Intel Skylake / Kaby Lake

Page 639 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
dullard

dullard

Elite Member
May 21, 2001
25,055
3,408
126
VirtualLarry said:
I have a feeling, that we as enthusiast observers, that don't (*well, most of us) actually work in the industry, and what we see as a simple tick-tock on paper, is actually probably a myriad different hectic schedules, and it was probably a massive undertaking to keeping them all aligned, to give us the illusion of the tick-tock working out just fine.

Now we're seeing the reality behind it.
Click to expand...
You are correct. Tick-tock only happened a few times at Intel. In the Pentium 4 days there were 4 ticks (Willamette, Northwood, Prescott, Presler). The Atom chips had 3 ticks (Saltwell, Silvermont, Airmont). Haswell had two ticks. Skylake has three.

It is a good goal in design to strive for: Make big new features, then refine. Then big new features, then refine. Software often follows that pattern with unstable releases (new features) and then stable releases (bug fixes only). But far too often, you need more than one round of bug fixes in software all while new features are ready to go. So, they just get all jumbled unless you purposely delay great new features just to stick to an arbitrary pattern. The temptation to release things when they are ready rather than delay products just to stick to an arbitrary pattern gets to be too great.
 
Arachnotronic

Arachnotronic

Lifer
Mar 10, 2006
11,715
2,012
126
mikk said:
Based on the CPUID it is clearly another Kabylake derivate but with a newer stepping than even CFL 6C.


Kabylake 4C 906E9 (Stepping 9)
Coffeelake 6C 906EAh (Stepping 10)
Coffeelake 8C ??? 906EC (Stepping 12)
Click to expand...

Can you confirm this info in the chipset driver? Where is BenchLife getting that screenshot?
 
IntelUser2000

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
dullard said:
You are correct. Tick-tock only happened a few times at Intel. In the Pentium 4 days there were 4 ticks (Willamette, Northwood, Prescott, Presler). The Atom chips had 3 ticks (Saltwell, Silvermont, Airmont). Haswell had two ticks. Skylake has three.
Click to expand...

Tick in Intel definition: Process shrink
Tock in Intel definition: Architectural change.

Silvermont went to an Out of order architecture with significant architectural changes over Saltwell and a 50%+ improvement in perf/clock. A "Tick" it is not. If you insist on calling based on Tick/Tock terminology Silvermont is both Tick and a Tock. Perhaps it gets glossed over for Atom because its not their main core.

Prescott is also a significant architectural change that brought process shrink as well. Unfortunately it was based on the ill-advised ideology of reaching 10GHz clocks so there was no performance gain, despite the architectural change.

Haswell had one tick. The other is a refresh.
 
IntelUser2000

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
VirtualLarry said:
To me, it just wouldn't make sense, to call both the 6C on 14nm++ process, and an 8C on 10nm+ process both "Coffee Lake". You would think that a totally different process, would warrant a new code-name, even if it's the same CPU Cores / same iGPU / same design, basically, just with higher core-count.
Click to expand...

Perhaps not. Perhaps yes?

http://cdn.wccftech.com/wp-content/uploads/2016/09/Intel-Coffee-Lake-is-Cannonlake.png

Ashraf made a reasoning that Coffeelake may be a backported Cannonlake core because of the naming on that slide.

Remember also what Intel's then-PC chief said about Broadwell? He said they made a mistake not having a full fledged desktop line and it hurt their profits. We know if they really did that, we wouldn't have seen Skylake-S in the same year, because they would have had to sell Broadwell-S chips. So despite the fact that it allowed them to pull in the next generation core, they would have preferred to sell full line of Broadwell ones.

With that line of reasoning.

Q1/Q2 2018 - Cannonlake mobile for low power
Q3/Q4 2018 - Cannonlake-S(full Desktop line based on Cannonlake 10nm)

That puts Icelake firmly into 2019. Considering the amount of problems they are having with 14nm and 10nm process, it makes sense to me. There was merely few months between the time when they said Cannonlake would be made into 4 core mobiles and when the rumors started swirling it would only be for 2 core ones. The things are changing so quickly and so bad they can't even see few months ahead. I thought it was weird when they seemed so optimistic they wanted Icelake by Q4 2018 when they can barely get Cannonlake by spring of that year.
 
B

beginner99

Diamond Member
Jun 2, 2009
5,210
1,580
136
Eddward said:
What a mess. Intel future shifted from perfectly aligned tick/tock roadmaps a few years back into unpredictable confusing mess with different overlapping architectures/chipsets/nodes/etc.
Click to expand...

A mess for the consumer but good for intels bottom line. The longer they can milk 14nm, the better for intel. And due to diminishing returns and power not being a huge issue in desktop it was only a matter of time till desktop will be limited to older process.

VirtualLarry said:
I have a feeling, that we as enthusiast observers, that don't (*well, most of us) actually work in the industry, and what we see as a simple tick-tock on paper, is actually probably a myriad different hectic schedules, and it was probably a massive undertaking to keeping them all aligned, to give us the illusion of the tick-tock working out just fine.

Now we're seeing the reality behind it.
Click to expand...

True. And what you describe, being able to see the reality behind is is what I call terrible management.
 
AtenRa

AtenRa

Lifer
Feb 2, 2009
14,001
3,357
136
beginner99 said:
A mess for the consumer but good for intels bottom line. The longer they can milk 14nm, the better for intel. And due to diminishing returns and power not being a huge issue in desktop it was only a matter of time till desktop will be limited to older process.



True. And what you describe, being able to see the reality behind is is what I call terrible management.
Click to expand...

The problem is that even Mobile in 2017 (Gen 8) still uses 2015 14nm process (even if its 14nm++).

As for the 8C 16T mainstream Intel CPU in 2018, im expecting to be on 10nm and launch at the end of H2 2018. If it is on 14nm, Intel will have a lot of problems with ZEN 2 at 7nm in H1 2019.
 
dullard

dullard

Elite Member
May 21, 2001
25,055
3,408
126
IntelUser2000 said:
Haswell had one tick. The other is a refresh.
Click to expand...
Use whatever definitions you want; it doesn't matter Two ticks or a tick and a refresh. My point is the same: the tick-tock concept almost never actually occurred. Is was just a short blip in processor history.
 
moonbogg

moonbogg

Lifer
Jan 8, 2011
10,635
3,095
136
Guru3d had a news item this morning mentioning Coffee Lake might come with an 8/16 CPU and be paired with the Z390 later this year. Sounds impossible, but you never know I guess. They will charge so much for that 8 core that it becomes obsolete I expect. Mainstream market doesn't want $600 CPUs, lol. That price would make sense though. Whatever the competition is charging, just double it and call it good over there at Intel.
Also, if this is true, I'd like to know how Intel suddenly managed to produce an 8/16 core with IGPU at 14nm. Remember? It was impossible due to heat. That's the only reason Intel didn't make a mainstream 8 core yet.
Also, 8700K @ $400 is too damn expensive. Its just too damn much. Ridiculous.
 
  • Like
Reactions: VirtualLarry and psolord
dullard

dullard

Elite Member
May 21, 2001
25,055
3,408
126
moonbogg said:
Also, if this is true, I'd like to know how Intel suddenly managed to produce an 8/16 core with IGPU at 14nm. Remember? Also, 8700K @ $400 is too damn expensive. Its just too damn much. Ridiculous.
Click to expand...
It would not be on 14nm, it would be on 14nm++. 14nm++ has dramatically less heat.

$400 is nothing to many, many consumers. People bought $400 CPUs (heck often $600 to $1000) for decades without blinking. And that was before inflation, so money meant more back then.

I'm still wondering what the exact US price will be. Converting from the Canadian price leak, it will be $395:
https://www.extremetech.com/computing/254521-rumor-prices-leak-intels-coffee-lake
Converting from the Euro price leak, it will be $391: https://www.forbes.com/sites/antony...ee-lake-cpus-six-cores-from-300/#314aa2b03586
Converting from the pound price leak, it will be $400: https://www.anandtech.com/show/1184...fee-lakes-cpus-published-400-for-core-i78700k
 
  • Like
Reactions: pcp7 and Phynaz
Glo.

Glo.

Diamond Member
Apr 25, 2015
5,705
4,549
136
dullard said:
It would not be on 14nm, it would be on 14nm++. 14nm++ has dramatically less heat.

$400 is nothing to many, many consumers. People bought $400 CPUs (heck often $600 to $1000) for decades without blinking. And that was before inflation, so money meant more back then.

I'm still wondering what the exact US price will be. Converting from the Canadian price leak, it will be $395:
https://www.extremetech.com/computing/254521-rumor-prices-leak-intels-coffee-lake
Converting from the Euro price leak, it will be $391: https://www.forbes.com/sites/antony...ee-lake-cpus-six-cores-from-300/#314aa2b03586
Converting from the pound price leak, it will be $400: https://www.anandtech.com/show/1184...fee-lakes-cpus-published-400-for-core-i78700k
Click to expand...
How about 389$ official MSRP from Intel for 8700K? ;)
 
  • Like
Reactions: pcp7
coercitiv

coercitiv

Diamond Member
Jan 24, 2014
6,187
11,858
136
Some info on upcoming NUCs via FanlessTech, needs some checking though: the KBL cores + dGFX resurfaces :)

yT9Y6vx.png


https://www.cnx-software.com/2017/0...2019-gemini-lake-coffee-lake-and-kaby-lake-h/
 
  • Like
Reactions: Jan Olšan
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom