Really, 15 degrees! I admit, I don't have screenshots which I know won't be good enough for some of you, but allow me to explain.
I'll start with... where I started from: I was using an unlapped TRUE with MX-2 TIM on my overclocked i7. I fold on my i7 which produces more heat than my benchmarking utility, systester (I run native linux and am currently ironing out issues with my windows partition, so none of the usual windows utilities could be used). Folding got my proc to around 80 degrees with an ambient of 21C. I have to admit, that reading should be taken with a grain of salt. My ambient temperatures aren't terribly stable and I also think folding sometimes produces different amounts of heat, as at the same ambient I could sometimes get something like 4C changes on my hottest core pretty quickly.
Anyway, the key thing to note is that my hottest core was usually right near 80. Pretty hot, I know.
So what did I do? Well I decided I wanted to get better temps so I lapped the TRUE. First time I've ever lapped anything, and it took quite a bit longer than I expected. I started with 400grit, ran out, went down to 220 (it was what they had). Even with the 220 it took quite a bit of lapping to get a smooth, all-copper surface. After that I went to 600, 1000, and finished with the 1500. If I had more 1500 I probably would have done a bit more with that grit. I was still seeing slow changes (for the better) in the finish on the base, but I was pretty happy and while it definitely wasn't a mirror finish I could definitely get clear reflections in it. Unfortunately the camera with the pictures is about 250 miles away right now, so I can't upload them. Don't worry, I'm a horrible photographer and you aren't missing much.
Lapping alone didn't get me 15C temperature drops though, I threw in a couple other ideas. I did the washer mod with a pretty thin (~1-1.5mm), but quite large washer. I was VERY happy with the effect this had on my fully tightened pressure. The true no longer feels like its ready to slip into a different position at the slightest touch. I can definitely still twist it, but it requires a deliberate force.
I also rotated the TRUE by 90 degrees (from the standard position most people use) so that the fans are now pointing up. I did this for two reasons, one of which probably made a real difference. I came up with the idea probably because I'd read about it before (although its usually dismissed as not likely to make a difference), but also because everyone knows that hot air rises. Well now the fans are just helping the hot air rise. I know that's not really making a difference but it makes sense to me and caused me to actually look at the heatpipes in the base of the heatsink. I realized that they run parallel to the die in the normal configuration, so only two or three are actually directly above the die.
The basis of the "line method" of applying thermal paste factors this rectangular nature of the die in. I understand there's some debate as to whether the line method actually performs well, which I'll mention in a moment. I do think that putting more heatpipes in contact with the die should increase performance. Now exactly how many more heatpipes are in contact with the 90 degree turned configuration, I don't know. Maybe the die is close enough to square that it doesn't actually change anything, I can't really tell from the pictures.
The last thing that I think helped my temps drop so dramatically was the fact that I feel much better about this mount than I do about the last one. Last time I used the line method, which artificially tends to result in the use of more TIM. I think that's the essential problem with the line method. If you could actually get the appropriate amount of time to form a line on the processor, I could see it actually working better. The problem is, its so hard (at least for me, I don't know about you guys) to fine tune your control over the TIM tube that much. This time I just used the dot method, and I put the least amount of TIM I could get to come out of the tube at once. It was still just a little more than I would have liked to have used, but I think it's better than it was before, and especially with the increased pressure from the washer mod squeezing extra TIM out, much better than before.
Anyway, when I fired the system up I was astounded. With my somewhat variable temperature readings I estimate (and thus have reported to you) a 15C temperature drop, which I think is pretty accurate. My hot core is usually around 65C now with ambient of 21, and I have yet to see any temperature starting with a 7, so its AT LEAST a 10C drop. I still say 15C though, because seeing 66C is pretty rare, and my ambients are above 21 when I do see it.
I'm most interested to see if the 90 degree twist actually offers an advantage. I'd love to test it myself, but I don't really have the time with school starting back up again. Has anyone looked at this in the past?
I'll start with... where I started from: I was using an unlapped TRUE with MX-2 TIM on my overclocked i7. I fold on my i7 which produces more heat than my benchmarking utility, systester (I run native linux and am currently ironing out issues with my windows partition, so none of the usual windows utilities could be used). Folding got my proc to around 80 degrees with an ambient of 21C. I have to admit, that reading should be taken with a grain of salt. My ambient temperatures aren't terribly stable and I also think folding sometimes produces different amounts of heat, as at the same ambient I could sometimes get something like 4C changes on my hottest core pretty quickly.
Anyway, the key thing to note is that my hottest core was usually right near 80. Pretty hot, I know.
So what did I do? Well I decided I wanted to get better temps so I lapped the TRUE. First time I've ever lapped anything, and it took quite a bit longer than I expected. I started with 400grit, ran out, went down to 220 (it was what they had). Even with the 220 it took quite a bit of lapping to get a smooth, all-copper surface. After that I went to 600, 1000, and finished with the 1500. If I had more 1500 I probably would have done a bit more with that grit. I was still seeing slow changes (for the better) in the finish on the base, but I was pretty happy and while it definitely wasn't a mirror finish I could definitely get clear reflections in it. Unfortunately the camera with the pictures is about 250 miles away right now, so I can't upload them. Don't worry, I'm a horrible photographer and you aren't missing much.
Lapping alone didn't get me 15C temperature drops though, I threw in a couple other ideas. I did the washer mod with a pretty thin (~1-1.5mm), but quite large washer. I was VERY happy with the effect this had on my fully tightened pressure. The true no longer feels like its ready to slip into a different position at the slightest touch. I can definitely still twist it, but it requires a deliberate force.
I also rotated the TRUE by 90 degrees (from the standard position most people use) so that the fans are now pointing up. I did this for two reasons, one of which probably made a real difference. I came up with the idea probably because I'd read about it before (although its usually dismissed as not likely to make a difference), but also because everyone knows that hot air rises. Well now the fans are just helping the hot air rise. I know that's not really making a difference but it makes sense to me and caused me to actually look at the heatpipes in the base of the heatsink. I realized that they run parallel to the die in the normal configuration, so only two or three are actually directly above the die.
The basis of the "line method" of applying thermal paste factors this rectangular nature of the die in. I understand there's some debate as to whether the line method actually performs well, which I'll mention in a moment. I do think that putting more heatpipes in contact with the die should increase performance. Now exactly how many more heatpipes are in contact with the 90 degree turned configuration, I don't know. Maybe the die is close enough to square that it doesn't actually change anything, I can't really tell from the pictures.
The last thing that I think helped my temps drop so dramatically was the fact that I feel much better about this mount than I do about the last one. Last time I used the line method, which artificially tends to result in the use of more TIM. I think that's the essential problem with the line method. If you could actually get the appropriate amount of time to form a line on the processor, I could see it actually working better. The problem is, its so hard (at least for me, I don't know about you guys) to fine tune your control over the TIM tube that much. This time I just used the dot method, and I put the least amount of TIM I could get to come out of the tube at once. It was still just a little more than I would have liked to have used, but I think it's better than it was before, and especially with the increased pressure from the washer mod squeezing extra TIM out, much better than before.
Anyway, when I fired the system up I was astounded. With my somewhat variable temperature readings I estimate (and thus have reported to you) a 15C temperature drop, which I think is pretty accurate. My hot core is usually around 65C now with ambient of 21, and I have yet to see any temperature starting with a 7, so its AT LEAST a 10C drop. I still say 15C though, because seeing 66C is pretty rare, and my ambients are above 21 when I do see it.
I'm most interested to see if the 90 degree twist actually offers an advantage. I'd love to test it myself, but I don't really have the time with school starting back up again. Has anyone looked at this in the past?
What is a washer mod? Does it apply to water blocks also? 
