June 25, 201312 yr Bought time... Took long enough... :lol: Thanks Stephen and Paul. May this help to make my journey into the unknown a little less scary. And Thanks to Max for the OP!
August 15, 201312 yr Just de-lidded my 4770K, using the hammer+vice method. I then applied Coollaboratory Liquid Pro between the IHS and die. Running Intel BurnTest 2.54, temps dropped from about 86C to 71C (temps for individual cores peaked at 68 / 69 / 71 / 71 C). Running the new LinPack 11 with AVX2 support, temps were 100+ C (6% throttle) and dropped to 81C after de-lid. That's at 4.3 GHz / 1.218V. -
August 15, 201312 yr Author Just de-lidded my 4770K, using the hammer+vice method. I then applied Coollaboratory Liquid Pro between the IHS and die. Running Intel BurnTest 2.54, temps dropped from about 86C to 71C (temps for individual cores peaked at 68 / 69 / 71 / 71 C). Running the new LinPack 11 with AVX2 support, temps were 100+ C (6% throttle) and dropped to 81C after de-lid. That's at 4.3 GHz / 1.218V. What cooling do you use? Those temps are still very high, though the drop from stock is decent.
August 15, 201312 yr What cooling do you use? Those temps are still very high, though the drop from stock is decent. Phanteks PH-TC14PE. With AVX2 Linpak, the drop was almost 20C which is pretty good. The only way to accurately test Haswell with AVX2 is to use this: http://software.intel.com/en-us/articles/intel-math-kernel-library-linpack-download You have to disable Hyper-Threading, then run linpack_11.0.5\benchmarks\linpack\runme_xeon64.bat If the CPU supports AVX2, It's extremely demanding and will run about 10-15C hotter than Prime95, Aida64 and other "lite" stress tests. You will however get over 200 GFlops on Haswell thanks you AVX2. A lot of overclockers just run Aida64 with CPU, FPU, RAM and Cache all checked, and don't realize that they interfere with each other, resulting in lower temps. -
December 30, 201312 yr You definitely need to apply TIM on the top (outer) portion of the lid. It should resemble something like this: CPU DIE -> TIM -> LID -> TIM -> HEATSINK/WATERBLOCK As for TIM, you should use the same for both. Like Tech states, Coollaboratory Pro is the best, though I use the Ultra variant and have zero problems with it. Recommendation... If you or anyone is going to go through the trouble of de-lidding, go ahead and lap your lid and heatsink too! Gives you peace of mind knowing you did everything possible you could to get the best temps for your OC. Trust me, if you don't, you will always to wondering if you could squeeze an extra C degrees off. This leads to disassembling and re-assembling which leads to problems in the long run. Do it all and right from the start. Just a not to those that have not lapped anything. The process is demanding, knowhow, and is time consuming. If you have never lapped any CPU or GPU I suggest you do not do it. Also, Intel CPUs are manufactured concave on purpose and many cooler heatsinks are convex to fit Intel CPUs properly, and therefore lapping is not needed. Note, I have not been on this forum for so long that I had to re-register, however, my avatar was kept. Go figure. Also, I have been overclocking for a very long time.
January 1, 201412 yr Author Just a not to those that have not lapped anything. The process is demanding, knowhow, and is time consuming. If you have never lapped any CPU or GPU I suggest you do not do it. Also, Intel CPUs are manufactured concave on purpose and many cooler heatsinks are convex to fit Intel CPUs properly, and therefore lapping is not needed. Note, I have not been on this forum for so long that I had to re-register, however, my avatar was kept. Go figure. Also, I have been overclocking for a very long time. IMHO, lapping is worthwhile, but I push my equipment hard and I don't mind taking an hour or two to get the most out of them. I saw in the neighborhood of a 10 degree drop after lapping.
January 1, 201412 yr To TechguyMaxC, Thank you for your response. Although not written well the thrust of my comment regarding lapping is that a novice should not try it. It is too easy to NOT to get a flat surface on a large air cooler heat sink. I have lapped CPUs and it seems easier to lap the CPU properly than it is the base of a large air cooler. I did not want people new to lapping that it is an easy process, a novice can easily lap improperly. Also, with a good air cooler I have found that proper airflow in the case, and using good TIM properly are very important to keep temperatures down. And, I think most novices use air cooling, although prebuilt water coolers are gaining in popularity.
July 25, 201411 yr Hi everyone! I am going to delid my Haswell i7 4770k using the vice method and I currently have Coolaboratory Liquid Pro. Though this post was mainly for Ivy Bridge, can I just follow it for Haswell? I'm really new to all this so sorry if asking dumb questions. After applying CL pro on the die, do I wait for it to dry or do I just proceed with putting the IHS on and then remounting it? I think I will also apply CL pro on the underside of the IHS because I have not read anywhere that the spacing issue between the IHS and die changed since Ivy Bridge. Does anyone know of anything new regarding this? Is it okay to just rely on the mounting bracket to hold the IHS and cpu together since I will clean off the adhesive material? Or should I use another adhesive after delidding and putting on CL pro to hold the die and IHS together before remounting? Thanks Will I may be answering one of my questions but perhaps maybe I shouldn't let the TIM dry before putting the IHS back on since the whole point of using TIM is to allow better contact (due to uneven surfaces or the tiny gaps) between their surfaces. And so it would be better to join them together asap to form as much contact as possible. Is this reasoning correct? Will
February 27, 20179 yr Author I just wanted to update this thread to let anyone who's curious about this process know that it is indeed still possible with Skylake and Kaby Lake chips and still delivers fantastic results.
Archived
This topic is now archived and is closed to further replies.