Tuesday, September 5th 2017

Intel Skylake-X HCC CPU Delidded by Der8auer, also not Soldered

Overclocking poster-boy Der8auer has seemingly gotten his hands on some early samples of Intel's Skylake-X high core count (HCC)HEDT CPUs. The upcoming 12 to 18-core enthusiast-class CPUs are being launched on the same X299 platform on socket LGA 2066 that Intel has already launched 4 (Kaby Lake-X), 6, 8 and 10-core parts already, and are supposed to bring Intel towards a level playing field - and then some - with competitor AMD's Threadripper CPUs, which boast of up to 16 cores.

From this delidding process with Der8auer's own delidding tool, Delid-Die-Mate-X, seems to result a die that is much larger - as expected - than Intel's 10-core i9-7900X. At the same time, it seems that Intel is still opting, again, for not soldering its enthusiast-targeted CPUs, which would result in better temperatures and, potentially, overclocking potential. The fact that Der8auer managed to delid the i9-7920X and didn't recommend against doing it likely means that there is minimal risk of damaging your CPU while subjecting it to this process. This is something the renowned overclocker did do when he recommended that users shouldn't delid their Ryzen or Threadripper CPUs looking for better temperatures, since the fact that these were soldered would likely result in both catastrophic damage and a much diminished chance of operating temperatures improvement through the application of special purpose thermal compounds. The Facebook post from Der8auer with the delidded 7920X likely serves as an appetizer for an upcoming delid video on YouTube, as has been the overclocker's MO.

Sources: Der8auer's Facebook, via Overclock 3D
Add your own comment

171 Comments on Intel Skylake-X HCC CPU Delidded by Der8auer, also not Soldered

#1
cadaveca
My name is Dave
phanbuey said:
So I agree with that... buuut there is like a 20C drop when you replace the tim with liquid metal, so there is something to be said for the solder.

Also the CPUs do eventually throttle at stock under avx load... or at least mine does.

Even with a -3 offset within a few minutes of priming the clock starts to dip. Mine is sitting on a AIO at 1.16v @ 4.6 Ghz and while it's fine, i could definitely push higher with a delid.
Right, but as mentioned above, it has been shown that it's not really the TIM that gives those better temps.. it's the removal of the silicone that holds the IHS on, thereby making the IHS that much closer to the chip, that matters.

I require AVX at the same speed for what I use my PC for at times, so I have no negative offset and a relatively high voltage and low speed. I would guess based on your given settings that your chip is only pulling 185W-225W. How much is the offset from "tJ" core readings to "CPU" readings? Plus or minus? My "CPU" reads about 5c higher than the highest core... if yours is @ like 8c or something, that throttle you are seeing might just be PROGRAMMED in, and not really due to ACTUAL temps. Having a clamp-meter giving power use over the 8-pin is really eye-opening, personally. Reading full system power consumption from the wall isn't specific enough.
Posted on Reply
#2
nemesis.ie
cadaveca said:


Like for der8auer's uses, benching sub-zero, yeah, maybe not the most optimal, but that's not how I use my chips, so I could care less about that aspect of it.
You mean you couldn't care less, right? ;)
Posted on Reply
#3
R-T-B
nemesis.ie said:
You mean you couldn't care less, right? ;)
In American english, it is often used interchangeably.

Dumb I know. :laugh:
Posted on Reply
#4
nemesis.ie
A bit like "anyways" instead of anyway, "prolly" instead of probably, "lay" down instead of lie down and likely a load of other stuff? Oh well. :)
Posted on Reply
#5
FR@NK
cadaveca said:
I have no negative offset
Your chip is stable at 4.5 running AVX512 prime95 with no offset?

I plan on getting a 7920x once asus releases their apex board but I fear it will be slower then my current 6900k@4.5GHz.
Posted on Reply
#6
cadaveca
My name is Dave
FR@NK said:
Your chip is stable at 4.5 running AVX512 prime95 with no offset?

I plan on getting a 7920x once asus releases their apex board but I fear it will be slower then my current 6900k@4.5GHz.
I find that my 6950X is faster, and it seems to have to do with the cache change and memory latency. Seems. I need to do way more testing still on that one.

And yeah, that's how I ended up at this clock and voltage; I can actually run 4.6 GHz @ 1.2V perfectly fine but AVX, man, is it hard on these chips. Some boards offer offset for the TYPE of AVX, too, which is nice, so that might be something that interests you; not sure.
Posted on Reply
#7
Mussels
Moderprator
trparky said:
But if Intel closed the gap between the die and the IHS the fact that it isn't soldered shouldn't make any difference. In the past it wasn't the fact that Intel chips use TIM (and not solder) that was causing the issue, it was that there was a gap between the die and the IHS.

Why? Think about it. Do we have to solder our heatsinks and/or waterblocks to the IHS to get better heat transfer? Of course not! Don't be silly. We apply TIM to these components just like Intel does but the difference between what Intel did and what we do is we clamp the heatsync and/or waterblock down really tight against the IHS. The TIM really isn't there to be the heat transfer medium, it's really only there to fill in the microscopic imperfections in the metal to facilitate better transfer of heat.

Now if Intel were able to manage to close the gap between the underside of the IHS and the die then the heat transfer would be just as efficient as it is between the IHS and our heatsinks and/or waterblocks.
its because of the small surface area.

Yes, the gap is definitely important - but anything reducing the heat transfer on such a small area drastically affects the maximum heat transfer.

This is a non issue for normal use of the chips, but these extreme CPU's (K chips, HEDT) really should cater to the enthusiasts... since that is their target market after all.
Posted on Reply
#8
9700 Pro
xkm1948 said:
Looks like a smaller PCB riding on the back of a bigger PCB. Weird huh?
Just like the socket 423 P4 Willamette CPUs back in the day :)

Posted on Reply
#9
Hood
trparky said:
But if Intel closed the gap between the die and the IHS the fact that it isn't soldered shouldn't make any difference. In the past it wasn't the fact that Intel chips use TIM (and not solder) that was causing the issue, it was that there was a gap between the die and the IHS.

Why? Think about it. Do we have to solder our heatsinks and/or waterblocks to the IHS to get better heat transfer? Of course not! Don't be silly. We apply TIM to these components just like Intel does but the difference between what Intel did and what we do is we clamp the heatsync and/or waterblock down really tight against the IHS. The TIM really isn't there to be the heat transfer medium, it's really only there to fill in the microscopic imperfections in the metal to facilitate better transfer of heat.

Now if Intel were able to manage to close the gap between the underside of the IHS and the die then the heat transfer would be just as efficient as it is between the IHS and our heatsinks and/or waterblocks.
The left over Intel silicon glue looks very thick, cleaning it off and relidding using a tiny amount of glue and a good relid press should get things much closer together, and drop temps dramatically.
Posted on Reply
#10
xkm1948
cdawall said:
Oh good god are we about to start the vega nonsense all over again?
Haha, it will involve @RejZoR constantly prasing on RyZen2 and ends up getting a i9 8900K again. All talk of supporting AMD with no action from the wallet.


Back to topic. I assume liquid metal plus deliding service might be super popular. We might even see some modded 7980XE on flea bay with liquid metal resealed in.
Posted on Reply
#11
phanbuey
cadaveca said:
I find that my 6950X is faster, and it seems to have to do with the cache change and memory latency. Seems. I need to do way more testing still on that one.

And yeah, that's how I ended up at this clock and voltage; I can actually run 4.6 GHz @ 1.2V perfectly fine but AVX, man, is it hard on these chips. Some boards offer offset for the TYPE of AVX, too, which is nice, so that might be something that interests you; not sure.
I have noticed that OCing the mesh to the max makes quite a difference.
Posted on Reply
#12
SARVAMANGALAM
i dont now.. levave me cold booth(ryzaripper and 7980XEcognac))
becouse _: Intel Xeon E5 2699 V4 ES QHUP 2.1Ghz 24 Core 55MB 145W LGA2011-3 CPU Processor
http://www.ebay.com/itm/Intel-Xeon-E5-2699-V4-ES-QHUP-2-1Ghz-24-Core-55MB-145W-LGA2011-3-CPU-Processor/272793357092?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2060353.m1438.l2649
48 cores for 1200usd on 2200mhz(write2,1but is always bit more) this ES xeon is it booth "cpu king" rit-reaper and it work on some cheap x99 asrock with ecc ddr4 memo

i have ES xeons witch work 5+ years 24/7/365 100%load or newer x5xeons on avx prime finders or other apps and i7 920 witch work now 7 years on same load..lol soo dont tell me some shh about rma..
i will wait to spring 2018 for more cpus from booth cpu producers .. 18 core will proably king but more expensive as always from intel. epyc will much more expensive,also mobo for epyc. so epyc its pic-out,
mobos for thripper X399 are expensive.. 250usd is enought.. why people need that 600usd asus primeros ?..show me why )) you get same preformance on cheaper

yes intel is one generation behind now and + kicked with glueing" . but amd must work hard becouse they cannot glue more..we dont need cpus like two men hand size )))) next time.. intel is on small process now, just look at this small 24 core, if intel start glueing amd have not chance.. but intll go bit other way .. also bigger cores and more sockets on mobos
more on serverhome..
Posted on Reply
#13
Nephilim666
SARVAMANGALAM said:

Your post may have given me my first migraine. Are you using a language translator that ignores punctuation?
Posted on Reply
#14
RejZoR
It's exactly that situation where every little counts the most.
xkm1948 said:
Haha, it will involve @RejZoR constantly prasing on RyZen2 and ends up getting a i9 8900K again. All talk of supporting AMD with no action from the wallet.


Back to topic. I assume liquid metal plus deliding service might be super popular. We might even see some modded 7980XE on flea bay with liquid metal resealed in.
Ok smartass, how was I suppose to buy Ryzen when 5820K/6800K was a thing? But I guess only you have the power to buy non existent products. Jesus, some of you are really thick.
Posted on Reply
#15
cadaveca
My name is Dave
Mussels said:
This is a non issue for normal use of the chips, but these extreme CPU's (K chips, HEDT) really should cater to the enthusiasts... since that is their target market after all.
Personally, I think that this idea could not be further from the truth. Most "enthusiasts", here on TPU even, can't afford HEDT for the most part. I mean, it's not like we have all sorts of users here with these chips, or X99, or X79, or even 1366. And I mean at launch. Like, with i7-920, at least, 1366 had an affordable chip for "enthusiasts", and then we clocked the crap out of them. But the market today is nothing like it was back then.

Like are you trying to imply that these HEDT CPUs are ONLY for enthusiasts? Then why are they priced at levels that most cannot afford?

I feel what you said is like saying that only enthusiasts buy high-end anything. I think, that those that have the money to afford such luxury are often NOT enthusiasts, purely by nature of who and what they are. Intel knows this, so what might seem to you as a snub to enthusiasts is Intel actually meeting the needs of the real people that buy such products, who, to me, clearly aren't enthusiasts.

What I see from most is "I AM ENTHUSIAST!!! I NO LIKE DIS, YOU SUCK, MAKE ME HAPPY!!!" and to me, that's fairly arrogant, because it ignores the rest of the world, who, actually, in this market, are the majority. NOT enthusiasts.
Posted on Reply
#16
Mussels
Moderprator
if you're going HEDT, you're not throwing a stock cooler on it. Cooling matters more there.
Posted on Reply
#17
cadaveca
My name is Dave
Mussels said:
if you're going HEDT, you're not throwing a stock cooler on it. Cooling matters more there.
Absolutely, but at what point does the TIM actually impact these chips? How they are sold, and run at stock, or at OC only?


The problem with leaving a trail of crumbs is that the birds get fat and lazy, and then they expect more.


phanbuey said:
I have noticed that OCing the mesh to the max makes quite a difference.
even going from 2400 MHz to 2700 MHz can have a significant impact on overall latency. But I found with my CPUs that there can be quite diminishing returns as you clock it up. You really need to see the change in power consumption over 8-pin from stock to manual settings that are exactly the same.
Posted on Reply
#18
Tsukiyomi91
to those who have either TR 1950X or i9-7900X, I respect you guys for picking a side & settled with it. Folks who didn't bought any of those & spouting claims etc are plain jelly while not looking at the bigger picture. Once u settled at any sides, stay there please. Moving from one camp to another makes you no different than a vagrant or those nomadic folks who keep moving every few months to find "greener pastures" to "settle in".
Posted on Reply
#19
Tsukiyomi91
Mussels said:
if you're going HEDT, you're not throwing a stock cooler on it. Cooling matters more there.
There's your answer everyone. If you don't have the money to invest in a good cooling setup for your godly-level silicon, don't even think of OCing on air & expect it to hit the magical number you're expecting it to hit with an inferior cooling system.
Posted on Reply
#20
Mussels
Moderprator
cadaveca said:
Absolutely, but at what point does the TIM actually impact these chips? How they are sold, and run at stock, or at OC only?


The problem with leaving a trail of crumbs is that the birds get fat and lazy, and then they expect more.


even going from 2400 MHz to 2700 MHz can have a significant impact on overall latency. But I found with my CPUs that there can be quite diminishing returns as you clock it up. You really need to see the change in power consumption over 8-pin from stock to manual settings that are exactly the same.
if you're selling a ferrari, you dont sell it with $2 windscreen wipers and tires from K-mart (does american K mart do that? do they even exist? Walmart? whatever)
Posted on Reply
#21
cadaveca
My name is Dave
Mussels said:
if you're selling a ferrari, you dont sell it with $2 windscreen wipers and tires from K-mart (does american K mart do that? do they even exist? Walmart? whatever)
You're right, and that's why these chips don't come with a cooler.


Sadly though, you'll find many of those cars that are priced that high, do have many such "inferior" parts. Especially in the interior. Once again car analogy fails to meet the needs for the PC market.

BTW, you might not be able to buy pilot sport cup 2 tires at kmart, but those that have money... got it by not spending it on silly things like that in the first place. Most people that buy those cars don't buy them to drive them these days; they are purely for investment. In other words, you buy them to sell them later, at a higher price. You can't do that with PCs... only on launches when supply is limited... but then it gets unlimited. Super and Hyper cars never get "unlimited" in numbers, which is why such analogies fail.
Posted on Reply
#22
R-T-B
cadaveca said:
See, I see Intel CPUs with paste TIM and high overhead and think "wow, they made good chips", and then I see AMD's soldered chips and think "hey, they NEED that solder, because these chips have no overhead", and that's good enough for me.
As much as I disagree that good chips should be fed paste as a TIM material, I get what you are saying here. It's sad but true: AMD really needed the solder. The Ryzen zeppelins are pretty thermally sensitive.
Once again car analogy fails to meet the needs for the PC market.
I feel the sudden need to kick you into a nameless pit screaming "This is TPU!" to reply to this.
Posted on Reply
#23
cadaveca
My name is Dave
R-T-B said:
I feel the sudden need to kick you into a nameless pit screaming "This is TPU!" to reply to this.
Meh. The proper approach to a car analogy for HEDT is to call them pick-ups, not Ferraris. Like, do you know what a dualie diesel costs? Anywhere from 3x to about 4-5 times the cost of a "normal" car. Normal mainstream CPUs cost $400 or so, right, HEDT costs what? 1200? 1400?

Do you see where I am going? That pick-up, does it come with steal wheels, or aluminum? Did you have to pay extra to get a color other than white, black or grey?


There are no Ferraris in the PC world.
Posted on Reply
#24
R-T-B
cadaveca said:
Meh. The proper approach to a car analogy for HEDT is to call them pick-ups, not Ferraris. Like, do you know what a dualie diesel costs? Anywhere from 3x to about 4-5 times the cost of a "normal" car. Normal mainstream CPUs cost $400 or so, right, HEDT costs what? 1200? 1400?

Do you see where I am going? That pick-up, does it come with steal wheels, or aluminum? Did you have to pay extra to get a color other than white, black or grey?


There are no Ferraris in the PC world.
I actaully agree the car analogies are weak, but they are part of our blood here now.
Posted on Reply
#25
Valantar
Cool Vibrations said:
Don't worry it's "not a furnace" so it's okay...

80c lmao what a joke
80C for a CPU under load is no problem whatsoever. 90-95, we'd be talking, but generally CPUs are perfectly happy at temps like that. We've become spoiled by the efficient, small CPUs of late. A decade ago, keeping an OC'd CPU below 90 degrees was impressive. Keeping a 300+W monster at 80 still is.
Posted on Reply
Add your own comment