CL-P / U, MX4...
...I have been putting the finishing touches on the 'siamesed' RIVE and MVE build...got 'a little bit sidetracked' putting the 3970X through some early competitive HWBot stuff (success)...but now back to the 'cosmetic improvements' re cable and cooling tube management...in the process of building this up, I removed the 3770K in the MVE which was running CL-U on both the die and the IHS...after 3 months or so, CL-U did not leave any visible marks
(serial # etching had already been cleaned off in prior mods), and a liberal application of Isopropanol on Kleenex cleaned up ALL traces
of CL-U from the previous application.
...both the 3770K and 3970X are great low-v performers (both validate at 1.36+ - v @ 5 GHz) but both had VERY concave
IHS' (a relationship to low v ? ?
) - way beyond what my other chips have... It is there where CL products can pose a problem...for the 3770K, I got around that as I did dome serious lapping of the IHS, starting with 600 grid and finalizing with 2000 grid sand paper...once done, a new thin layer of CL-U on both die and (lapped) IHS dropped temps even more...combined with the siamesed triple-rad loop, 5 GHz at Cinebench has a.) fairly uniform core temps and b.) stays in the high 50ies C range for max core temp now.
...but the 3970X initially had some temp issues
with even a 'solid' application of CL-U..
..in spite off correct water-block to IHS mounting, there clearly remained a 'cavity' between the IHS and the block. I knew from previous experience that MX4 is much better (and safer!) to fill such cavities, and once I applied MX4, temps went down dramatically...by over 20 C in Cinebench at 5 GHz for the Sandy-E. I suppose I could have applied 'more CL-U' into the cavity, but the problem is dosage...too much and it will start to run out and hit electronically sensitive components...not enough and the cavity remains in some shape or form...and lapping a brand-new $1000+ CPU was not in the cards (...yet, may do so later)...also keep in mind that the 3970X has a much bigger die / IHS than the 3770K (and thus potentially bigger cavity with a concave IHS). So MX4 solved the problem nicely.
...per multiple earlier post I have done on CL and MX products, both my own experience and repeatable YouTube experiments between CL-P, CL-U and MX4 suggest that CL-U has the best temp performance, followed closely by CL-P and MX4 (in some tests, MX4 actually beats CL-P)...bottom line: If your IHS i/ water-block is not too concave / convex and involves no aluminum, CL-U is the best choice, followed by CL-P...but if there are such shape issues and you don't want to lap (and in the process remove CPU serial info etc), then go for MX4...
...finally, whether CPU or GPU get TIM applications, in all cases, copper or Nickel/copper are 'ok' for CL and other liquid metal products, aluminum is definitely NOT...some caution is in order as many GPU 'blocks' actually have flattened copper (or nickel/copper) bottoms that are embedded in an aluminum assembly, with some minor amount of aluminum also touching the die area...in that case, either stick to MX4 entirely, or apply it on the aluminum bits (usually very thin stripes) before applying CL products on the die. Edited by Joa3d43 - 4/26/13 at 1:19pm