Just finished assembling my RIG;

Case: Corsair Obsidian 650D

MoBo: ASUS Z87-PRO

i7-4770K overclocked to 4.6Ghz, Core voltage= 1.248 (stable)

ASUS GTX780-OC edition

Corsair Hydro Series H110 280mm radiator mounts

Max Temp on OCCT reached 76 deg C on highest core, 68 deg C on lowest core.

do you suggest that I de-lid before pumping it up to 4.8 GHz?

Really great read Charles.

 

You're giving me the urge to de-lidd my 3770K.

 

I was about to a few months ago but a split second before I dived in, the minimal performance increase as a result de-lidding put me off. 

 

If I do it will be for fun, just to see what I can achieve. I may lap also.

 

My preference on top of the IHS has always been Coollaboratories diamond paste. You've tempted to try MX-4 next time.

 

Looking forward to seeing your results.

 

Max Temp on OCCT reached 76 deg C on highest core, 68 deg C on lowest core.

do you suggest that I de-lid before pumping it up to 4.8 GHz?

 

Hi Ali A. ... no ... I do not even recommend going to 4.8GHz ... at 4.6GHz I'm sure MSFX is chugging along without any hiccups whatsoever.  In the end, that should be the goal ... don't chase fps ... chase a smooth operating flight sim under all conditions. 

 

With that said however, for me, this morphed into something much more than getting MSFX to run well, it was already doing that at 4.4GHz (for me anyway ... I don't run a bunch of add-ons ... I just like to fly).

 

About halfway through my build I started to realize .. hey ... I want to try this, then that, then another thing.  Soon I just wanted experience using Coollaboratory Liquid Pro, then the Corsair H100i, then using four Noctua F12 fans to set up a push-pull airflow through the radiator, then delidding, then lapping, then the carrot of being able to follow NickN's up and coming jaunt through Haswell overclocking ... ad infinatum. 

 

So, if you are like me and you want to try out all of these things ... go for it ... learn by the mistakes that I made (like staying away from the top left corner of the CPU) and gain experience.  I can say pretty safely now that I know the physical layout of the CPU under the IHS that I can delid one with little or no chance for destruction. 

 

My only advice would be to possibly reconsider using CLP under the hood ... it is without doubt the best ... but in order to minimize risk I'm not sure that MX-4 would be a better choice (I bet you wouldn't be able to tell more than a degree or two).  And, without doubt, shifting that IHS from the right to the left during the mounting process on the motherboard would definitely screw things up.  MX-4 is non-conductive and non-capacitive ... thereby reducing risk significantly.

 

 

Looking forward to seeing your results.

 

Martin ... me too ... if I've gone through all of this and my temps sill hit 95C on the Prime95 test ... I will be very disappointed to say the least.  If, after I've performed the next step of continued prep of the IHS and the bottom of the H100i cold plate and then applying the MX-4 the temps are still high I can only assume that the CLP TIM between the IHS and the CPU heat pad was too thin and the two surfaces did not come directly in contact with one another.  But, I don't think that is going to happen.  Removing the resin should bring the two surfaces into tight contact with one another.  In addition to that, continuing to tighten the cold plate down during the mounting process should serve to squish the IHS onto the CPU thermal plate just a little more. 

 

Results on Wednesday (two day delivery from Newegg).

 

Cheers!

 

C. T.

This is what I'll be using on Wednesday evening to stress everything out ... IntelBurnTest v2.54.

 

Anyone have any comments about the validity of this test ... HLJAMES ... you out there B)? 

 

Thanks,

 

CT

Ali A.:

 

What I meant to say is that I'm pretty sure that the MX-4 might be a better choice over CLP under the IHS ... it takes away the stress of wondering if any conductive material accidentally got onto those components to the left of the CPU's thermal pad ... I bet you wouldn't see a couple degrees of difference between the two ... if any at all ... but I'm not going to tear that IHS off to test that theory ... I can say that with certainty.

 

C. T.

 

 

one have any comments about the validity of this test ... HLJAMES ... you out there ?

Test is OK......press on.....be careful!

 

HLJAMES

Ali A.:

 

What I meant to say is that I'm pretty sure that the MX-4 might be a better choice over CLP under the IHS ... it takes away the stress of wondering if any conductive material accidentally got onto those components to the left of the CPU's thermal pad ... I bet you wouldn't see a couple degrees of difference between the two ... if any at all ... but I'm not going to tear that IHS off to test that theory ... I can say that with certainty.

 

C. T.

 

Thanks for your input C.T.

enjoying your thread(s) on this project, and enjoying your detailed reports and photos of every step you do, very informative, thanks again.

 

best of luck on your final tests ;-)

Though I very much admire the work you are doing to delid, and not to discourage anyone else but I feel it is a waist of time -for me.

 

Based on the fact that after doing this repeatedly and seeing little result gained with my Haswell I5 (temps went down but OC limit was the same).

 

Second, to replace the I5 the 4770K I picked up is a good chip and doing very well @4.4 with ultra quiet 750 rpm Aircooling D-14. With water the chip will likely go to 4.8-or 5 but I can live without the noise of the pumps and fans needed to keep it up there, but who knows I might be tempted to push it on air, maybe see what she will do at 950 RPM and 4.6 better then many who have water and have delided) but I'm not deliding.

 

Just saying after doing the delid thing for so long I don't see the merit in it any longer, I don't worry the TIM will dry out under the HIS and I'm not trying to compete in Benching contest anymore and while it was "fun" I see no point of it for FSX.

 

to see even 10% more FPS (from 30 to 33 or 20 to 22 I would need to spend considerable more time and money that I feel is better saved for other things like family and friends - And other FSX add-ons Ha.

 

I think your post and work are really good and helpful so I hope that you can trust that this is just another take from a friendly someone who has had the same itch and scratched it.

 

Cheers! 

As an update from Corsiar on the ampere capacity of the H100i coolant block (pump):

 

Just as an update, there are two slots on the side of the pump.  Corsair provides a y-splitter that inserts into each slot, meaning that each slot can support two fans and therefore H100i can support up to four fans total.  As per Corsair technical support each slot can provide 2 amps of current to two fans. 

 

I connected two Noctua NF-F12s to slot one (which performed as a push pair throwing air up into the radiator from the case) and two more on top pulling air through the radiator.  As per Noctua, each F12 pulls .5 amps; therefore, four fans @ .5 amps per fan totals 2 amps ... which is well within the design capability of the N100i.  This pump could handle four Delta fans  (.8 amps / fan) ... which pull 110 cfm ... it would sound like a freight train though.  The better choice would be to buy two more SP120's (the default fan for the H100i) and put them on top of the radiator also creating a push/pull arrangement.  Unfortunately, those fans put out 30db each for a total of 120db for the H100i push/pull arrangement.  This is why I swapped in the Noctua fans.  Each one of those are pretty quite ... around 16db each.  When you add that to AS Suite 3 fan control, the fans are barely audible unless there is a killer load on the CPU like a stress test of some flavor.

 

Anyway, in turn, the H100i is connected directly to the CPU_FAN header leaving the CPU_OPT fan header free to control another set of fans.

 

C. T.

Yeah thats my experience with a retail chip (not delidded).

The water cooling industry understands what you are trying to do. They have already got the components you need to take Haswell / Ivy Bridge to 5.0GHz.

 

HLJAMES

 

There's only so much the cooling industry can do. The two problems with Haswell are:

 

1. TIM acting as an insulator rather than conductor of heat, trapping the heat under the IHS instead of transporting it to your water cooling loop. Your expensive water cooling system will just be circulating cool water around, while the CPU itself is boiling under the IHS

 

2. "Voltage wall" - a lot of chips need increasingly higher voltages as you move beyond 4.2 - 4.4 GHz, causing very high temps regardless of the cooling used.

 

The only way to get a 4.5+ overclock is to play the silicon lottery. If your chip does 4.6+ GHz at ~1.2V, you won't have to de-lid or invest in expensive cooling.

 

 

Based on the fact that after doing this repeatedly and seeing little result gained with my Haswell I5 (temps went down but OC limit was the same).

 

I would tend to agree. If you're not willing to risk your CPU, don't bother.

 

Before de-lid, my CPU would do 4.3 GHz at 1.218V and 92-93C in the most demanding AVX stress tests. After de-lid, I'm able to dial in 4.4 GHz / 1.24V and still stay at around 80C in the same stress tests. However, at 4.5 GHz, the voltage required is so high that my temps are right back to where my pre-delid temps were at 4.3 GHz. So in terms of overclocking, the de-lid gave me 100 MHz, or 200 MHz if I''m willing to put up with the pre-delid temps.

 

The good thing is that at 4.4 GHz, the heat is now extremely manageable. Outside of stress tests, in actual games and applications (where temps are 20 - 30C lower than stress tests) I have a very cool, quiet system and no longer need to turn up the fan speeds.

 

If done right, de-lidding to get this benefit is almost free and takes a few hours (20-30 minutes for the de-lid, then cleaning everything up, applying the Liquid Metal and putting it all together). I think de-lid is best applied when you have an overclock that you're happy with, but you would like your CPU temperature to be lower (either to lower the fan speeds to reduce noise, or because the temps are too close to TJMax at 100C).

Pigs In Space:

 

You are absolutely in the ballpark there ... did you see the post earlier to warn lurkers as to the potential frivolity of this endeavor?  There are six very valid points as to why one should walk away from this exercise and be happy with a very achievable 4.2GHz that the MAXIMUS series of boards can provide by entering two settings in the BIOS ... which is a VERY stable state (mine passed all stress tests without blinking) and a "cool" thermal signature with Vcore and Temperature combination very addressable by a good air cooled or AIO water cooled solution.  I'm not sure the H100i could handle a 5.0GHz OC ... but I hope with a P/P fan on the radiator it will handle 4.6GHz ... time will tell.  I do draw the line at the H100i though, I'll be damned if I'm pulling all of this apart to put in a custom cooling loop.  No sir ... now way! 

 

Also, when I put my chip through its paces at 4.4GHz (adaptive voltages ... which threw some pretty big numbers its way) I also had a stable overclock and Prime95 pushed the temps to 95C on a couple of cores.  That's not too shabby ... probably should have stopped there!  But that will be my test base ... do the same test with a de-lid and see if the temps drop by the advertised amount of 15C ... approx. ... based upon what HLJAMES says ... the temps might stay the same ... hope not!

 

I did a test going from 4.2 to 4.4Ghz and the increase in fps was TWO ... approximately ONE fps per 100MHz.  So, going from 4.2 to 4.7 (my personal goal) would add 10 fps at KATL GATE-3 (assuming the relationship of MHz to fps is a linear function).  Seeing as how I'm currently at 19 fps ... 10 more would put me over my .cfg limit of 25 ... actually, I could live with 4.6GHz ... but that just doesn't sound very impressive given the level of work does it ... I might have to demand 4.8GHz to feel good about myself :rolleyes: ... but that would be a miracle.

 

C. T.

I did a test going from 4.2 to 4.4Ghz and the increase in fps was TWO ... approximately ONE fps per 100MHz.  So, going from 4.2 to 4.7 (my personal goal) would add 10 fps at KATL GATE-3 (assuming the relationship of MHz to fps is a linear function). 

 

Actually, I'm not sure it works that way. 4.2 to 4.4 GHz is an increase of just under 5%. If you got 20 FPS at 4.2 GHz, this 5% increase would translate to 21 FPS at 4.4 GHz. This is what you should have seen, but there are many explanations why you didn't see exactly that. However, the jump from 4.2 to 4.7 GHz is is around 12%, which would increase your FPS from 20 to 22.4 FPS.

 

When you think about it, this makes sense. If each 100 MHz added 1 FPS, it would only take a 2 GHz CPU to achieve 20 FPS. The relationship is linear, but it's linear related to the percentage of clock speed you increase by, not by absolute clock speeds. To double FPS from 20 to 40, you would have to double your overclock to 8.4 GHz. To add 50% (10 FPS), you would have to overclock to 6.3 GHz. The difference between e.g. 4.2 GHz and 4.6 GHz is actually very small and not worth worrying about. It's mostly about bragging rights.

JimmiG: 

 

That pop you heard was of my balloon of expectations exploding :unsure:  ... well, if the only benefit is that I drop the temps from 95 to 80 @ 4.4GHz running Prime95 then the absence of fan noise would be worth the delidding I suppose.

 

I'm also primarily looking forward to having the rig ready to work through NickN's Haswell Overclocking guide.  I want to be able to follow the whole thing ... at least now the only limitation I have is the one imposed upon me by Intel.

 

Sigh ...

Charles, Stay focused, Press on into the darkness!

 

HLJAMES

Well my rig has been undergoing BIOS settings changes and stress testing for the last couple of days to find stability at 4.5Ghz and from there on I plan on pushing for 4.7Ghz where I'll probably call it a day since I don't really plan on delidding at the moment.

 

At 1.2v on my 4770K I was able to boot into windows at 4.6 and everything seemed stable. Passed a quick run of IBT but failed on Prime after around an hour I believe.

 

Currently I'm running 1.22 vcore for 4.5Ghz and everything else auto I was able to pass IBT, Aida for 12 hours but Prime crapped out after about 4 hours. Uncore is currently at default 35x.

 

I've found that setting uncore coltage to 1.15 (as recommended in gigabytes guide) and CPU input voltage to 1.8 has added stability for me in prime (I'm looking for min 12 hrs stable same as I did with Sandybridge).

 

Have a feeling I may be able to drop vcore a little since changing uncore voltage/CPU Input seems to have added some stability, but time will tell.

 

Haven't run any tests on FSX yet.

 

From what I can tell so far it seems like I've ended up with a pretty decent chip.