Explain yourselves. Drilling HS for a probe....

[omaticrail]

I love this site, and this forum. Overclocking and CPU cooling are seriously habit forming. However, I have a small gripe, or maybe just puzzled thoughts that need direction.

I just finished reading several of the o/c.com articles on thermal testing and the questionable methods by which people measure and interprete temps. While it all sounded very good, something doesn't jive at all.

People will take a perfectly good HSF or waterblock, and spend hours carefully lapping it to a mirror perfection, then apply the most expensive thermal grease, and employ some perfected clamping method, all to achieve the optimal thermal mating of two components. All this is done to achieve a measurable improvement in core temps, and eek out one more notch on the overclocking hall of fame.

NOW I read that those same people are perfectly willing--even suggesting--drilling a tiny hole in the HSF to house a thermalcouple for optimal temp reading. I'm not sure how small a hole a thermalcouple can slip into, but even a 1mm hole covers about 1% of the core's surface, and you're putting it right at the hottest point. Filling a visible hole with AS3 is most certainly worse than using the same compound to correct the block's original, non-visible imperfections you spent so many hours getting rid of.

So, what am I missing here? Did these guys key their own car-door to demonstrate how good a paint job they did?

 

[rivercom9]

I dont know what some people are thinking. I wouldnt drill a "tiny" hole into my cpu. Just too dangerous for me and it seems that the means is not worth the effort.

 

[Hoot]

Having drilled such heatsinks as my highly valued MC-462a for a thermocouple, I conceed that yes, you lose 1% of the surface area, but I assure you it has very little impact upon the overall cooling effectiveness. It can not be equated to having a rough finish on the heatsink. All those micro-grooves or scratches add up to some serious loss of contacting surface area. They may be small, but they make up for that in quantity.

Hoot

 

[SkyHook]

Not In The CPU

They are not drilling the hole in the CPU, they are in most cases drilling as small a hole as is necessary in from the side of the heatsink base to locate the tip of the probe just above the center of the core. At least that is the method that has been explained to me in past conversations.

 

[racecar12]

I just set my probe next to the cpu core. I even trim the plastic coating back to get it that extra mm like this.

 

[Diggrr]

AMD pic from their own pdf titled:AMD Thermal, Mechanical, and Chassis Cooling Design Guide

Personally, not having a drill press, I'd do the 'edge touching' method. Plus, I don't have a #53 drill bit.

 

[rivercom9]

Diggrr, what is that design guide for? For temp controlled fans for something based on the temp of the base of the hsf?

 

[Spartacus51]

Assuming I do this (although it may be difficult on a cross-drilled WB, I already drilled a lot of holes) Where do I get a probe, and for how much?

 

[Diggrr]

Just for installing an accurate thermocouple on athlon systems.
There's also guides on other ways to do it, the one I gave was the origional one that I found months ago.

Another method they outline is to use an ultra slim probe, and thread it between the socket and mobo surface on the side opposite the locking lever. Use special tape to tape and thermal goop, fixing the thermister to the exact center of the underside of the cpu.

The link to their tec pdf's is here ...sit back, relax, and read yourself to sleep.

 

[Malakai]

get a compunurse. like 15-18 bucks on various websites listed on the frontpage
-Malakai

 

[Diggrr]

That's what I'd do too Malakai, I wasn't posting the above info to say that's the way it has to be done, just to clear up that "core drilling" is NOT the way to do it.

Using a DigiDoc, or CompuNurse is the way I'd go, just puting the sensor at the core's edge like racecar above is.
Though I do wish I had Hoot's tools to play with! I'll submit, he's probably got the accuracy thing down pretty well with his methods.

 

[Mr B]

AMD pic from their own pdf titled:AMD Thermal, Mechanical, and Chassis Cooling Design Guide

Personally, not having a drill press, I'd do the 'edge touching' method. Plus, I don't have a #53 drill bit.

Borrow the #53 from Bill the Cat in your avatar.....

 

[rivercom9]

LOL Mr.B. Is that the name of the cat in Diggrr's avatar? Where is he from?

 

[Diggrr]

He's from an 80's comic strip in the newpaper..."Bloom County". I used to read it every morning in government class in highschool, cause the teacher always had the paper.
He acted alot like BobCat Golthwait after a good knock in the head (if you can imagine that)

 

[The Highlander]

I dont think is that effective or even necessary to drill a hole on the HSF. Hello we just want to have a relative mesurement...

If I drop 5C on the heatsink itself it means i dropped 5C on the core...

The temp of the HS may not be the same as the core but the differential are the same (the errors are cancelled, that is why many specific mesurements are differences)

The point of it is that if you measure your CPU temp at the farthest point of the HS and it measures 38C at full load and it locks up, maybe you need to lower the temp... If you drop 5C on the same spot... You will drop 5C on any spot providing that is the same initial and final measurement spot.

So i dont think is that necessary to drill a hole.

If we were writing a scientific article about the C/W of the processor then we might need to...

If we are comparing 2 heatsinks... the difference is always enough and perhaps with the proper mathematical (statistical) calculations we figure out by how much.

Just my oppinion...

 

[vandersl]

Not true, Highlander.

That wouldn't even be true if temperatures dropped off linearly, much less exponentially.

A thermistor embedded in a HS 5mm above the core will not read a temperature remotely close to the core temperature. It might read 1/2 the core temperature. So if the core was 20C above ambient, it might read 10C above. Raising the core temperature 6C would raise the thermistor temperature 3C.

This is what we so lovingly call 'temperature compression'.

But I agree - leave the hole drilling to the heatsink reviewers

 

[rivercom9]

I must agree with vandersl on the temp readings.

 

[omaticrail]

racecar12's approach looks far more apealing than drilling. How did you hold them buggers in place? Thermal epoxy? Can they be connected directly to the mobo sensors header, or do I need to assemble one of those diode readers between the two?

 

[The Highlander]

Lets say you get a second increment of 6C on the core and you get 3C on the heatsink up again.. Its still a relative measure and you have risen by 2x the same amount. That is why people calculate the C/W of the heatsinks with processors emulators!!!

THe heatink at the contact point will be almost the same temp that the processors... Themochemistry. The reason you dont get such a linear approach in temps is because the temp is disipated and that is why you get such an exponential form... But you can always adjust your curves with some math and you get a conversion factor and you will get your linear diff in temps.

If you get an raise by infinite on the core's temp you will also get a raise in temp by infinte.

The reason you get that so called "compresion temp" is because all metals have different thermal resistances... and it takes a while for the temp to get from one place to another.

 

[racecar12]

racecar12's approach looks far more apealing than drilling. How did you hold them buggers in place? Thermal epoxy? Can they be connected directly to the mobo sensors header, or do I need to assemble one of those diode readers between the two?

I use a small dab of clear rtv that can be removed easily to secure the probe to the cpu. I am not sure how you would setup the prob with the motherboards SMB. I have got a couple of these Enermax temp monitors which gives you 2 probes and monitors along with a hard drive bay for 16 bucks at newegg.

But when setting up the sensor, make sure the wire leading away from the probe is not too fat possibly causing the heat sink not to sit properly on the cpu core. When you trim back the platic shield to get that extra mm, it brings the solder point, where the probe is soldered to the wires, and heat shrink in just far enough to cause resistance. Just cut the first layer heat shrink off. The individual soldered wires have heatshrink to protect them and that cutting off the first layer was enough for my w/b to seat properly. Alway double check thought. It would be a shame to loose a cpu for something like this.