Monday, August 10th 2015
Intel Skylake De-lidded, Reveals Tiny Die
When Japanese tech publication PC Watch got under the hood (lid) of a Core i7-6700K quad-core processor, what they found was an unexpectedly small silicon, that's shorter in proportion to its width, than previous dies from Intel, such as Haswell-D, and Ivy Bridge-D. It's smaller than even the i7-5775C, despite the same 14 nm process, because of its slimmer integrated graphics core with just 24 execution units (compared to 48 on the i7-5775C), and the lack of an external 128 MB SRAM cache for the iGPU.
The substrate Intel is using on the i7-6700K was found to be slimmer than the one on the i7-4770K, at 0.8 mm thick, compared to 1.1 mm on the latter. The thicker IHS (integrated heatspreader) makes up for the thinner substrate, so it shouldn't cause problems with using your older LGA1150 coolers on the new socket. Intel is using a rather viscous silver-based TIM between the die and the IHS. The die is closer to the center of the IHS than its predecessors were. PC Watch swapped out the stock TIM with Prolimatech PK-3 and Cool Laboratory Liquid Pro, and found some impressive drops in temperatures at stock speed (4.00 GHz) and with a mild overclock (4.60 GHz).
Source:
PC Watch
The substrate Intel is using on the i7-6700K was found to be slimmer than the one on the i7-4770K, at 0.8 mm thick, compared to 1.1 mm on the latter. The thicker IHS (integrated heatspreader) makes up for the thinner substrate, so it shouldn't cause problems with using your older LGA1150 coolers on the new socket. Intel is using a rather viscous silver-based TIM between the die and the IHS. The die is closer to the center of the IHS than its predecessors were. PC Watch swapped out the stock TIM with Prolimatech PK-3 and Cool Laboratory Liquid Pro, and found some impressive drops in temperatures at stock speed (4.00 GHz) and with a mild overclock (4.60 GHz).
112 Comments on Intel Skylake De-lidded, Reveals Tiny Die
With the current processors, they aren't at risk of self destructing from heat, they can both handled higher temperatures and are putting out less heat. The only area putting TIM under the IHS would help is the enthusiast market, and you have to face facts, the enthusiast market doesn't even show up as a blip on Intel's customer charts.
I have seen such examples in my home country. When a given satellite provider bought sports content which was locked only to its own TV channels. The legal authorities decided it wasn't fair and now those channels broadcast for more providers, and respectively for more customers.
Expensive sports content is similar to extremely expensive lithography machines from ASML, plus huge engineering costs...
Either way. Access to Fab Plants are not the problem regarding AMD's lacking competitiveness.
I think its more you just have a very skewed out look on the whole thing. No one is stopping AMD from being successful. Why should Intel give AMD access to their own fabs? That is totally ass backwards thinking when it comes to competition. They are Intel's fabs, of which Intel can do what they want with them. AMD uses TSMC and Global Foundaries.
Intel has left the door wide open for AMD to get back into a competitive state, but its AMD year after year that shoots themselves in the foot.
AMD has the full right to turn to the only other fab that is capable - that is Intel..
And I don't see what your problem is - look at Apple and Samsung, not the same situation with different outcome?
Just admit that you are wrong.
Also Samsung has their own fabs too. Which is why some people think if Samsung bought AMD, it would be a good thing for AMD all together. Have access to millions in R&D and massive fabs.
Apple uses Samsung and TSMC fabs.^^This
Another is with Intel bundling their CPUs with a GPU so that neither the lower end cards from either AMD or Nvidia may get sold in greater numbers.
Another one is also to enable sales. Like having "your" consumers spend more and more often.
Like with the recent Skylake where you have some reviews including a "benchmark" Dolphin, that shows 60-70% diff. between "Sandy Bridge" and Skylake where they also state that "Sandy Bridge, Your Time Is Up" and such. This certainly has nothing to do with that that those with a SB PC should feel/think their system is not adequate enough any longer, that they are dated, that they are behind and such should change = buy, right? Psychological rather than rational(relevancy).
Not merely a "business/corporate" issue but a consumer issue as well. Foolish consumers, serventile consumers and consumers licking corporate ass, where they cover for them helps them as well i think. For me its not a AMD vs Intel thing, but a Consumer/user/purchaser one.
As to bundling a gpu in the cpu, that's what AMD's APU are, and they were doing it first. Intel jist played catchup. Because Intel is a well-run company, the made up for lost time quickly.
You're right about AMD not having much for funds to do R&D, marketing, etc. But it's not Intel's fault. Intel has merely done what every company's shareholders demand: Make money. Have you read up on why AMD is where they are now? They have done it to themselves. Bad decision, bad strategy, bad management.
Can they recover? Maybe, with alot of effort, but they are going to have to do it themselves. I hope they do, because no company is ever going to help a competitor, and we need them to compete fiercely.
@RejZoR, if I were you I'd quit waiting and buy the lower end Haswell-E (5820k). It doesn't clock as well as a 920 but it ain't bad and it'll give you 6 nice, soldered cores if nothing else.
2. Solder can be used on all CPUs, it has nothing to do with complexity at all.
3. The issue with the TIM is that it will crack or rather stop being able to transfer heat effectively after a few LN2 sessions. That's why you ca read -115'C on the IHS but +11'C real CPU die temp. At that point the CPU becomes useless for pretty much everything as even stock clocks and VID will run into upwards of 90'C.
4. INTEL has been talking overclocking to us for a while, even more so since 2013. All initiatives and even in the press decks they highlight overclocking. By design in fact, SKylake-S K SKUs was made to allow more overclocking (hence the step-less bclk, more memory ratios, more voltage control, same power plane for Uncore/CPu clock etc.)
The TIM used is in contrast to this ideology and message of promoting extreme OC. The situation is worse than with Haswell or any other CPU before because this time. You will not make 6GHz with a CPU that has not had the TIM replaced. If you do, the CPU throttles and performance tanks (6.3GHz performs like 5.4GHz for instance)
5. For air cooling, the current solution is probably fine for most people, but once you go XOC, it is utterly useless.
Bottom line is: if noone tried to delid them you wouldn't know what the hell is under the IHS and would happily use them without moaning and deal with the temperatures.
And besides, extreme overclocking is way overrated. Who gives a shit about LN overclocking? Do you use a PC like that all day long? :rolleyes:
I have degraded a 3930K by using 1.5v daily and that's 32nm, imagine what it's gonna happen to a node this small with high voltages daily.