Anyone here delid an i7-4790K?

[Sasqui]

What socket clip are you referring to?

The foxconn trap clip that locks the CPU in the socket... on virtually every motherboard. The top of the IHS is pretty thick, more that a mm I'm guessing. Take that off, and the heatsink has to sit that much lower to make contact with the die

 

[GelatanousMuck]

The foxconn trap clip that locks the CPU in the socket... on virtually every motherboard. The top of the IHS is pretty thick, more that a mm I'm guessing. Take that off, and the heatsink has to sit that much lower to make contact with the die

With a bare die mount you remove that clamping mechanism.

http://forums.anandtech.com/showthread.php?t=2285595

 

[trog100]

i took the top off my 4790K not just to change the tim.. i did it to make sure it was assemble properly.. intels (cheap mass production) way of doing it aint that clever.. there is no pressure between top and chip and there is too much paste..

so i have answered the question of why i took mine off and then put it back.. i did think about not putting the top back but decided any tiny gain would not be worth the trouble..

trog

 

[Sasqui]

With a bare die mount you remove that clamping mechanism.

http://forums.anandtech.com/showthread.php?t=2285595

Yes.

i took the top off my 4790K not just to change the tim.. i did it to make sure it was assemble properly.. intels (cheap mass production) way of doing it aint that clever.. there is no pressure between top and chip and there is too much paste..

so i have answered the question of why i took mine off and then put it back.. i did think about not putting the top back but decided any tiny gain would not be worth the trouble..

trog

Wasn't for me either. The die and IHS was pretty tight so no real improvement. Intel must of figured out tolerances, or it was an accident... Bawhahaaa!

 

[trog100]

Yes.



Wasn't for me either. The die and IHS was pretty tight so no real improvement. Intel must of figured out tolerances, or it was an accident... Bawhahaaa!

the only way to find out is try it and see.. i gained a good 10 C maybe a little more i think they can vary somewhat..

trog

 

[peche]

Come on AMD!

no thanks ... nothing gooood for me on that side...

The reason I can run a 5ghz overclock on a CPU running TIM under the heat spreader is the cooling, I am running below ambient chilled water cooling.
My coolant temperature is peltier cold side generated, running on a preset temperature relay to keep my coolant temperature in the 10c target zone, which in a 23c ambient room temperature is 13c below ambient.
I am running below ambient but still above any condensation developing temperature, so that is not a problem.
And that sir is why I can run a 5ghz rock solid stable overclock.

There is a link to my cooling in my sig, if you are curious.

thanks again sir

 

[cat1092]

Normally I don't like to revive long dead Topics, yet since I have an i7-4790K that peaks at 4.6GHz (although at high voltage & temps under load), want to delid to reduce both. And then try to get to 4.8GHz stable, with some luck, 4.9 or 5.0GHz.

Have been able to boot at 4.8Ghz with only multiplier adjustment, and disabling a few things that more OC'ers does. Yet as soon as I fire up the Intel Extreme Tuning Utility & start to perform a benchmark, half way through, sometimes almost done, never fails, BSOD. The best I can do w/out crashes is setting multiplier at 47 & that'll allow me to run the bench, my best score so far was 1224 on 09/17/2017. It can be seen below that how many times I came close, took over 5-6 months to beat my previous best of 1221.

https://hwbot.org/user/cat1092/

Some may question why I want to make a 5.0GHz run (or close), well it's the last best Intel CPU I'll own, because AMD still builds theirs the old fashioned way, with a soldered on IHS. Come two years, it'll be the equivalent of today's Ryzen 1800X for me.

I use my PC for some general usage, also for folding@home, everyone knows, the higher the performance, the more PPD earned. And I earn far less points with the CPU than the GTX 1070 FTW, so need to make that chip perform.

Too, it's kind of a bragging rights thing, I'd guess that over 80% of those who delids (or even purchases a high performance CPU) are doing so for that purpose. Although with Haswell, things were a bit different, the closest model to it was the 4790 (non-K). Ran at 3.6GHz with 4.0GHz Turbo, going to the K version was a massive leap w/out OC'ing. There would be a few other Haswell i7's that had close performance to the vanilla 4790 (the 4770 included with my XPS 8700 is close, once a Top 10 CPU on Passmark), still the 4790K was the cream of the crop. According to Passmark, even with DDR4 RAM & new chipset/MB, still edges the i7-6700K by a hair (non-OC'd).

Has the standards changed since this was discussed? Am a bit confused whether to use clear nail polish or the liquid electrical tape I just learned of tonight. I'd assume it's best to pick up a bottle to be on the safe side.

Will be using the Rockit88 1150/1151 delid/relid kit & won't be using so much glue, just a very light touch of RTV sealant on the corners & no more. Have the Thermal Grizzly Conductonaut liquid metal compound, will be plenty to do all three of my Haswell's (there's the other Devil's Canyon, i5-4690K in another build).

So if I can run at 4.6Ghz with stock TIM & Noctua NH-D15 cooler, wouldn't it seem that an added gain of 400MHz be possible with liquid metal, combined with tweaking voltages as I bump the multiplier upwards? While that may seem to be a tall order, it's not a lot to ask for & expect, the cooler running chip should be able to pull it off. Of course, it'll take 24 hours of folding at each step after some Prime95 tests to ensure stability. Looking to move up 100MHz at a time, make adjustments as needed, until I run against the wall.

If that wall happens to be having a system running at 5.0 or maybe 5.1GHz, it'll all be worth the effort. Than will move onto the oldest, the XPS 8700, of which CPU temps has naturally risen, even with a now copper core cooler with downwards blowing Noctua NF-A9 fan & exhaust fan spinning at 2.5x of that when new, plus added an intake fan for more airflow. Stronger flow at that, the exhaust is connected to the CPU with a PWM splitter, the intake is plugged into the native exhaust & running almost as fast.

Thanks for any responses.

Cat

 

[Folterknecht]

...

 

[cat1092]

Thanks for the video, have the Rockit88 tool already, ordered two of their copper IHS replacement minutes ago, along with one relid guide that works with the kit.

https://rockitcool.myshopify.com/products/copper-ihs-for-lga-1150

Works with both 1150 & 1155 CPU's, there's a different for 1151

https://rockitcool.myshopify.com/products/copper-ihs-for-lga-1150-1151

Figured to go with this over the one already installed, copper tends to run cooler (gets rid of heat faster) & will save time on cleanup. Also larger surface area. May get a liquid cooler, depending on how this works out.

Cat

 

[jboydgolfer]

@cat1092 the rockit88 works for 1150 & 1151,(and possibly more, but these are the two i did) i did it this past week.*edit* maybe you were referring to that copper ihs thing. Nvrmnd

using that tool makes it SO easy, there literally only human error left to chance. I recently did a couple of my Chips, and a 4790 a well, i dropped 28c, but my Stock TIM was dried aF. the 4790 has a ton of capacitors, or whatever they are on the left side of the chip, i just scotch taped it off on one side on the IHS, and surrounded the ship with scotch tape on the Wafer.

Another point I think that's worth mentioning, when you put your chip in the device and you have it all clamp down ratchet the piston forward until it contacts that IHS, then put a piece of tape that shows that line (when its starting to push the IHS) that you can see while the device is closed ,because my 4790 didn't make that pop sound it just slid off , and if i weren't paying attention and just twisting waiting for that pop sound i couldve possibly broke it. Or you can just take note of the lettering on the top of the CPU as it moves to the side

I put VERY little RTV on for sealant, I can hold the chip to my eye, and see light right through it. I spread it thin, on each of the corners with a tooth pick, and it doesnt even squeeze out when i vice it down. if your scraping RTV after it dries, youve added too much

i recommend using a fine tipped marker, and putting a line, or marking on the corner with the Golden Arrow, it helps avoid any possible human error, as putting the lid on wrong is one of the few things left for a person to screw up. like this on my 8600k

 

[infrared]

Excellent info from jboy and falter re delidding.

I use my PC for some general usage, also for folding@home, everyone knows, the higher the performance, the more PPD earned. And I earn far less points with the CPU than the GTX 1070 FTW, so need to make that chip perform.

Just wanted to pick up on this, using a CPU to fold is pretty much a waste of time, even with it overclocked as hard as you can the PPD will be dismal compared to a GPU. A better way to contribute to medical research using your CPU would be to install boinc and join our WCG team, it's CPU only and you'd be making a much bigger contribution in that than you would in F@H. A lot of us run both folding@home on GPU and WCG on CPU at the same time, you just need to set WCG to leave a thread free to feed the GPU (if ppd drops, just free up a second thread). Just a suggestion, apologies if you're already aware and just prefer f@h.

 

[flexy]

This is an old topic, but I recently again got into thinking about delidding...in the meantime there are also many great tools out there to help you.

But let me tell you my story, this was on my old Haswell 4770k which I so badly wanted to delid...

* It was a TOTAL NIGHTMARE and I *almost* destroyed my CPU. (Or at least that's what I thought...)

First, ignore all the delid Youtube videos where you see some idiot with a carpet knife or even a utility knife, since according to my experience these knives are WAY too thick. You won't cut into the sides of your chip with a box cutter or a carpet knife. And if you do, then with a very high risk for damage.

When I tried to delid mine, I thought I was well prepared and watched many videos, but then in reality it turned out a total pain in the ass.
I started to slowly trying to cut, until at some point I slipped. Obviously I freaked out. I realized there is no way to do this with a box cutter, if at all then with a razor.

When I turned my chip over after the slip, I spotted several (2-3) contacts on the chip had been "flaked off". (In the mean-time I realized that these "flaked off" contacts were likely not because of my slipped knife, but I think that this is how I got the CPU since I bought it used back then. If I had cut the CPU, the dmg would look different, IMO).

I was of course at first 100% convinced that the chip is destroyed since I thought I did the dmg, but low and behold I could reassemble the PC and it worked. It then did all sorts of tests and can't find a problem. I looked up these exact pins on schematics, and it seems to be redundant/non important pins.

But the point why I then never ever considered any other delidding (with vice or even with these tools) is that the CPU obviously is already "gnarly" and already lost pins...so I didn't do anything else like putting it in a frickin' vice and whack....who knows, more pins might come off. I am glad that thing works so I haven't touched it ever since.

That Haswell gets insanely hot...OCCT, Prime95...I mean you all know the usual offenders. (I am running at 4.4@1.250), 1.250 is really the upper limit on air.

But here is the deal: So the CPU gets 85-90C in OCCT, but despite these temps shocking you never see these temps "in real life". You won't see 90C in BF4 or whatever game. I simply accepted that Haswell gets crazy hot in synthetic tests. So I just don't use them. For testing I am now using Real Bench, or maybe X264 stability test. Yes, it's still "warm" (80s) when I run X264 test...but at least the PC works and I am fine with my 4.4Ghz. I'd likely delid a new CPU with a tool, but not this one that's already lost some pins for whatever reason. Too much risk, IMO. (I also did some calculations back then and concluded that even if I had delidded and say it would run 20C or so cooler, the chip doesn't have much headroom for a higher OC. I'd likely have to feed it 1.45 or whatever for 100-200Mhz more, which I now can't do because of temps, but ultimately not worth it for me.)

 

[Bones]

Have a tool I made originally to delid PS CPUs to redo the dried out TIM when fixing overheating issues such as the red or yellow LED of death with those. The tool was originally for another purpose in how thin yet stiff the metal/blade was. That made it perfect for the job I need it for.
Had to shape and reshape the config of it a few times to get it right but now I can delid a PS3 CPU with ease..... And can do it to a PC CPU too. Used it to delid my 7700K and worked perfectly, in fact that went easier than doing a PS3 CPU normally would have.

Was also a good thing I had several older/dead PS3's to work out the kinks with it, I did manage to kill couple of boards due to knocking components around trying to manuver the tool around when cutting, PS3 CPUs are soldered into the board and you can't remove them so you gotta work with and around everything there.
Got those issues worked out by refining the config and also have to say it has yet to cut into the PCB of the chip itself, solved that problem right from the start. If it's going to try and cut anything it will be the lid, not the PCB of the CPU.

So.... If you are really determined to make something work it's possible, what I have wasn't originally intended to delid chips like my Kaby but worked perfectly when tried and the change in temps afterwards showed it as worth doing.
Not OC'ing the chip, it's more of a lower temp ='s a longer life with less degration for the chip overall.

 

[Vayra86]

This is an old topic, but I recently again got into thinking about delidding...in the meantime there are also many great tools out there to help you.

But let me tell you my story, this was on my old Haswell 4770k which I so badly wanted to delid...

* It was a TOTAL NIGHTMARE and I *almost* destroyed my CPU. (Or at least that's what I thought...)

First, ignore all the delid Youtube videos where you see some idiot with a carpet knife or even a utility knife, since according to my experience these knives are WAY too thick. You won't cut into the sides of your chip with a box cutter or a carpet knife. And if you do, then with a very high risk for damage.

When I tried to delid mine, I thought I was well prepared and watched many videos, but then in reality it turned out a total pain in the ass.
I started to slowly trying to cut, until at some point I slipped. Obviously I freaked out. I realized there is no way to do this with a box cutter, if at all then with a razor.

When I turned my chip over after the slip, I spotted several (2-3) contacts on the chip had been "flaked off". (In the mean-time I realized that these "flaked off" contacts were likely not because of my slipped knife, but I think that this is how I got the CPU since I bought it used back then. If I had cut the CPU, the dmg would look different, IMO).

I was of course at first 100% convinced that the chip is destroyed since I thought I did the dmg, but low and behold I could reassemble the PC and it worked. It then did all sorts of tests and can't find a problem. I looked up these exact pins on schematics, and it seems to be redundant/non important pins.

But the point why I then never ever considered any other delidding (with vice or even with these tools) is that the CPU obviously is already "gnarly" and already lost pins...so I didn't do anything else like putting it in a frickin' vice and whack....who knows, more pins might come off. I am glad that thing works so I haven't touched it ever since.

That Haswell gets insanely hot...OCCT, Prime95...I mean you all know the usual offenders. (I am running at 4.4@1.250), 1.250 is really the upper limit on air.

But here is the deal: So the CPU gets 85-90C in OCCT, but despite these temps shocking you never see these temps "in real life". You won't see 90C in BF4 or whatever game. I simply accepted that Haswell gets crazy hot in synthetic tests. So I just don't use them. For testing I am now using Real Bench, or maybe X264 stability test. Yes, it's still "warm" (80s) when I run X264 test...but at least the PC works and I am fine with my 4.4Ghz. I'd likely delid a new CPU with a tool, but not this one that's already lost some pins for whatever reason. Too much risk, IMO. (I also did some calculations back then and concluded that even if I had delidded and say it would run 20C or so cooler, the chip doesn't have much headroom for a higher OC. I'd likely have to feed it 1.45 or whatever for 100-200Mhz more, which I now can't do because of temps, but ultimately not worth it for me.)

Well spoken. Delidding is mostly an enthusiast type of move where you just want to squeeze more out of a chip and run it as cool as possible. Both aren't necessary in any way, and if you're not having fun doing such things, just don't. The TIM is fine as it is - yes even on Coffee Lake.