Intel Skylake / Kaby Lake

Page 601 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Sweepr

Diamond Member
May 12, 2006
5,148
1,142
131
Need everyone to step back from the AMD-Intel debate for a bit. [Not a literal command]

Core i5-8250U, 8350U, i7-8550U, 8650U are Kaby Lake-R[efresh], manufactured on 14 nm +. Business is business, but manufacturing process all lined up in 7th generation, when all processors were 14 nm +. Now 8th generation has 14 nm +, 14 nm ++, and 10 nm.

Is the per-core Turbo frequencies of Kaby Lake-R known yet?

And Facebook stream in an hour.

Close friend obtained Core i7-8650U Turbo clocks directly with Intel. 4.2 GHz for 1-2 core and 3.9 GHz 4-core Turbo. Huge gen-to-gen improvement and the fact that they managed to pull this while still on 14nm+ is impressive.
 
Last edited:
  • Like
Reactions: Phynaz

PeterScott

Platinum Member
Jul 7, 2017
2,605
1,540
136
Well, it blows away the 7600K, IMHO. Just because a task is "lightly threaded", doesn't mean that the user isn't running a dozen of them at once, A/V in background, music, web browsing, Skype, etc.

Most of those things like music, web browsing, etc... are trivial loads, and most users won't be doing dozens of them, that is going to be some kind of ultra rare edge case.

The real use for the >4 core CPUs is some kind of fully threaded rendering task, and then you don't have to contrive edge case scenarios. If you are rendering you will use all available cores to the fullest.

But if you are not regularly doing some kind of rendering, you probably don't need more than 4 cores. I do a fair bit of video compression, so I wouldn't mind more cores. So I would probably choose a Ryzen 1600 over a 7600K for similar money.

But I would probably choose 8600K 6 core over the Ryzen, for some combination of good enough rendering and better single thread performance.

I will have to see benchmarks compared to my use cases when it comes down to a final decision.
 
Mar 10, 2006
11,715
2,012
126
15W with quads was always possible at 14/14+. They had to lower base clocks to fit the 15w TDP. Intel did not need to do it in the past as there was no competition. Now with Raven Ridge 4C/8T at 15w TDP scheduled to launch by year end thats no more the case. Anyway consumers can rejoice that quad core comes to ultrathin notebooks.

Actually, I think they added more cores because it's hard to push ST perf up meaningfully.
 
  • Like
Reactions: Ajay

dooon

Member
Jul 3, 2015
89
53
61
Coffee Break.
tzEBc3T.png

https://www.facebook.com/Intel/videos/10154709486191850/
 

psolord

Golden Member
Sep 16, 2009
1,875
1,184
136
Actually, I think they added more cores because it's hard to push ST perf up meaningfully.

You are correct, but the very aggressive turbos that have been leaked, if true, will do exactly that.

4,7Ghz for single core boost, when the threading is low, is just crazy.

The 8700k will probably be the only cpu I will not fiddle with even for 100Mhz more, for day to day use.
 
  • Like
Reactions: Phynaz and Zucker2k

dullard

Elite Member
May 21, 2001
24,998
3,325
126
Why would you expect a big IPC increase? Intel has been massaging, refining and tuning the same Core uarch for a decade now. There really isn't much left to pick. With process shrinks they can throw more transistors at it, but any gains are not going to be some huge breakthrough.
He did say "Tangible" not "Big".

The easy IPC gains are in memory limited tasks. DDR-2666 is 11% faster than DDR-2400. For those memory bound applications, IPC can go up as much as 11% from that change alone.

No one should expect big IPC changes though.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
There will be no IPC increase, only additional cores.

Traditionally, process refinements like Kaby Lake and Coffee Lake have been referred to as CPU stepping instead of new codenames. Skylake, Kaby Lake, and Coffee Lake are fundamentally the same architecture.
Yes, Intel changed the way they do things.

I believe the leaks show an IPC increase with CL, though.

This is technically the 4th iteration of 14nm. The Haswell shrink to Broadwell, then Skylake, Kaby Lake, and now Coffee Lake.

Hopefully Intel has it perfected with CL.
 

Bouowmx

Golden Member
Nov 13, 2016
1,138
550
146
Close friend obtained Core i7-8650U Turbo clocks directly with Intel. 4.2 GHz for 1-2 core and 3.9 GHz 4-core Turbo. Huge gen-to-gen improvement and the fact that they managed to pull this while still on 14nm+ is impressive.

When U is faster than H..
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
15W with quads was always possible at 14/14+. They had to lower base clocks to fit the 15w TDP. Intel did not need to do it in the past as there was no competition. Now with Raven Ridge 4C/8T at 15w TDP scheduled to launch by year end thats no more the case. Anyway consumers can rejoice that quad core comes to ultrathin notebooks.
Ryzen isn't the reason for everything.

I keep talking about base clocks, and I keep getting told that chips never run at base clocks so they don't matter.
 
  • Like
Reactions: pcp7
Mar 10, 2006
11,715
2,012
126
Ryzen isn't the reason for everything.

I keep talking about base clocks, and I keep getting told that chips never run at base clocks so they don't matter.

Base clocks matter in mobile, but they tend to represent kind of the "worst case" scenario.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
He did say "Tangible" not "Big".

The easy IPC gains are in memory limited tasks. DDR-2666 is 11% faster than DDR-2400. For those memory bound applications, IPC can go up as much as 11% from that change alone.

No one should expect big IPC changes though.
I'm also expecting impressive memory speeds with the 8700K. :D
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
Base clocks matter in mobile, but they tend to represent kind of the "worst case" scenario.
Wouldn't we need to compare the 2 core turbo speed of the 4 core chips to the 2 core chips base clock?
Of course the base clock of a 4 core chip might be lower than that of a 2 core chip, but what speeds do they each run at with 2 cores in use?
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
Actually, I think they added more cores because it's hard to push ST perf up meaningfully.

Increasing ST perf was always hard after Haswell. It did not suddenly become so as you seem to suggest.

Ryzen isn't the reason for everything.

I keep talking about base clocks, and I keep getting told that chips never run at base clocks so they don't matter.

Sorry but Intel did the minimum necessary for mainstream desktop all these years because there was no competition. Things will pick up pace now that competition has returned to the PC industry.

1. Did you ever see Intel release a Medium core count die based HEDT CPU all these years - No
2. Did you see a 6 core Intel mainstream desktop CPU all these years - No
3. Did you see a Intel 4C/8T 15W notebook CPU all these years - No.

Were all of these possible. Definitely. Did they happen - No. So I do not agree with you on this.
 
  • Like
Reactions: USER8000

pantsaregood

Senior member
Feb 13, 2011
993
37
91
Yes, Intel changed the way they do things.

I believe the leaks show an IPC increase with CL, though.

This is technically the 4th iteration of 14nm. The Haswell shrink to Broadwell, then Skylake, Kaby Lake, and now Coffee Lake.

Hopefully Intel has it perfected with CL.

Process iterations alone don't yield performance improvements. Broadwell and Skylake used the exact same process, while Kaby Lake modified the process, and Coffee Lake modifies the process as well.

I would expect no IPC gain. Higher memory works just as well on Skylake/Kaby Lake, so that's not a valid point.
 

TheF34RChannel

Senior member
May 18, 2017
786
309
136
Are they giving any real details? When will we see the specs (although that is mostly known already), reviews, prices, and availability?

I would like to know this as well, especially when precisely I can hold one in my hand. It doesn't seem like Intel spilled the beans. Unless that October coffee cup is their not so subtle clue, but that's a month wide window; we need an exact date!
 

PeterScott

Platinum Member
Jul 7, 2017
2,605
1,540
136
Increasing ST perf was always hard after Haswell. It did not suddenly become so as you seem to suggest.

Sorry but Intel did the minimum necessary for mainstream desktop all these years because there was no competition. Things will pick up pace now that competition has returned to the PC industry.

1. Did you ever see Intel release a Medium core count die based HEDT CPU all these years - No
2. Did you see a 6 core Intel mainstream desktop CPU all these years - No
3. Did you see a Intel 4C/8T 15W notebook CPU all these years - No.

Were all of these possible. Definitely. Did they happen - No. So I do not agree with you on this.

Did we see an Intel roadmap in 2015, showing >4 core desktop chips. YES!

Coffee lake 6C is not a response to Ryzen, it's a plan B, to 10nm being too late to deliver desktop Cannon Lake .

>4 cores is niche use, IGP is mainstream use, Intel concentrated on improving IGP first, before core counts, which is the sensible tradeoff.
 
  • Like
Reactions: pcp7

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
There will be many of those, just as there are many Ryzen users with GT710 or HD5450s right now.

Um what? Ryzen users with GT710? Most of them seem to have old 7950s or 290s or whatever as a minimum. Haven't seen any with an HD5450.

Looks like it is going to be a hell of a chip for a "rush job". Crushing their competitor in single core performance, while matching it, or better, in multicore performance...with 25% fewer cores. Just imagine what intel could do if they took their time.

Considering who is their competition and what was their reputation before March 2017, I wouldn't be THAT excited. The main accomplishment here is that they managed to extend Skylake/Kabylake's core count without screwing up any other part of the design. And extra L3 is nice.

It would be nice to know what Intel could do "if they took their time". Sometimes it seems like we haven't really seen that since Sandy.

Coffeelake launch in Oct 2017.

So much for August.
 
  • Like
Reactions: .vodka and raghu78

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
Lacking an igp, and being slower than the Intel i7 chips in lightly threaded tasks and gaming, the 1800x should be selling well below $300 to be "viable." The faster 7700k (in gaming and light-threaded tasks) is only $279 at microcenter.

Source? (and not something running DDR4 2xxx)? Here is mine: http://www.anandtech.com/bench/product/1852?vs=1826 <-- The 1700x is about the same price as the 7700k. Notice that they are both neck in neck, despite the 1700x being 1 GHz slower. Ryzen has strong IPC.

At any rate, expect more of the same out of CL. I do hope that future Intel platforms improve IPC, thermals, and raise core count. I've been on a 2600k @ 4.6 GHz (and I can push it to 5.0 GHz if I bump voltage) due to the lack of a compelling upgrade. The extra 10-20% or so performance of newer offerings just doesn't justify upgrading RAM, CPU, and motherboard. This year I am FINALLY going to upgrade, but to Threadripper rather than CL. I have the money to throw at a new machine, Intel just won't give me a compelling offering. I can't believe how much IPC has stagnated since Broadwell.
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
I like performance and don't care who delivers it. There is no question a highly clocked i7-8700K is going to be a beast for typical client workloads (read: 6c/12t or fewer). I doubt it will replace any of my Zeppelin-based (Zen) systems for heavily MT tasks, however. Especially not in the perf/$ or perf/W metrics for running 24/7/365 at stock clocks.

OC should give it a run for the money in heavily MT tasks, but will throw perf/W out the window. Which matters for me because I have so many rigs and have to consider cooling for them.

Still planning to pick up an 8700K+, though I'm unsure if I should wait for Z390 or not.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
Process iterations alone don't yield performance improvements. Broadwell and Skylake used the exact same process, while Kaby Lake modified the process, and Coffee Lake modifies the process as well.

I would expect no IPC gain. Higher memory works just as well on Skylake/Kaby Lake, so that's not a valid point.
Well, I am expecting the 8700K to be able to use memory speeds that the 7700K could not use.
 

StinkyPinky

Diamond Member
Jul 6, 2002
6,761
777
126
I can take October if it stays that way. I just hope it doesn't creep until November or worse. Hopefully it's not October 31.
 

IndyColtsFan

Lifer
Sep 22, 2007
33,656
687
126
The 7740x, 7800x & 7820x also have to be sold, don't they? If they release(d) a mainstream octa core their whole lineup & proposed lineups from a few years back would lose value instantly, faster than the last ethereum flash crash.

Using this logic, Intel would never release products with x cores, because it might kill sales of existing or previous models. CFL will kill sales of the 6700 and 7700 while at the same time, severely impact sales of the 7800. It might have an effect on the 7820 as well, though I'd expect that to be minimal.

Intel were sandbagging on more cores, they can't increase IPC on the fly, as I said to another poster previously. A hypothetical mainstream octa or hexa core not only eats into the existing 6700(k) & 7700(k) sales but also devours the solderless 7740x, 7800x & 7820x in most cases. You don't cut the nose to spite the face, not when you're Intel. This (CFL) is obviously a compromise IMO, we can agree to disagree here.

It might kill sales of the 6700 and 7700, but it might also entice the existing base of 6700 and 7700 users to upgrade. There are still a ton of Sandy, Ivy, and Haswell users out there as well - you might notice I'm still running an overclocked 2600k. That's because I didn't feel there was value going from one quad core to the next, especially with the minimal gains the last 6-7 years. CFL offers a ton more value to me, so this year I'm upgrading (though I can't rule out the 7820 either).
 

psolord

Golden Member
Sep 16, 2009
1,875
1,184
136
Using this logic, Intel would never release products with x cores, because it might kill sales of existing or previous models. CFL will kill sales of the 6700 and 7700 while at the same time, severely impact sales of the 7800. It might have an effect on the 7820 as well, though I'd expect that to be minimal.



It might kill sales of the 6700 and 7700, but it might also entice the existing base of 6700 and 7700 users to upgrade. There are still a ton of Sandy, Ivy, and Haswell users out there as well - you might notice I'm still running an overclocked 2600k. That's because I didn't feel there was value going from one quad core to the next, especially with the minimal gains the last 6-7 years. CFL offers a ton more value to me, so this year I'm upgrading (though I can't rule out the 7820 either).

You are correct. Coffeelake will severely impact the sales of these cpus.

However it will greatly increase the sales of....the whole Coffeelake series and as a direct consequence the sales of Intel in general as well! :D