Cooling lapped my ultra-120 extreme (pics and temp results)

When my Ultra-120 X and I have to say I'm a little puzzled. The base where it should contact the heat spreader is not smooth at all, it's actually grooved! You can see a scratch which is where I gently ran my thumb nail over the surface; I could feel the rough edges.

Have a look for yourself:


Anyway, others encouraged me to lap it which I've never done before. After wrestling with the idea for a couple of days as well as reading many articles/guides, I decided to give it a go. $20 worth of sandpaper, a $2 piece of flat glass, and 4 hours of careful work (and sweat) later, I was left with a pretty darn flat HS. You can see by the pictures that this particular one was quite concave instead of being flat which isn't good for keeping contact between the HS and IHS of the CPU.



Did it work you're probably wondering. The temp data as measured in coretemp for a ~1 h x264 encode (uses all 4 cores with a CPU load of >99 %). I had coretemp log the temps and I averaged the whole data set per core for the 2nd pass of the 2-pass encode (the 2nd pass is the most CPU intensive). Room temp for both experiments was ~23 C.

Code:

Before lapping the HS:

Core 0: 51.9
Core 1: 51.4
Core 2: 45.6
Core 3: 45.6

After lapping the HS:

Core 0: 49.9
Core 1: 49.4
Core 2: 44.0
Core 3: 44.4

Delta:

Core 0: 2.0
Core 1: 2.0
Core 2: 1.6
Core 3: 1.2

System specs: Q6600 @ 9x333=3.01 GHz (stock voltage), P5B-Deluxe in an Antec p182 case.

EDIT: there is a companion thread to this one wherein I lapped the IHS on my Q6600 too. This gave the largest gains in temp reduction. Read it here.

 

Why are your temperatures different?
http://www.tomshardware.co.uk/forum/235216-11-lapped-ultra-extreme-pics-temp-results

 

I was going to say "ooooo busted" untill i read the one on toms.

"By the way, I added a constant of 15 to each core in speedfan since it incorrectly displays temps for quads by 15 °C. "

 

Not worth lapping then

 

Really?

Combine this with the lapping of the Q6600 in his other thread and the temp drop on core0 is over 10%. People will often pay a great deal of money on an air cooler/water cooling to achieve those kind of results.

I say: welcome to the forums graysky, congrats on your results and thanks for sharing them with others here who may benefit from doing the same

 

I don't think it's worth the risk of messing up a decent HSF and/or the 4 hours solid, hard work required for a 2'C delta.

 

I wouldn't really say it was time invested well either. But the sake of shiney'ness. That's always a bonus! And welcome to the forums.

 

Goes to show that Thermalright makes better surfaces than Intel.. Damn IHSs.

Anyway, just please don't go bashing him about poor results and what not, it's called enthusiasm and it's good he posted this. The cooler doesn't really have any warranty to be destroyed (how could it break up w/o own mistake?) so why not?

 

No, no, no - fair play to the OP, anyone who laps their HSF's properly has my respect, but there's a big difference between someone doing it of their own accord and recommending everyone starts doing it.

 

Pfff. Just get watercooling.

**actually, good job. If I had air, I'd rather lap my heatsink than my CPU**

 

I'm with WiskeyAlpha on this, sorry. If you consider that the maximum safe operating temperature for a Q6600 is 62C, and grasky was registering temps up to 52C/67C then even a few degrees reduction is a good thing indeed (I'm not sure that they would reach 67C though, grasky, unless you have a G0 stepping CPU? I think you need to check that measurement error adjustment. Try a comparison with CoreTemp's report, which is generally spot-on accurate).

Exactly. This is a modding forum, remember? When we mod, we experiment. Sometime we see big results, sometimes we see small results. The important thing about experimenting is that it is done, so that we know rather than assume. Thanks, grasky.

 

Nobody is 'bashing him'. All i was saying is that as far as i am concerned it is a waste of time!

 

people who 'bash' are to too scared to mod, or tbh to lazy! lol!

 

I think experience shows over the years that most software temperature monitors (or maybe more to the point the diode in the die) rarely report with any great accuracy. IIRC, AMD stated that the temperature diodes on their X2 line of processors were accurate to +/- 10C. In terms of overclocking that could be the difference between a very cool chip and a very hot one. I think that the newer processors (i.e. conroe +) are more accurate, but then recent controversy regarding the delta between cores on the Wolfdale (E8xxx) line has once again lead me to question the accuracy.

Granted, the real use comes when comparing the temperature delta in a "before and after" scenario, exactly like graysky has here, and to a lesser extent checking that you are operating within "safe" temperatures. It's almost pointless comparing temps with others though. You could be reading higher but actually have a lower temp than someone else.

As overclockers, I see it as a risk we take. Without using a calibrated laser thermomitor (and even then you can't get the "actual" die temps) how will you ever know? Provided I'm 100% stable and my temps aren't absurdly high, I'm happy.

That's hit the proverbial nail on the head Nexxo. It's something that a lot of people seem to be forgetting; and of late the forums look more and more like a technical support forum than a modding one. Don't get me wrong, I asked my fair share of questions when I started out at bit-tech (and still do ) and we have a great deal of people on bit-tech with in depth technical know-how and it's great to share that with others. However, I find some people's reaction to this post surprising at best.

Rather than welcoming the poor fella to the forums and giving him some credit for sharing his work with us (which could stand to help many people here) he gets canned and told it's "not worth it then". He never said everyone should do it or that you'd be silly not to, yet he seems to have got the reaction that he was silly for trying.

I recently volt modded my GTX and the result is that it now sits highest on bit-tech's 3DMark06 leaderboard. Would I therefore recommend that everybody with a GTX do the same? Of course not. But I shared it with others on Bit-tech as it is a Modding forum as Nexxo pointed out. At the same time I'd be willing to give anybody who wanted to try the same my support and advice without hesitation. However, if I'd had replies similar to graysky, I would've been more that a little dismayed.

/RANT

Ah, that's better

 

lapping my TRUE is always one of the things on my TODO list, but i can never be bothered to do it.......

what i want to know is that why after lapping, there's still a temperature difference between Core0,1 and Core2,3? if i were to lap my TRUE, that is exactly the issue i want to resolve.

 

Even if the temp measurements were accurate, there could still be a lot of factors between the cores, things like position on the die are particularily important. Then you've got position compared to the socket, orientation with respect to the HSF... Many factors that I have no idea about too. Basically, lots.

 

no matter how many re-sits, it's ALWAYS higher, i always thought it's because of my TRUE not flat....... doesn't seem to be the case here

 

Well external factors other than your HSF can't be excluded - even if it's in exactly the same place, at maximum effectiveness, it won't eliminate other effects.

 

Yeah, speedfan (which I no longer use) did have the incorrect constant that was 15 °C off. Coretemp on the other hand like HWMon does not have this problem with the Q6600.

Right on, dude. The whole reason I spent the time to photograph the process and write up the posts is to help spread the knowledge to other folks either too timid or try it or to those who don't even know what it means to lap a surface.

I actually wrote this up into a formal article you can see over at techarp at this URL.

One comment I'll make is that if you lap the HS, you should lap the IHS on the CPU as well since both surfaces really need to be flat to maximize your thermal xfer. For those who missed it, you can read about the IHS on my Q6600 in this thread. More importantly, have a look at the colorful graph I made comparing unlapped/unlapped vs. lapped/unlapped vs. lapped/lapped (HS/CPU). It really made a difference for my machine and you can bet that I'll do this on future builds as well

 

see next post.