want to leave that attached.

[tuyen]

Originally Posted by hUMANbEATbOX

Another thing to consider is that the temps are all estimations, calculated by software. By no means are the readings perfect.

The temperature readings are not estimations, nor are they calculations.
They are read directly from the thermal sensors built into each CPU. They're accurate to within a fraction of a degree.

[graysky]

Originally Posted by GFORCE100

2) At 9x333 I set the lowest voltage possible in the BIOS, which is the VID of the CPU, Q6600 #2. I set it to 1.2500V. Under Prime95 25.5 Small FFT same as before, the actual load Vcore is 1.18V and my temps, according to Core Temp 0.95.4 are 55/55/51/51. The PWM CPU fan doesn't spin 100% due to the fact the CPU temp isn't that high, according to Abit's uGuru it's only 41C. PWM temp is also very cool at 46C (these under Prime95 25.5).

9x333 @ 1.25 (1.18 measured) is too hot... you might wanna either lap them or upgrade to a ultra-120 ex IMO

[tuyen]

Originally Posted by GFORCE100

Question #1 Why are the temps so high on Q6600 #2 if its running on such a low Vcore at 3.6GHz?

Those temps are normal, given your parameters.
You're pushing your CPU at 1200 MHz over the rated spec...and you're using a less-than-ideal cooler.

My Q6600 runs at 3000 MHz right now (333 x 9) at 1.264 V with DDR2 running at 1112 MHz.
Under full load with Prime95, I get temps of 67, 66 , 63, and 61 for the cores, using a Zalman 7000 (all copper) which I took from my old Socket 478 system.
Under typical loads while playing games, the hottest core never goes above 44-46 degrees. I know I can get those temps about 10-15 degrees lower with a new Thermalright 120 Extreme...but for now I'm using the Zalman.

So given the fact that you're running 600 MHz over me, I would say your temps are well within the expected range for that heat sink. And also keep in mind that each CPU is different as far as how much it'll overclock and how hot it'll get.

Don't waste your time with lapping unless you have noticeable (and severe) curvature on the heatsink...and don't even think about lapping your CPU unless you're willing to forego the warranty.

If you're using Arctic Silver 5 (and you've applied it properly to the CPU), then you're running at near-optimum conditions. You could spend a few dollars on something like Masscoll Shin-Etsu X23 which typically gets you 4-5 degrees cooler than AS5, and is currently rated as the best thermal compound on the market.

If you want the absolute lowest temperatures on air cooling, you should pair up the X23 with a Thermalright 120 Extreme. That's the best possible solution currently available for air.

[hUMANbEATbOX]

Originally Posted by tuyen

The temperature readings are not estimations, nor are they calculations.
They are read directly from the thermal sensors built into each CPU. They're accurate to within a fraction of a degree.

Yes the sensors are accurate, but they are interpreted by software, which isn't perfect. Use Coretemp and TAT on the same machine, they apparently read from the same sensor. On some machines, the readings are quite close, on others (like mine), they are about 3-4c apart from each other.

There is NO software reading that is 100% accurate.

[tuyen]

Originally Posted by hUMANbEATbOX

Yes the sensors are accurate, but they are interpreted by software, which isn't perfect. Use Coretemp and TAT on the same machine, they apparently read from the same sensor. On some machines, the readings are quite close, on others (like mine), they are about 3-4c apart from each other.

There is NO software reading that is 100% accurate.

TAT hasn't been updated in nearly two years.
It doesn't even support quad CPUs.
It can't even properly identify a quad (it shows my Q6600 as being a Pentium M).
Bottom line - it's old and outdated, and therefore unreliable.

Coretemp is what you should be using for the most reliable and accurate readings from each CPU's core.

[GFORCE100]

Originally Posted by tuyen

If you want the absolute lowest temperatures on air cooling, you should pair up the X23 with a Thermalright 120 Extreme. That's the best possible solution currently available for air.

Well the only thing I was thinking, considering I already have the Zalman fan bracket installed, and hence don't need to take the motherboard out for this, would be a Zalman Reserator XT watercooling kit. I'm not however sure how good these are with quads, no real reviews of the thing out there sadly. It would make a lovely match to my Zalman GT1000 case, also in titanium.

I already had a IFX-14 and wasn't impressed by it, and that's 120 EX performance albiet larger and more cumbersome to handle.

If the Zalman 9700NT clip was stronger, I'm sure I could shave some more points off as it's not pushing down overly hard.

[hUMANbEATbOX]

Originally Posted by tuyen

TAT hasn't been updated in nearly two years.
It doesn't even support quad CPUs.
It can't even properly identify a quad (it shows my Q6600 as being a Pentium M).
Bottom line - it's old and outdated, and therefore unreliable.

Coretemp is what you should be using for the most reliable and accurate readings from each CPU's core.

Guess what, it reads my original e6300 purchased right at launch as a Pentium M as well.

Bottom line is, I stand by what I said. There is no perfect software reading, and your assertion that Coretemp is accurate to within a fraction of a degree is wrong. It is a nice indicator, but for truely accurate temps, you need a setup like this: http://overclockers.com/articles1312/

He uses that setup to test heatsinks as seen here: http://overclockers.com/articles373/p4sum.asp

There is a reason why he hasn't updated his test setup with a C2D+Coretemp. It isn't as accurate.