De-lidded i7 6700k

[ZenZimZaliben]

So I have read this article: http://www.guru3d.com/news-story/intel-skylake-de-lidded.html . The gains seem to be pretty big. Add in watercooling and this could make a huge difference.

I am wondering why not run it naked with a copper shim? The IHS/TIM seems to be very poor on skylake. I need to find out the thickness of the IHS and look into a copper shim for my waterblock. Or if just replacing stock TIM in IHS will be good enough.

Delidding looks really easy this time around. Will not be the first time I have delidded or lapped a CPU. Nice thing about skylake is all the components other than the cpu have been moved back to the board so no chance of accidentally razoring off a component hidden under the IHS.

Thoughts?

 

[fullinfusion]

IMHO I think its a waste of time "unless" you have a very hot chip to begin with. A 4° drop isn't much IMO. Why not see how it rocks out for awhile and go from there. You always have great loops there Zen so why the Interest in delidding it?

@cadaveca how's the temps on that ES chip you have?

 

[ZenZimZaliben]

IMHO I think its a waste of time "unless" you have a very hot chip to begin with. A 4° drop isn't much IMO. Why not see how it rocks out for awhile and go from there. You always have great loops there Zen so why the Interest in delidding it?

Because I really want to break into the 5+ghz realm as a 24/7 clock.

 

[manofthem]

IMHO I think its a waste of time "unless" you have a very hot chip to begin with. A 4° drop isn't much IMO. Why not see how it rocks out for awhile and go from there. You always have great loops there Zen so why the Interest in delidding it?

@cadaveca how's the temps on that ES chip you have?

4*C drop is nothing but that 16*C drop from Cool laboratory stuff is pretty awesome. That's the stuff I used last time and it was great, though a little different to work with.

However I think it's a waste for the vast majority of people.

 

[cadaveca]

Because I really want to break into the 5+ghz realm as a 24/7 clock.

I think that this clock will be rare, and added cooling won't be the factor that dictates whether it is possible or not. If you really want 5 GHz, then you'll likely need to bin CPUs till you find one with 1.2 V default VID.

 

[ZenZimZaliben]

4*C drop is nothing but that 16*C drop from Cool laboratory stuff is pretty awesome. That's the stuff I used last time and it was great, though a little different to work with.

However I think it's a waste for the vast majority of people.

Well I already have the cooling system in place so I am already not part of the vast majority. Naked and Shimmed I bet it will be a 20c drop.

I think that this clock will be rare, and added cooling won't be the factor that dictates whether it is possible or not. If you really want 5 GHz, then you'll likely need to bin CPUs till you find one with 1.2 V default VID.

Maybe...We don't know that yet for sure. I have seen a few screen shots of non ES running 5.2Ghz and stable at 5ghz. But those screen shots have all been off chinese sites....so...

 

[radrok]

Unless you get a golden chip that is 5GHz stable (stable on everything is pretty much impossible) at safe voltages be ready to swap your CPU for a new one in cycles, because silicon degradation will become an issue.

I've been there, swapped so many CPUs on X79 era because of that.

 

[EarthDog]

This is only worth it if you are actually temperature and not voltage limited. You will more than likely need a well above average sample to run 5GHz stable under 1.45v.

My stock voltage(VID) is 1.31v for the current CPU. I was able to run 4.9GHz @ 1.449v and it wasn't too stable. It passed Cinebench, WPrime, HyperPi, and Super Pi 32M, though.

I wish you luck, you will need it.

 

[ZenZimZaliben]

From the few shots I have seen it seems like temperature is playing a big part on limiting OC. If voltage is the main issue then yeah I will be SOL. I am hoping with the vregs being moved off the core there will be a greater opportunity for OC'ing. This is a big deal really with the vregs off chip.

 

[EarthDog]

It wasn't for mine... custom loop (totaling 5x120mm of rad, CPU only). I think I peaked a bit over 70C running Wprime and Cinebench against it several times. I am sure it would raise a couple of C if I left it looping for hours on end, but... 70C leaves 20C+ to go, no?

 

[ZenZimZaliben]

It wasn't for mine... custom loop (totaling 5x120mm of rad, CPU only). I think I peaked a bit over 70C running Wprime and Cinebench against it several times. I am sure it would raise a couple of C if I left it looping for hours on end, but... 70C leaves 20C+ to go, no?

What wasn't for you? If you saying Voltage vs Temp then I would say you are talking about one of the chips in your sig? If that is the case then I could say the same thing...Since my i7 950 oc'd 1Ghz+ so should skylake.

 

[EarthDog]

The explanation didn't tell you I was talking about...

temperature is playing a big part on limiting OC

MY CPU wasn't temperature limited?

EDIT: Sorry, I have not updated my system specs here because we have not formally moved to Skylake as our review platform (ironing out W10 and new games/benchmarks). I am however talking about the 6700K I have. Sorry for the confusion there. I assumed being in a 6700K delid thread, that would be assumed. My fault!

 

[cadaveca]

From the few shots I have seen it seems like temperature is playing a big part on limiting OC. If voltage is the main issue then yeah I will be SOL. I am hoping with the vregs being moved off the core there will be a greater opportunity for OC'ing. This is a big deal really with the vregs off chip.

Not temps. 5 GHz will be about 130W.

130W is minor, Haswell-E at stock, for example.


You are right, there is more overhead, heat-wise, but because voltage regulation is moved off of the chip, the voltage supplied to the CPU at stock is high in comparison to past generations. My 4790K was 1.020V @ stock... my 6700K is 1.25V. EarthDog's is 1.3V. Degradation happens quickly at just over 1.4V. That's not a lot of room. Also, that small increase in voltage possible means that clock increases are just as limited.

It is possible that Intel did not give out good chips on purpose, but this seems silly. Since the CPU is not on sale yet, you can say safely that all results currently are "hand-picked".

Intel can work this design some, add more cores, make the silicon able to handle more current, but this is 1st-gen 14nm, and is prone to have issues.

this is my 6700 K CPU wattage @ 4.8 GHz running wPrime(middle meter is CPU, outside two are PCIe plugs into GPU):

It seems the lowest result so far is 4.8 GHz, really. My 4.8 GHz is rock-solid. Maybe you'll get a good chip, but again, the stock VID will tell you all.

 

[EarthDog]

For reviews, we are likely going to take ours down to 4.7Ghz just in case we (heh, I - since I am not the CPU reviewer I have to buy my own.......) get a dud.

 

[ZenZimZaliben]

So are either of you able to play with BLCK more now? I know just tossing a higher multi is easier, but wondering if increasing BLCK might help stability vs just high multi. The BLCK is supposed to be unlocked now, or at least able to move beyond 100. So rather than run 50x100 you run 48x104..take some load off the cpu multi and place on blck/fsb. I know the end result is the same but you are changing freq at different points...just wondering.

 

[cadaveca]

So are either of you able to play with BLCK more now? I know just tossing a higher multi is easier, but wondering if increasing BLCK might help stability vs just high multi. The BLCK is supposed to be unlocked now, or at least able to move beyond 100. So rather than run 50x100 you run 48x104..take some load off the cpu and place on blck.

Yeah, it's unlocked, and most motherboards have a similar chip on them that helps with this. 500 BCLK should be common for those running LN2.

Yet, there is a limit right now with speeds at just over 6 GHz, LN2-wise. So we are basically looking at the 2600K in terms of process maturity in regards to MHz scaling. We are now back to the point where what motherboard you buy for overclocking to the limit is really going to matter, when it comes to CPU power delivery. Also, DDR4 clocking is another thing to look at. But BIOSes are still a bit raw, so maybe that will improve as time goes on, I am not sure. If things go well, I'll end up with several retail chips and can get a better idea of what the clock scaling would be like, but I'm pretty confident in saying that most newer AIO coolers are more than enough to max out these chips for 24/7 use. Some might end up a bit hot, and air coolers might struggle at times due to the small die size, but because the IHS is thicker, maybe that will help with heat dispersion. The TIM is an interesting factor, but with such low power consumption for what is relatively high clocks, I think most users can re-use their existing Haswell cooler when upgrading and end up with better temperatures at max OC.

 

[EarthDog]

I ran mine up to 225Mhz on the MSI Gaming 7. Anything over 250 and it wouldn't work. I have not tested it under cold at this time. I have a Giga board landing today and we will see if I can get past 250 on it...

It really doesn't 'take stress off the CPU/multi' though... If you hit a multi wall, it could help in that case. Otherwise, it just helps with granularity of the overclock (CPU/Memory).

500MHz... wow. I doubt that, but I hope I am proven wrong!

 

[cadaveca]

500MHz... wow. I doubt that, but I hope I am proven wrong!

Pretty sure 555 or something is what I saw on HWBot. I could have been mistaken though. Maybe that was the person;s ranking and it was 450's... . I'll have to take a look.

EDIT:

Here's a link:

http://hwbot.org/submission/2939911_drweez_reference_clock_maximus_viii_extreme_552.27_mhz

 

[ZenZimZaliben]

I ran mine up to 225Mhz on the MSI Gaming 7. Anything over 250 and it wouldn't work. I have not tested it under cold at this time. I have a Giga board landing today and we will see if I can get past 250 on it...

It really doesn't 'take stress off the CPU/multi' though... If you hit a multi wall, it could help in that case. Otherwise, it just helps with granularity of the overclock (CPU/Memory).

500MHz... wow. I doubt that, but I hope I am proven wrong!

That is exactly what I meant. Multi Wall. I had several chips, older that i had to drop multi by 1x and raise FSB and suddenly it would work at Freq that was not attainable. I want to say it was my Q9650 or Q6600...long time ago.

 

[EarthDog]

Unless you go cold, you won't hit a multi wall. I tried 5GHz on my chip up to 1.5v and tried BCLK/Multi combos... nada. There wasn't a hole/wall on my chip... at least with ambient cooling. Perhaps there is one higher? Not sure. I have seen 64x on some screenshots, but that was with LN2 of course.

EDIT: 470 is the highest at Hwbot at this second. Well, at least for 'reference clock' anyway. Perhaps another bench had higher???
http://hwbot.org/submission/2940107_fugger_reference_clock_maximus_viii_gene_469.99_mhz

EDIT: There is a 6.8GHz validation as well (68x multi). So I doubt there is a multi wall like sandybridge had.
http://hwbot.org/submission/2940515_toppc_cpu_frequency_core_i7_6700k_6801.63_mhz

 

[radrok]

I think your best bet is phase change, that will help with stability and durability. If you are willing to put up with noise and power consumption.

There's quite a few good unit on the market able to handle 300W+ at sub-zero temperatures.

 

[EarthDog]

That really isn't a 24/7 solution though...

 

[ZenZimZaliben]

Yeah I am not going to get into sub-ambient cooling. I had an old peltier cooler a ways back on a opteron 165 and it was just a freaking hassle with dew point and insulation. You know you have problems when your motherboard starts sprouting mushrooms. j/k...lol

 

[cadaveca]

Yeah I am not going to get into sub-ambient cooling. I had an old peltier cooler a ways back on a opteron 165 and it was just a freaking hassle with dew point and insulation.

So you run 200 BCLK with 3600 MHz ram @ 18 multi, 4.8 GHz CPU @ 24x multi, 4.2 GHz cache @ 21 multi. Sounds great to me.

 

[ZenZimZaliben]

Yeah I will find that optimal balance..just wanted 5ghz to be my end point...But 4.8 isn't going to be anything to cry about.