FRONTPAGE Intel i7 4790K Devil's Canyon CPU Review

Woomack said:

Actually I recommend to use more fans to get rid of the moisture. I had all tubes wet when I was benching on water with ice. Idle temps were ~0-5*C but it was long time ago, for sure not on haswell.
I was using about 2kg bricks of ice ( frozen water in plastic bottles ) and not much water in the bucket. In this way it was melting slower and keeping lower loop temps. I was also switching water from time to time as water in the bucket was heating up by ~10*C.

Click to expand...

i only got the condensation from where the fan was blowing onto the cpu area, on the other side there was nothing. but the next time i will use more ice and less water.

Shell said:

The heat might be from current handling issues and not the logic itself operating, aka a design shortcoming from being haswell.

Click to expand...

Seems to be the case.

many people reporting the same thing out of this chip.

You are correct, so much going on in such a small place heats the little bugger up.

outhouse said:

delidding and lapping changed little. have to bleed out air bubbles and check again, but still 99C with 1.325v

Click to expand...

You have to use liquid metal, ie clp/clu to get good temps (tim1 only). clp has highest conductance. CLP/CLU for tim1 (ihs to die). but dont use it for tim2, tim2 doesnt matter what you use, PK1, mx2, or whatever non-conductive tim. tim2 has such a huge surface area wb to ihs, versus critical tim1 with small surface area of hot spots on die to IHS.

Also make sure all the black adhesive is completely scratched off, credit card and gentle scratching is best way.

Then cover the exposed vrm with nail polish.

And then apply a very thin layer, extremely tiny drop on die and spread with enclosed qtip. I used CLP, primarily liquid gallium 40 w/mk. Though get the drop on tip of syringe then touch to die...dont squirt over die.

Since it took 2 days for CLP to arrive, I tried phobya hegrease first, I got the same temps as intel. Intels tim is very good, just way too thick for tim1. I couldnt get hegrease spread thin enough. I then used PK1 very thin, much thinner than using as tim2 (wb to ihs). And got ~10C better temps, but still large spread and still no impressive. After CLP, 23c temp drop.

Many on oc.net tried other tims with only mild improvement, but all the 20-30C drops were liquid metal clp/clu. It can be applied very thin and be effective, has much lower contact resistance, and much higher bulk conductivity.

Prior to delidding 4.7 ghz w/1.287v, prime 28.5 on small ffts instantly was 95C, then over 7 mins increased to 98C.

After delidding 4.9ghz w/1.38v core prime 28.5 small ffts instantly (30 seconds) max at 83C. Before delidding would have log/100C on all 4 cores instantly.

great tip using the nail polish


Im surprised no one has tried to lap the IHS inside to copper.


Just hitting it with 1000 and 1500 polished it up quite a bit and did reveal its flaws.

I'm still debating delidding. My main concern (besides possibly killing the cpu...) is whether the TIM will dry or require reapplication at some later time. I'm a set it once perfect, and forget it type. I hate having to go back and redo something over and over again, even if it is required maintenance. And I'm kinda confused on the CLP/CLU because of this. Some sites/reviews say that temps will start to creep up over time and require reapplication, some don't. I can't tell if it was just improper application that is causing the problem, or if its actually a TIM lifespan. What would be the consensus here?

I've used CLP on 3770K over a year ago. It's in my brother's PC for about a year without issues with temps etc.

CLP = "Alloy of the metal components gallium, indium, rhodium, silver, zinc and stannous".

CLU = "Alloy of the metal components gallium, indium, rhodium, silver, zinc and stannous, bismuth; suspended in a graphite-copper matrix

I don't think these alloys are quickly getting worse in thermal conductive.
Most new TIMs are prepared to work at least 5 years without change. I'm not sure what about liquid metals but I doubt you have to replace it every year.

Mpegger said:

I'm still debating delidding. My main concern (besides possibly killing the cpu...) is whether the TIM will dry or require reapplication at some later time. I'm a set it once perfect, and forget it type. I hate having to go back and redo something over and over again, even if it is required maintenance. And I'm kinda confused on the CLP/CLU because of this. Some sites/reviews say that temps will start to creep up over time and require reapplication, some don't. I can't tell if it was just improper application that is causing the problem, or if its actually a TIM lifespan. What would be the consensus here?

Click to expand...

Some thinner paste tims were not meant for tim1 use and will suffer from pump out. There are 200 people just on one delidded thread alone, many have been running delidded since ivy came out. No one using liquid metal clu/clp has had issues with having to replace or with temps increasing over time. CLP can form an almost bond over time, and may be difficult to clean off, but temps are not increasing over time. Some thin tim pastes like mx2, PK and others people were having to reapply after month or so as temps started increasing. That same issue was on 2 different forums and with same pastes.

But that is other reason for using liquid metal, it does not suffer from the harsher tim1 location pump out effect, at least not for year and half or so tested so far. But really if not using liquid metal tim, I wouldnt delid.

rge said:

. But really if not using liquid metal tim, I wouldnt delid.

Click to expand...

Agreed.

Done two CPU's and it made no difference at all.

I thought the lapping the IHS would have atleast picked me up a few C, but no. No gain at all.

What would you all suggest then to use to bond the heatspreader back onto the cpu? Would liquid electrical tape be enough to cover the caps and keep the hs bonded on? Crazy glue or any other type of cynoacrylate? If I'm gonna delid, I want to do it this weekend, so I can also start on reinstalling Windows (ugh).

Most use nail polish to cover vrm components, because it has been done before and works. There is one person that used electrical tape, but I havent heard how it worked. also nail polish comes with a perfect brush for applying.

I was originally thinking same, ie I wanted to glue IHS back down. But you dont want to create a gap via adhesive, so you would have to put IHS on first, then glue edge. I ended up not glueing ihs and I wouldnt if I were you.

after get liquid metal on die (very thin), then put die/pcb back in mobo, then very carefully put ihs on but maybe 1mm too high, and use your finger to hold ihs in place and close bracket, when you close bracket the ihs will slide down about 1mm because bracket mechanics. The bracket holds ihs on very tight. apparently after running for a while, the ihs kind of bonds to die with clp anyways from some peoples claims. also if you get a bad mount, you will want to redo, another issue with trying to glue down.

rge said:

I was originally thinking same, ie I wanted to glue IHS back down. But you dont want to create a gap via adhesive, so you would have to put IHS on first, then glue edge. I ended up not glueing ihs and I wouldnt if I were you.

.

Click to expand...

Was thinking silicone would work for this, and leave some crush room required.


Its always that 1mm of slide that bothers me once delidded and trying to close it back in.

May try a little oil in the ears, just enough for the tabs to slide on, instead of pushing the IHS off center.

Did order CLU tonight. Thought it was CLP and wasn't paying attention


getting old I guess.

Isn't CLU supposed to be the "better" of the two? I was planning on ordering CLU as well, since it appears from the reviews that it gets just as good temp as CLP, and spreads much easier.

Some months ago I got Phobya version of CLP which is exactly the same, in the same package but cost 30% less.

rge said:

Most use nail polish to cover vrm components, because it has been done before and works. There is one person that used electrical tape, but I havent heard how it worked. also nail polish comes with a perfect brush for applying.

I was originally thinking same, ie I wanted to glue IHS back down. But you dont want to create a gap via adhesive, so you would have to put IHS on first, then glue edge. I ended up not glueing ihs and I wouldnt if I were you.

after get liquid metal on die (very thin), then put die/pcb back in mobo, then very carefully put ihs on but maybe 1mm too high, and use your finger to hold ihs in place and close bracket, when you close bracket the ihs will slide down about 1mm because bracket mechanics. The bracket holds ihs on very tight. apparently after running for a while, the ihs kind of bonds to die with clp anyways from some peoples claims. also if you get a bad mount, you will want to redo, another issue with trying to glue down.

Click to expand...

Not electrical tape. Liquid Electrical tape. There are many different manufacturers, but for the most part they are all the same substance. I have plenty of that laying around doing nothing but drying up, so I might as well put it to use before its too late. Although not really meant to be used as a glue (it can peel off very easily, especially when used on smooth surfaces) it can at the very least hold the HS in place, and it dries very flexible, so should be squishy when setting the holder down.

Mine finally arrives tomorrow. I cheated though and bought an already tested one... She does 4.8GHz Wprime/XTU and 1.32v (on air).

Mpegger said:

Not electrical tape. Liquid Electrical tape. There are many different manufacturers, but for the most part they are all the same substance. I have plenty of that laying around doing nothing but drying up, so I might as well put it to use before its too late. Although not really meant to be used as a glue (it can peel off very easily, especially when used on smooth surfaces) it can at the very least hold the HS in place, and it dries very flexible, so should be squishy when setting the holder down.

Click to expand...

yep, I meant liquid electrical tape, nobody actually put tape on. The person that used liquid tape just used it to cover vrm components, not attach ihs though.

It is the adhesive that prevents good contact between IHS and die so wouldnt put electrical tape or silicone in same place intel does. Someone used a single piece of paper as a shim to get same crappy temps as intel....you really dont want anything between IHS attaching to pcb...unless you just put something on edge after putting IHS on.

And CLU and CLP should give roughly the same temps, probably more people using CLU. CLU is easier to spread and easier to remove. CLP caused almost a bonding between ihs and die over time and more difficult to remove IHS...I used CLP for that reason, hoping it forms a bond, and it has slightly higher thermal conductance.

Earthdog...sounds good...post your results when you get it.

Woomack said:

Some months ago I got Phobya version of CLP which is exactly the same, in the same package but cost 30% less.

Click to expand...

I read a 25 page thread on that and CLU and CLP.

It is not the same and did not stack up that well in testing against either CL product.

rge said:

And CLU and CLP should give roughly the same temps,.

Click to expand...

It was close, depending on how well it applied.

In general his test indicated CLU had a hair lower temps, but in another CLP had won.

Both stained and hardened with time, CLP being more prone to hardening. But a few complained of CLU as well.


I think I will take your advise and not use it on the outside of the IHS.

CLU is CLP that is suspended in graphite copper matrix, per manufacturer CLU will have slightly worse thermals, but better spreadability. But fact that several sites tested one or other better (I saw that same thread) just shows that error of mounting differences exceed performance differences. CLU thermal conductance is listed as 38.4 w/mk, CLP as 82 w/mk (~same as intels indium solder at 87w/mk). CLU seems correct given gallium is 40/wmk, CLP seems inflated specs even though minor constit have very high thermal cond..

And mpegger, here is guy that used liquid electrical tape, he showed pics with it on vrm components...he just posted his is working fine. Nail polish melts at 120C, most electrical tape melts 200+C, so either should work fine....my wife must have a thousand bottles of nail polish so I had that on hand. Though I realize you were speaking of using it to reattach IHS.

If you guys do glue ihs, be interesting to see what temps you get, I havent seen any temps yet from someone that tried to glue ihs back. But again, even a thin piece of paper in same place as black adhesive was causes 20C worse temps...so dont know how you would glue IHS on without negating the purpose of delidding....unless reattach ihs first then try to put some glue at edge to help hold it.