Intel Skylake / Kaby Lake

Page 327 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
If the 3.7Gh pentium with HTT comes true the whole i3 class will be hard to be argued for...
60-70% more expensive for barely 15% more speed(before o/c) is a hard sale,unless the i3 will have some other important feature.
You can typically get an i3 with more cache than available on a Pentium such as the i3-63XX with 4mb, and i3 has AVX.
 
Aug 11, 2008
10,451
642
126
Jeez,2.7Ghz is celeron speeds, coupled with the slowest ram and not even DDR4 but DDR3 and a 300w PSU (bare minimum for the 1050ti) ,if the harddrive is a 5400rpm one then we're set...

Yes ok it's good value for $400 with a windows licence but it's also right up there with the worst things you could get.
Actually there are a *lot* worse things you could get (i3, pentium, or even atom desktops, or pretty much any AMD cpu). I still would take that cpu over a faster clocked i3, it turbos to 3.1 on all cores. 5400 rpm hard drive and the really slow ram are pretty lame though. But put a 1050Ti in there, and it would be a decent gaming system for an entry level gamer.
 

AtenRa

Lifer
Feb 2, 2009
14,001
3,357
136
Thanks for the link, I see the Core i5 2500K has the following turbo behavior:

1C 3.7
2C 3.6
3C 3.5
4C 3.4

This compared to Core i5 6400 turbo:

1C 3.3
2C 3.3
3C 3.2
4C 3.1

That makes the lead of Core i5 6400 over stock speed Core i5 2500K larger than I thought.

Core i5 6400 at 3.1GHz (4 Cores) is close to 4GHz Core i2 2500K due to ~30% higher IPC(Faster Caches etc), which is much better than the Core i3 6300 (3.8GHz) in almost every game.
Core i5 7400 at 3.3GHz will be even better (close to 4.3GHz SandyBridge), for $20-30 more its way better than the unlocked Core i3.
Also, using non-K OC BIOS, you may be able to OC the locked i5 as a bonus making it even better choice over the overpriced unlocked Core i3.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Actually there are a *lot* worse things you could get (i3, pentium, or even atom desktops, or pretty much any AMD cpu). I still would take that cpu over a faster clocked i3, it turbos to 3.1 on all cores. 5400 rpm hard drive and the really slow ram are pretty lame though. But put a 1050Ti in there, and it would be a decent gaming system for an entry level gamer.

It can also mount an 2.5" SSD.

In fact, I just realized if the 2TB 3.5" drive was mounted in the top position (and a 2.5" SSD used in the lower position) a longer dual fan power connector-less GTX 1050 Ti like this one could be used:

14-487-292-06.jpg


So not just the shorty cards like this GTX 1050 Ti:

14-487-290-01.jpg
 

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
Got another idea, folks.

You know how people keep using the term 14FF or 16FF to refer to 16nm process, as if finFET is something extremely significant and revolutionary? So why won't we use a similar term to distinguish Intel's 10nm from other people's seemingly equal 10nm?

Call TSMC and Samsung 10nm: 10FF.
Call Intel 10nm: 10QW or 10III-V or 10QW-III-V.

With QW stands for quantum well. Of course we don't yet know if that will be the case.
 
Mar 10, 2006
11,715
2,012
126
Got another idea, folks.

You know how people keep using the term 14FF or 16FF to refer to 16nm process, as if finFET is something extremely significant and revolutionary? So why won't we use a similar term to distinguish Intel's 10nm from other people's seemingly equal 10nm?

Call TSMC and Samsung 10nm: 10FF.
Call Intel 10nm: 10QW or 10III-V or 10QW-III-V.

With QW stands for quantum well. Of course we don't yet know if that will be the case.

You really think that Intel 10nm will have III-V or Quantum Well?

Sorry to burst your bubble but it's almost certainly just a straightforward evolution of 14nm :(
 

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
You really think that Intel 10nm will have III-V or Quantum Well?
I've seen many arguments/data/projections/forecasts/... pro and con, and in my view the pros win out.

Sorry to burst your bubble but it's almost certainly just a straightforward evolution of 14nm :(
Where does the "almost certainly" come from? Intel has their mouths firmly shut, so it's anyone's guess. But given how long it's been in development, I think they have solved all the problems. Even Global Foundries has at least SiGe at 7nm.
 

Sweepr

Diamond Member
May 12, 2006
5,148
1,142
131
Nice leak from BenchLife!

- Coffee Lake confirmed to be Intel's first mainstream hexa-core CPU
- 149mm² for a 6C/12T Coffee Lake, only an extra 23mm² compared to the quad-core version
- 6C+GT2 is actually smaller than 4C+GT3e (149mm² vs 185mm²)
- Coffee Lake-S will be available for desktops earlier, in February 2018, about a year after Kaby Lake-S
- Intel lists a 37.5 x 37.5 package for LGA 1151, so no socket change for now?
- Coffee Lake-X belongs to Socket R (HEDT) - just like Kaby Lake-X - looks like Intel plans to release special versions of its mainstream processors on the enthusiast platform from now on - giving us a choice between higher core count or the best single thread performance
- Graphics are still Gen 9/Gen 9.5, makes us wonder if these CPU cores are based on Skylake or Cannonlake microarchitecture
- No dual-core SKU. Will duallies return with Icelake or is Kaby Lake-S the last desktop family to sport 2C/4T chips?


coffee-lake.jpg


https://benchlife.info/intel-coffee-lake-with-14nm-process-will-launch-2018-11192016/
 
Last edited:

jpiniero

Lifer
Oct 1, 2010
14,610
5,227
136
Nice leak from BenchLife!

Just couldn't help yourself even though it was already posted, huh.

Sandy/Ivy and Haswell/Broadwell are also 37.5x37.5 so that doesn't mean much on the socket end. Seems pretty crappy if it's just the extra cores and nothing else, not even the GPU gets updated.
 

Sweepr

Diamond Member
May 12, 2006
5,148
1,142
131
Just couldn't help yourself even though it was already posted, huh.

No need to be rude, I was just commenting on the leak, and posting what mikk didn't.

Seems pretty crappy if it's just the extra cores and nothing else

Actually that's what some people have been asking for years. Many don't care about the iGPU or the rest.
 

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
https://benchlife.info/intel-coffee-lake-with-14nm-process-will-launch-2018-11192016/


Basically we have confirmation that CFL is based on KBL architecture and comes with 6+2 for desktop as well. They said CFL-S comes around February 2018 and CFL-H around April. Nice to see the Die size of these processors. 6+2 only 149 mm².
CFL 4+2 with Gen9.5 is only 4mm² bigger than SKL 4+2. So much for the talk of 14nm+ density.

Those release dates, though. KBL didn't even make 2016 for a big part. Prolly 'cause Intel has no immeditate plans to transition to 10nm for desktop.

And also: why doesn't CFL have Gen10?
 

mikk

Diamond Member
May 15, 2012
4,141
2,154
136
CFL 4+2 with Gen9.5 is only 4mm² bigger than SKL 4+2. So much for the talk of 14nm+ density.


How can you know the exact number from SKL when we didn't get any offical numbers from a leak or Intel itself? 14nm+ density should be the same as before, performance should differ.
 

jpiniero

Lifer
Oct 1, 2010
14,610
5,227
136
You know, I think Skylake 2+2 is something like 93 or so... I wonder if Coffee Lake 6+2 only has 8 MB of L3 still.

WiGig sounds kind of cool although it's pretty limited in terms of distance.
 

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
Any one else wants to guess what the bubbles mean concretely :)?

intel-14nm_large.png


My guess would be

At 65nm they add tock (2nd bubble)
At 22nm they add refresh (3rd bubble)
At 14nm they add a fourth wave (coffee lake) with 3rd wave being the optimization on + process
At 10nm they add 10nm++ (after again a fourth wave cause they point to last bubble)

However it seems a bit odd that there would be 5 full generational products of 10nm while their PAO only prescribes 3. And given that those bubbles go well past the 2020 mark, it seems they will still release stuff on 10nm even after 7nm is online. So that goes a bit against the trend where companies were proud they had their fastest node ramp ever (eg Intel 45nm).
 

Dave2150

Senior member
Jan 20, 2015
639
178
116
Just couldn't help yourself even though it was already posted, huh.

Sandy/Ivy and Haswell/Broadwell are also 37.5x37.5 so that doesn't mean much on the socket end. Seems pretty crappy if it's just the extra cores and nothing else, not even the GPU gets updated.

I think updating this thread is very important to Sweepr (for whatever reason). He always reposts things even if someone else reposted them first.

Judging from his dedication, he either thinks of this as a job, is a major Intel shareholder, or both. Either way best leave him to it :)