Intel Skylake / Kaby Lake

[Shivansps]

Intel were originally planning 10c max (and 44 pci-e lanes) for Skylake-X.

There is a 12C version of that slide as well i belive, the thing is, if they in fact limited LGA2066 to 10C, the max LCC die, what whould have happen to the other part of HEDT sales that is Workstations? They sell 2011-3 Workstations up to 22C with Broadwell-E, the entire Xeon E5 V4 lineup.

They cant do what they used to do on Broadwell-E and lower. on that aspect SKL-X whould have been a massive regresion over BDW-E.

 

[blue11]

Show me the benchmarks that you're referring to that measures core clock vs uncore scaling. Don't need LN2 speeds to show its effect.

Go underclock the uncore on your 6900K to 1.5 GHz (2.4 * 3.6 / 5.8) and run Cinebench or whatever. Let me know how your throughput wasn't affected by the uncore that "doesn't matter."

 

[jpiniero]

Coffe Lake was planned long time ago, in fact i think we know about it before we knew about Ryzen. You may complain about the details of non-LCC, but if they come out this year AMD had nothing to do with them.

You're thinking of Cannonlake. Since they canned the desktop version, they needed something new. Hence Coffee Lake.

 

[swilli89]

Why can't you admit that CPU design cycles are years long, and Skylake-X is not a response to Threadripper at all?

CPU Design - Yes. But taking an existing chip design and steering it into a slightly different socket does not take years when you have literally thousands of engineers. Especially when your HEDT platform is basically an enterprise platform to begin with.

And what do you say about this picture? It literally proves your argument is wrong.

 

[blue11]

You're thinking of Cannonlake. Since they canned the desktop version, they needed something new. Hence Coffee Lake.

Has there been any recent news on Cannonlake? Things are looking dire for 10 nm.

 

[krumme]

Core 2 Extreme 4 cores
Sandy Bridge-E 6 cores
Ivy Bridge-E 6 cores
Haswell-E 8 cores
Broadwell-E 10 cores
Skylake-X 18 cores

Intel was predictable like clockwork until yesterday. Then they jumped by 8 cores instead of 2.

Not to continue the discussion but you miss perhaps the most important processor imo the 6c core i7 980x from 2010 that dominated the charts completely at its time:
http://www.anandtech.com/show/2960
999 usd.
And the 4c 965 extreme from 2008 at 999 usd also.
Also a huge jump up from core 2 arch.

 

[wildhorse2k]

May i remind you that X99 had up to 22C on client and now that HEDT and Xeons are running on diferent sockets is not trival to bring a higher SKU?

Some of Gold and Platinum Xeons will be for R4 socket which is variation of LGA 2066 and there will be motherboards that support both, for example https://www.techpowerup.com/233864/gigabyte-c422-ws-bridges-the-gap-between-core-x-and-xeon-gold . So you do not really lose ability to use Xeons as long as you plan ahead. This also means it is not that difficult to rebrand a Xeon into Skylake-X HEDT CPU.

 

[jpiniero]

Has there been any recent news on Cannonlake? Things are looking dire for 10 nm.

Things have gone pretty dark on it. No rumors of any Xeon D either, which would be rather nice given the shrinkage. Yields must still be extremely horrible.

 

[TheF34RChannel]

Has there been any recent news on Cannonlake? Things are looking dire for 10 nm.

None that I have heard about. It seems to have been diminished entirely as a mobile part. Ice lake anyone?

 

[frozentundra123456]

It does exceed 200W. Burning through AVX2 Prime95 but still. Source


hwbot submissions suggest otherwise. I'm not pulling this out of thin air. Plus Intels own 10/15% result with +/- 5 or 7% uncertainty.

Lets wait for more data from various sources in a variety of tests done side by side to determine the difference in IPC. And you do understand the principle of a bell curve, right? 10% plus or minus 5% does not mean 5%. It means the change could be as low as 5%, but it could also be as high as 15%. I am not even sure where you are getting that information from. The only claim I recall Intel making was an increase in performance in mobile, mostly due to higher clocks and the ability to maintain turbo longer.

 

[formulav8]

Things have gone pretty dark on it.

Yeah looks possibly like Intel laid an egg on that process so far.

 

[imported_ats]

Really. I guess all that support over the years for overclocking competitions and whatnot was just liked the faked moon landings?

Just because Toyoda has a racing organization doesn't mean they support street racing. Likewise Ford isn't going to replace your engine under warranty when you put a turbo charger on it and blow it out doing a 1/4 mile.

Intel/AMD's support of overclocking is: sure you can do it but don't cry to us when it doesn't work.

 

[imported_ats]

Of the TIM thing, yes, they did cheaper it out, but im going to wait for actual reviews and OC results, to see how much it does affect.

These conspiracy theories really need to stop. It is completely improbable that Intel switched to paste based TIM in order to save a couple pennies on a product with an ASP above 1k. It is highly probably that they switched to TIM for a real viable engineering reason.

As for the specific TIM they chose, it should be pointed out that Intel designs their parts for 10+ year total lifetimes and FIT rates well into they 6 digits and beyond.

 

[jpiniero]

Yeah looks possibly like Intel laid an egg on that process so far.

The thing is, there haven't been any rumors or leaks on what's exactly wrong with it. We can assume yield is terrible, but it's only an assumption. Icelake of course is probably using EMIB so even if the yield is terrible still by then they should be okay.

 

[Ajay]

[At the bottom of the graphic] With Intel, there is some insurance- at a price (last time I looked it was cheap for 4 core K processors). Anyway, I hope this is the end of the car analogies.

 

[majord]

Anyone worked out the die size
yet?

 

[LTC8K6]

I am so tired of this. Is that your only defense? You just defend profit maximizing no matter what companies do? No matter how much they gouge price or cheap out on materials?

You (a consumer and enthusiast) simply just say "Its called business"? What's the point of any of this discussion if that's your response. We are talking about a CPU that costs a few grand. Its marketed to overclockers. But Intel wants to save a few cents per $2,000 chip. Now those same people have to void their warranty to get the cooling ability that they should be demanding out of the box.

Do you have any actual thoughts as to the issue that I highlighted?

Don't buy it? Then Intel will lose money, right? That 5 cents they saved with the TIM will lose them a much bigger sale, right?
That's how it all works. People either buy your widgets or they don't.
It's business. If they don't buy, you won't be selling your widgets for very long.

Personally I think it's ludicrous to believe Intel uses TIM to save pennies on a processor. I don't understand the thinking that arrives at such a belief.
I know that Intel uses TIM because overall it's the best solution for Intel.

 

[Arachnotronic]

Don't buy it? Then Intel will lose money, right? That 5 cents they saved with the TIM will lose them a much bigger sale, right?
That's how it all works. People either buy your widgets or they don't.
It's business. If they don't buy, you won't be selling your widgets for very long.

Personally I think it's ludicrous to believe Intel uses TIM to save pennies on a processor. I don't understand the thinking that arrives at such a belief.
I know that Intel uses TIM because overall it's the best solution for Intel.

LTC8K6, spot on.

 

[tamz_msc]

Lets wait for more data from various sources in a variety of tests done side by side to determine the difference in IPC. And you do understand the principle of a bell curve, right? 10% plus or minus 5% does not mean 5%. It means the change could be as low as 5%, but it could also be as high as 15%. I am not even sure where you are getting that information from. The only claim I recall Intel making was an increase in performance in mobile, mostly due to higher clocks and the ability to maintain turbo longer.

Let's not confuse things - bell curve refers to normal distribution, these results are from official Intel slides and pdfs and these refer to uncertainty in estimates. When your estimates are of the same significant figures as your uncertainty in said estimate, they are useless.

 

[tamz_msc]

Go underclock the uncore on your 6900K to 1.5 GHz (2.4 * 3.6 / 5.8) and run Cinebench or whatever. Let me know how your throughput wasn't affected by the uncore that "doesn't matter."

Well I know for a fact that uncore on quad cores doesn't affect Cinebench scores, and I provided a source for that.

So I ask you to provide sources and you ask me to do my own tests?

I'm not interested at all in cache frequencies because for my use they aren't critical.

There is at least one guy out here with a 6900K - maybe he's interested in testing your claims.

 

[imported_ats]

Personally I think it's ludicrous to believe Intel uses TIM to save pennies on a processor. I don't understand the thinking that arrives at such a belief.
I know that Intel uses TIM because overall it's the best solution for Intel.

And it should be pointed out that the TIM they do use is some of the best TIM on the market. The pump out with thermal cycling is all but non-existent and the degradation with time is also basically non-existent. Year 1, it is probably behind in thermal performance, but it quickly catches up and surpasses other TIMs as they get subjected to pump out and degradation. Intel isn't designing chips for people who think that replacing TIM on a yearly basis is a normal thing and they never have.

 

[tamz_msc]

[At the bottom of the graphic] With Intel, there is some insurance- at a price (last time I looked it was cheap for 4 core K processors). Anyway, I hope this is the end of the car analogies.

You bringing in overclocking competitions to the mix doesn't mean anything for the average Joe who buys a retail processor.

Intel doesn't endorse overclocking.

Intel has received feedback from customers who want to implement overclocking on eligible processors, that because of the lack of any replacement coverage for the eligible processors, the risk of overclocking is too great. We understand this position, and while we cannot endorse overclocking, we want to provide a limited remedy if issues arise as a result of their decision to enable overclocking.

It's clear as day.

 

[frozentundra123456]

Let's not confuse things - bell curve refers to normal distribution, these results are from official Intel slides and pdfs and these refer to uncertainty in estimates. When your estimates are of the same significant figures as your uncertainty in said estimate, they are useless.

Yes, and that uncertainty is normally distributed. 10% +/- 5% does not equal zero as you are trying to imply. It means the difference can range from 5 to 15 % with the most likely value still being 10%, i.e. the center of the distribution curve.

 

[blue11]

Well I know for a fact that uncore on quad cores doesn't affect Cinebench scores, and I provided a source for that.

They don't affect the scores because (1) on client CPUs, the uncore is already 50% faster than HEDT/server, and (2) overclocking 4.4 to 5 GHz is too small to notice any L3 starvation anyway. You are making IPC claims based on OC of over 60%, which isn't a comparable situation at all.

I am using a locked processor, so I can't adjust my uncore frequency to run the benchmarks you demand. On the other hand, since you are such an expert on overclocking, it should be trivial for you to open whatever OC app you have and type in a lower uncore clock. If you don't want to be convinced, of course nobody can convince you.

 

[moonbogg]

The lack of solder makes these chips look ridiculous, especially given the price and high-end reputation of this segment. When you buy Intel's high-end, you are used to expecting the best. Looks like that can't be expected anymore with this release. That TIM is garbage and will limit the thermal performance of the chips. It makes them unattractive, period. And to preempt those who are about to say it, of course I won't be buying this. What's that now? I should shut up about it? Yeah, like that's going to happen. Admit it. This release is badly tarnished now and it shocked everyone that intel would cut such a corner as this with such a high class of product. All the excuses are not to be believed. They are BS, straight up. Intel has a reason for doing it and its got nothing to do with longevity of the chips. It more likely has everything to do with making sure in the eyes of enthusiasts these chips have much less longevity, precisely until the soldered refreshes come out.