Relidding adhesive woes
I have a der8auer delid/relid kit waiting for me at home along with Thermal Grizzly Conductonaut liquid metal, some clear nail polish and some masking tape in prep for delidding my 8700k.

I couldn't find the Loctite Super Control gel that RockitCool recommend and it seemed wasteful to buy a giant tube of black silicone just for a few dabs on the corner of the IHS, especially since you can't effectively reseal it.

So I bought UHU Strong&Safe All-Purpose adhesive because it states it can handle temps up to 100°C.
https://www.amazon.com/PURPOSE-ADHESIVE-SUPER-Strong-Safe/dp/B004QMADSW

Now I've read a million opinions on superglue and relidding, some say it's okay, others saying it's hard to delid again to replace the LM in a few years, still others say acetone usually does the trick, some even claim superglue is corrosive (??). So can I use this glue or what?
最近の変更はElthraelが行いました; 2020年7月29日 3時26分
< >
16-16 / 16 のコメントを表示
I got the Loctite 5910 black silicone gasket maker I posted above. Worked great.

TL;DR - delidding is totally worth it. Average temps dropped by 13°C, fan RPM dropped to 70%, the system is cold and dead silent.

So yeah, I delidded the processor on Monday and ran the first tests on Tuesday. Wow, what a difference. All info below is average temp of all 6 cores measured with HWinfo64 after a 2 hour stress test with RealBench (I find it more realistic than Prime95 since it stresses the whole system not just the CPU).

Before the delid, my temps with fans set to 100% at all times and stock settings (i.e. 4.3GHz all core turbo with voltage around 1.0 to 1.1V) averaged around 65°C with peaks to the low 70s.

Post delid, I set my fans to a default "silent" present on the mobo and my average temps were 59°C with the highest peak being 64°C. Now that might not be the big "OMG 20 DEGREEZ" gain that some people report, it's "just" 6°C, but the takeaway is this: before I reached 65°C with high-ish peaks with fans blasting at full speed. Now I have a comfy 58°C with fans at virtually silent.

The OC results are more impressive, however.

I found that the only stable overclock pre-delid was 4.9GHz all core with 1.32V idle Vcore and an AVX offset of -200Mhz, which under load reached a very stable 1.296V due to lvl 6 LLC. I found it needed a minimum of 1.344V under load for 5.0GHz, and pushing the voltage any further caused the CPU to throttle - it was just too hot, reaching the TjMax of 100°C on Core 5 after just 15 minutes.

So let's first compare apples to apples.
4.9GHz @ 1.32V (1.296 under load), fans set to 100% constant, 2h stress test.

Pre-delid average temps: 85°C
Post-delid average temps: 72°C
Change: -13°C

Pre-delid maximum peak temp: 92°C
Pre-delid maximum peak temp: 76°C
Change: -16°C

I was happy with this, but then I said screw it and went to see where the limits are. I couldn't get my chip to run stable 5.2GHz even at 1.42V (didn't want to push it further because at that voltage the overshoots from LLC were 1.44V which is close to the safe maximum voltage), so I instead opted for a nice, comfy 5.0GHz @ 1.38V, 1.36V under load (still LLC 6). No AVX offset. I then cranked the uncore ratio to 1:1, which is technically a cache overclock since stock is 46 max. Then I went and applied a custom "silent" PWM fan curve and gave it another whirl.

So here's what I consider the final, "optimal" result:

Average temp after 2 hour stress test: 80°C
Max peak temp: 85°C
Change vs. pre-delid custom OC: -5°C (smack bang in the middle)
System noise: "girlfriend doesn't notice the PC is on even at full load" levels :cleanfloppy: :lunar2019laughingpig:

So basically, I went from a system that was uncomfortably hot at 4.9GHz OC even with AVX offset and fans at constant 100%, to a reasonably cool (considering voltage and performance) and very silent constant 5.0GHz system. Or, I went from a warm stock system to a very cool and dead silent stock system. Totally worth it!

BTW, since I'm running 1440p, I'm very rarely CPU bound so the 5.0GHz OC makes very little difference. At 1080p, there was a HUGE difference (like 20+FPS average) in Far Cry 5 and DOOM Eternal, but on 1440p, the difference was more or less zero, zilch. So I won't be keeping the OC, but I have saved the preset if I decide to upgrade my GPU (which might be soon since Ampere drops in a few months, apparently).
最近の変更はElthraelが行いました; 2020年8月6日 10時27分
< >
16-16 / 16 のコメントを表示
ページ毎: 1530 50

投稿日: 2020年7月29日 3時24分
投稿数: 16