Intel Skylake / Kaby Lake

Page 101 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Tovarisc

Member
Jun 12, 2015
50
0
0
i7-6700K review: http://iyd.kr/758

Haven't looked through it all, but they've actually tested 4790K, 5775C and the 6700K at the 3.6GHz to compare IPC. Looks pretty decent, actually... Around 13% higher IPC than Haswell.

EDIT: Only 9% faster than the 4790K with both at stock, though. Guess that's due to the higher boost frequency of the 4790K.

Something funky going on with the gaming results, though. They're actually down...

I was just about to comment on gaming performance too. 6700K basically dominates your usual computing tests in comparison to previous-gen and gives that nice performance increase, but in games it just doesn't perform being at same level or worse.

I wonder what gives and causes drop in gaming performance?

9-9-9-24 with DDR3, 16-18-18-36 with DDR4.

Would memory cause so noticeable difference in gaming performance?
 
Last edited:

Qwertilot

Golden Member
Nov 28, 2013
1,604
257
126
I really don't think judging this by purely the top end unlocked quads is entirely fair when it looks like those chips are beginning to cease to make proper sense.

It looks very arguable whether you get enough over the locked 65w versions for the added cost and/or power. Unlocked Broadwell was keeping up fine at 65w of course.

The speed on the 35w versions is fun for SFF desktops, very nice for laptops etc but this is all also going to eventually benefit a lot of multi core high performance chips. Look at stuff like the Broadwell Xenon D and presumably the eventual Skylake E.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Would memory cause so noticeable difference in gaming performance?

Yes.

metro-1.png

hitman-1.png

thief-1.png


Also more from Anandtech here:
http://www.anandtech.com/show/7364/memory-scaling-on-haswell/8
 
Last edited:

Brunnis

Senior member
Nov 15, 2004
506
71
91
The start of that review makes it look like they are comparing a 4.4GHz 4790k to a 4.2GHz 6700k. Am I interpreting that right?
Yep, that's the stock single core turbo frequencies of those chips.

Would memory cause so noticeable difference in gaming performance?
Possibly, but the thing is that they're using 3000MHz DDR4. Those timings at 3GHz should produce almost identical latencies as DDR3 1600MHz 9-9-9... So it still doesn't explain why it makes for worse results than Haswell.
 
Last edited:

Absolute0

Senior member
Nov 9, 2005
714
21
81
13.25% IPC increase over HW is big

HW is what, 15% IPC increase over SB?

(1.15)(1.1325) = 30.24% IPC increase for SB-->SKL

So 4.2 Ghz SKL = 5.47 Ghz SB
 

inf64

Diamond Member
Mar 11, 2011
3,685
3,957
136
I was just about to comment on gaming performance too. 6700K basically dominates your usual computing tests in comparison to previous-gen and gives that nice performance increase, but in games it just doesn't perform being at same level or worse.

I wonder what gives and causes drop in gaming performance?



Would memory cause so noticeable difference in gaming performance?

It's a bit odd, especially since Broadwell does perform well in games thanks to that eDram it has. 13.2% IPC jump over HW in compute is really nice achievement but really not a reason for HW users to upgrade. I doubt even IB users should care. SB users on the other hand might get a solid boost especially if they don't OC.
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,710
182
106
Devil's Canyon was a generational bump itself

Ok, well if you consider Devil's Canyon (Haswell Refresh) a separate CPU generation, then when comparing Haswell->Skylake it is 3 CPU generations we are talking about.

And we got what... 10% or 15% performance increase across those three CPU generations? I can't believe how low we have set our expectations if that is something to cheer about.
 

Abwx

Lifer
Apr 2, 2011
10,854
3,298
136
Nice. :)





2274AE3A55BED85832C6C0






For IPC comparisons remmember Core i7 6700K and Core i7 4790K are not running at equal clocks.
3.6GHz Skylake-S was 13.25% faster than 3.6GHz Haswell on average (17 benchmarks). About what you would expect from an Intel tock.

FP benches scales perfectly with frequency, if the scores for the 4790K@3.6 are genuine then the stock frequency score is at 4.3/4.38GHz for Cinebench and Povray respectively.

So either the score at stock is wrong or the score at 3.6 is wrong, but both cant be right at the same time.
 

Sweepr

Diamond Member
May 12, 2006
5,148
1,142
131
Ok, well if you consider Devil's Canyon (Haswell Refresh) a separate CPU generation, then when comparing Haswell->Skylake it is 3 CPU generations we are talking about.

And we got what... 10% or 15% performance increase across those three CPU generations? I can't believe how low we have set our expectations if that is something to cheer about.

Let me repeat this since you pretend you didn't understand (trolling?). Devil's Canyon was always meant to replace Haswell, you will not find references to a quad-core + GT2 graphics desktop Broadwell anywhere. Broadwell-K belongs to a new segment, lower TDP 4C+GT3e/GT4e at a higher price bracket.
From a Tock vs Tock perspective Skylake-S is totally in line with its predecessors (if not slightly better). Also your misleading 5% a year mantra might have to be revised and I'm not even talking about the higher-clocked locked 35W-65W Skylake-S parts (compared to 35-65W Haswell Refresh), there's more to Skylake-S than Core i7 6700K you know.

Arachnotronic said:
>13% perf/clock improvement over Haswell? Looks like a good tock to me.

Indeed.

Qwertilot said:
I really don't think judging this by purely the top end unlocked quads is entirely fair when it looks like those chips are beginning to cease to make proper sense.

It looks very arguable whether you get enough over the locked 65w versions for the added cost and/or power. Unlocked Broadwell was keeping up fine at 65w of course.

The clock boost + IPC bump should make the 35W parts more than fast enough for most people. 'T' lineup is looking very tempting, hopefully there will be some reviews later this month or early September.
 
Last edited:

Fjodor2001

Diamond Member
Feb 6, 2010
3,710
182
106
Let me repeat this since you pretend you didn't understand (trolling?). Devil's Canyon was always meant to replace Haswell, you will not find references to a quad-core + GT2 graphics desktop Broadwell anywhere. Broadwell-K belongs to a new segment, lower TDP 4C+GT3e/GT4e at a higher price bracket.

Doesn't matter. Broadwell was still a new CPU generation no matter how you try to spin it. The fact that Intel decided to release only a few placeholder Broadwell SKUs on desktop since it got so delayed does not change that fact.
 
Last edited:

Sweepr

Diamond Member
May 12, 2006
5,148
1,142
131
Doesn't matter. Broadwell was still a new CPU generation no matter how you try to spin it. The fact that Intel decided to release only a few placeholder Broadell SKUs on desktop since it got so delayed does not change that fact.

Lol, nobody is saying it isn't no matter how you try to spin it.
3rd time now, here we go: Broadwell did not replace any Haswell part on desktops. Instead of Broadwell, desktops got Haswell Refresh and Devil's Canyon.

By the way Skylake-S was 10.6% faster than Broadwell-K with eDRAM (clock per clock), quite a bit more than I was expecting. :)

2370A03A55BED85B341E7F
 
Last edited:

Headfoot

Diamond Member
Feb 28, 2008
4,444
641
126
I wonder if there will be a Skylake + eDRAM SKU? Maybe that's the thing they're holding back so that Kabylake has something to give it a boost
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Doesn't matter. Broadwell was still a new CPU generation no matter how you try to spin it. The fact that Intel decided to release only a few placeholder Broadwell SKUs on desktop since it got so delayed does not change that fact.

Broadwell was never to get anymore desktop SKUs than the C models. Stop making up stories. We know you hate the company, move on.
 
Aug 11, 2008
10,451
642
126
This arguing about "generations" is kind of ridiculous. It is simple. Haswell originally came out in May I think, 2013. It is now 2 years later, so whether you consider Haswell refresh or the basically aborted Broadwell desktop the next generation, Skylake is for all practical purposes the second generation past Haswell. So if the 10 to 15 percent gains hold up, it is OK, not great, but not terrible as the usual suspect(s) will try to portray it. Basically you are looking at similar to ivy/haswell progress. Ivy to Haswell was something like 5% plus 8% or 13%, and Skylake is looking to be 10 to 15% over Haswell.

It is just so obvious that they need a mainstream hex core that it almost smacks you in the face. I dont think Intel is deliberately holding back IPC gains, but it seems they have pretty much wrung out all they can from tweaking quad cores, except for the igp of course.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,229
9,990
126
It is just so obvious that they need a mainstream hex core that it almost smacks you in the face. I dont think Intel is deliberately holding back IPC gains, but it seems they have pretty much wrung out all they can from tweaking quad cores, except for the igp of course.

I think that if mainstream Core CPUs (with IGP) expanded to six cores, then they might well eclipse the -E series CPUs. Reason being, would be game streaming. Six cores for gaming, plus QuickSync encoding for the streaming, could be quite a potent package.
 

tenks

Senior member
Apr 26, 2007
287
0
0
Non of the above are games. In games the difference between a 4GHz Sandy and 4GHz Broadwell at 4K is ZERO in the vast majority of games. You are way too much GPU limited in that resolution. Even at 1080p the differences are not worth mentioning. For most people with Sandys, it will be better to upgrade their GPUs even if they only game at 1080p.

You're comparing a 95w chip made for the desktop (Sandy) to a 65w Mobile-first Chip (Broadwell). ANd broadwell still keeps up. Imagine if it was designed for the desktop and given 30 more watts?

That's why Skylake, a full 95w chip meant for the desktop, is going to be a great upgrade for Sandy owners.
 

majord

Senior member
Jul 26, 2015
433
523
136
The only thing giving a 13% average is those couple synthetic of SIMD Sandra results.

Applications only is much lower..

8% over HW clk/clk
4% @ stock frequencies, which Is pretty paltry

The later indicating the 4790K must turbo higher even with MT loads

Handbrake was a big improvement - 12% clk/clk

Clearly been some big changes in key areas of the uARch, so I think we'll see varying results when reviews come out depending what they throw at it

the lower turbo frequency is odd though given it really needs more frequency to distance itself from HW.
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,710
182
106
Broadwell did not replace any Haswell part on desktops.
There are Broadwell desktop SKUs. The fact that Intel decided to skip some SKUs because Broadwell got so delayed does not matter. Broadwell is a new Intel CPU generation no matter how you try to spin it.
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,710
182
106
Broadwell was never to get anymore desktop SKUs than the C models. Stop making up stories. We know you hate the company, move on.

Doesn't matter. Broadwell is still a new Intel CPU generation. What SKUs Intel decide to release based on each CPU generation is another story.