Delidding 4790K thread

For die to IHS, used 1/2 rice grain and used the brush. For IHS to heatsink, I used about 1 1/2 rice grain and used brush again. IHS is huge, so needed a lot more CLU than I expected. Did the same for GPU.

But I may remove it later and use regular TIM on the IHS and GPU, leave the CLU under the IHS since I don't like the idea of CLU drying up and welding to copper.

 

i hear ya, id probably do the same, thus make sure to have a nice clu application between ihs and die and then use a regular tim for gpu and ihs/heatsink

Sent from my Nexus 5 using Tapatalk

 

Well my CPU arrived safe and sound to the guy doing the delidding. A week from now we will have two data points using CLU from the chip to IHS and gelid from IHS to sink.

 

I still wish Meaker would post a diagram of how he would attempt direct to die.

 

Meaker works in mysterious ways....

 

@Zymphad

@Jaybee

It takes time to create great art, especially when no one commissions it


There, a Meaker diagram especially for you. Not to scale obviously.

 

Oh ok, add some non-conductive material back of the motherboard and put the backplate back on. Hmm that would be interesting.

Has anyone done this before? Even on a desktop? Just trying to figure out how you would keep the CPU from moving once you do this, since the holder presses down on the IHS.

Also I won't be trying this as it would involve removing the entire motherboard, and I am lazy.

This might be something Mythlogic for example may want to provide as a service to those why buy from them.

 

its another meaker-diagram, classic!

Sent from my Nexus 5 using Tapatalk

 

I'm sure it will appear again some time in the future

 

Delid is complete here is what he sent me post delid. I'm not familiar with the chip in a desktop so no idea of what he is saying is true or not but anyways here it goes


"Your delid is complete and here are your results. NOTE: My water cooler took a dump, so this was tested with a STOCK intel cooler which is ****...but as you can see with a 4.4 GhZ (I had C states on so the frequency changed all the time on realtemp) your overclock your idle temp was around 25C and max temp is 85 C on the hottest core... So this delid went very well normally on a stock cooler at stock speeds you can hit 100C at times so if this were on a water loop I'd think the temps would be at around 65-75 C for the same setup. Note being that is in a laptop I have no idea how it will perform, but if you can get indigo extreme to work you could gains another 3-5 C on cooling. Note the 4 corners are all that have been sealed as requested delid went smoothly.. "

I will have the chip back Thursday or Friday at which point I will post before and after temps at exactly the same settings.

Edit added picture of temperatures..

 

Don't use CLU from IHS to heatsink. CLU does meld with the copper, avoid it.

Also CLU doesn't make good contact between IHS and heatsink, too much imperfection in our heatsinks, but great for die to IHS, using leftover Noctua NH-1 I had around, it's making better contact and I'm getting 41-43 C idle, 52-54 watching Netflix. Impressive.

Delidding was worth it.

 

If you have the temps saved it would be nice to see the before and after temps in something like the XTU stress test using the same settings...

 

Sadly did not save.

 

That paste is notorious for pumping get out in low pressure mounting situations so I would suggest ordering some GC Extreme and using that in place of the Noctua paste or you'll notice temps increasing over time.

Crazy delidding the 4790k, I have one in my desktop that replaced a really bad 4770k (wouldn't even do 4.3 totally stable) and have it at 4.7GHz 24/7. Great chip.

 

I'm done with this delidding thing. Used Coollaboratory Liquid Ultra between die and IHS. And used Phobya Nanogrease Extreme between IHS and heatsink. After 30 minutes of P95, I got 92C with no thermal throttling. This is good enough for me. I have no doubt that P95 after letting it sit for an hour will cause it to start thermal throttling because P95 blows.

Also now it's idling at 45-47C while watching Netlfix. I wonder what folks with lottery winning binning would get with delidding using CLU and Phobya Nanogrease Extreme?

I don't think it will get better unless somehow Clevo offers a direct to die cooling solution.

 

I posted my results in the other forum. Not sure if you saw it or not. I feel like I got a pretty solid CPU since I can run 4.9 ghz stable before throttle now.

 

@Zymphad: what kinda clocks and voltages is that at? stock? if so then thats still pretty toasty.... u should reaaaaaally try undervolting

Sent from my Nexus 5 using Tapatalk

 

All I can get out of my 4930K is 4.3ghz so don't feel too bad, your single core performance will outstrip mine.

 

who needs single core performance nowadays? and yeah right, we all feel sorry for your lowly hexacore meaker

 

Tri-quad core performance matters more in games than hex core speeds often In a 4 threaded task yours will out pace mine.

 

No thermal throttling so 100% @ 4.2

 

Someone really ought to do a delid and run a true bare die mount with CLU in order to probe the thermal limits of the cooling system.

 

thus far it looks like the thermal system can handle up to 75-80W of TDP on the CPU side before thermal throttling sets in... thats with a good paste job (gc extreme), machine propped up, max fans

Sent from my Nexus 5 using Tapatalk

 

I'm sure if you did a bare die mount with CLU you could easily squeeze at least another 30W or more from that.

 

just not fully convinced yet that a bare die mount would offer a significant difference over a delidded cpu with clu as the ihs tim...

Sent from my Nexus 5 using Tapatalk

 

Every mm can make a big difference along with removing thermal transfer layers.

Now if someone wanted to donate a machine for me to play with while I am in the states then feel free lol.

 

Well unless you're also using CLU between the IHS and the heatsink, a bare die mount with CLU will most definitely improve thermals.

Otherwise heat transfer from the IHS to heatsink would still be the bottleneck.

 

You are effectively bottlenecked by the poorest layer it's true.

 

I have much improved temps since have a few changes.

I had to replace MOBO (P770ZM-G), in order to upgrade to LG IPS eDP. May have the backplating and height etc more level? Not sure.

Had to redo the CLU, this time I applied a thin a layer not only the die, but also on the underside of the IHS. I also applied CLU to the IHS but also the underside of the heatsink. This was also done for the 980M. 980M as said before was replaced with fujipoly and replaced the pads used around the CPU.

I also upgraded BIOS to the latest XMG for W10 support.

I have 81 C for Intel's XTU stress test.. Core Frequency 4.19, Cache, 3.99 and 98% Utilization.

Idle temps not much changed. 47-53C, oh well.

 

Idle temps wont change so much as the smaller temperature difference will flatten any differences up to a point.