Finally ready to start putting up some numbers.....and there are a lot of numbers to be put up
Let's run down what I tested and how I tested:
I have two different CPUs for testing, an extremely hot running E6700 and a very hot running QX6700. I was spending a lot of testing time up at TjMax on both processors....and one of the HSFs couldn't even hold my E6700 at or below TjMax with the lowest fan speed (though it was still totally stable....just the temperature data was useless). I ran the E6700 at 3555MHz (395x9) at 1.56v loaded on a Blitz Extreme. I ran the QX6700 at 3160 (395x8) at 1.46v loaded on the same Blitz Extreme. When swapping processors, only two BIOS settings were changed: multiplier and voltage. This was for simplicity....neither CPU is anywhere near maxxed out for being Prime/OCCT stable at these clocks, I just wanted to consistently create a lot of heat and have it run regardless of the temps it was enduring. On request, I can list more complete BIOS settings....I am, admittedly, pushing on the northbridge pretty hard (200 strap and 1.74v for starters), but it shouldn't affect results.
I used OCCT to load and record temperatures....it was getting essentially identical load temps and temperature outputs to a manual test of Prime95 + CoreTemp, but it's fully automatic and repeatable, and therefore more desirable for this kind of testing. Linpack was not used because, well, it was too hot--most of the time it would go over TjMax in no time at all (making all data useless), and sometimes it would crash (never fun). I suppose the next batch I can use Linpack implemented via OCCT 3.0b and use some lower clocks/voltages
I have four HSFs here to test: a new Copper TRUE, an older TRUE which I have already lapped (amateurishly....just to rid it of its poor stock base), an older unlapped IFX-14 (the backside cooler was not used in this test, it doesn't do much in my experience though), and a new Xigmatek 1284EE with stock push-pin mounting and the "Crossbow" hard mounting.
On each set up, I tested 8 fan setups: 1) single Noctua P12 (1300RPM), 2) dual Noctua P12s, 3) single S-Flex F (1600RPM), 4) dual S-Flex Fs, 5) single Ultra Kaze 2000 (2000RPM), 6) dual Ultra Kaze 2000s, 7) single Panaflo U1A (2750RPM), and 8) dual Panaflo U1As.
Each test was repeated on two separate mounts and the data has been averaged. I really wanted to do three or more, but considering how much time two took, three was out of the question for my sanity, let alone more than three. Intake ambient temps were taken throughout testing and used to calculate deltas. I tried to maintain them between 23-24.5C for all testing to minimize the amount of error incurred by varying ambients.
I also did a few modifications of the HSFs, and are noted when appropriate. Here's a summation of notation:
Lapping1 = amateurish lapping job.
Lapping2 = a nice lapping job (mirror finish).
Lapping3 = significant base material removed and then lapped to a mirror finish.
Hi-Pres = washers or duct tape were used to increase mounting pressure significantly.
I've graphed four different sets of data extracted from the same tests: average sustained load, peak sustained load, average short load, and peak short load.
Average sustained load is the average of dTs over the span of 0:20 and 10:00 of the OCCT load.
Peak sustained load is the maximum delta between the span of 0:20 and 10:00 of the OCCT load (any core of the 2 or 4).
Average short load is the average of dTs over the span of 0:00 and 1:00 of the OCCT load.
Peak short load is the maximum delta between the span of 0:00 and 1:00 on any core of the OCCT load.
Some notes:
1) E6700 is lapped to a mirror shine, QX6700 is not lapped at all.
2) Pushpin 1284EE was not run on the QX6700 because one of the pushpins refused to maintain pressure that far along into the testing. I'm actually a big fan of pushpins--easy to mount (takes 5 seconds for any HSF), but these were amazingly low quality. I've enjoyed using pushpins from Intel and Scythe and Thermaltake sinks, but these were just bad, I can see why people have wanted to upgrade to the Crossbow.
3) Lapped 1284EE was not run on the E6700 because it performed worse than stock on the QX6700 and I didn't waste my time since it already performed so poorly on the E6700. Sorry (well, not really).
4) I really want to take my regular TRUE's baseplate down to within a millimeter or two of the heatpipes, but after a long, long time lapping just to get to "Lapping3" I felt it was unpractical to such significant time for so small of a potential gain.
5) All HSFs were bought via retail channels, none of the manufacturers know who I am, let alone arranged a review with me--hopefully these results are fully indicative of what you can expect from retail product.
6) My tests are just one set of results and while I've put a lot of effort into ensuring the quality of the data--it doesn't make other legitimate results any less valid, even if they disagree.
7) My main direction in this review isn't to find what is "best" but rather what is most useful in certain situations.
8) Ceramique was used and allowed no curing time. Upon removal, all mounts were considered at least "good" based on visual inspection. For the
'normal' HSFs, I used a dot in the center and for the HDT, I used a line perpendicular to the heatpipes.
9) All single fan configs were in pull (and the IFX-14 had the fan in the center) while all dual fan configs were in push/pull (with the IFX-14's fans in the center and the 'far' end of the airflow).
...So how about some results?
Sustained load tests first:
E6700 395x9 1.56v loaded, averages
E6700 395x9 1.56v loaded, peak deltas
QX6700 395x8 1.46v loaded, averages
QX6700 395x8 1.46v loaded, peak deltas
Short load tests:
E6700 395x9 1.56v loaded, averages
E6700 395x9 1.56v loaded, peak deltas
QX6700 395x8 1.46v loaded, averages
QX6700 395x8 1.46v loaded, peak deltas
Reply With Quote
Originally Posted by Apokalipse
Ive been wondering how a copper true compares to a normal true performance wise. Man, this must have taken forever. I dreaded haveing to do the half asked lapping job on my true. This would have driven me insane.
You might try it with the Backside cooler and 3 fans, what would be the difference
Bookmarks