Intel Skylake / Kaby Lake

[-sandro-]

are all mobile cpus soc systems?

 

[dark zero]

You do realize that that's not how you add up percentages - or percentage points to be more exact.

Presuming the numbers are right, the performance gain from Nahlem to Skylake would actually be 1.25 * 1.15 * 1.1 * 1.1 * 1.15 = 2.00028 ~ 1.00028 or 100.028% improvement.

That's why I said the first part.. One thing is only IPC, other is REAL performance and at min. It can get from 75%, but hardly I see going doubling the performance

 

[Fjodor2001]

That's why I said the first part.. One thing is only IPC, other is REAL performance and at min. It can get from 75%, but hardly I see going doubling the performance

Huh? Care to translate that into plain English?

And what about your calculation... did you understand the error you made when adding up cumulative performance increases?

Then there's the whole issue of the performance increase numbers from one generation to the next being BS in the first place.

 

[crashtech]

Huh? Care to translate that into plain English?

"I made an arithmetic error."

 

[dark zero]

"I made an arithmetic error."

Not that at all. I've just forgot to add some more factors who nerf that percentages to the real performance. That's all.

 

[cytg111]

wtf

edit: google up compound interest please.

 

[frozentundra123456]

Do you have a source? The graph doesn't look far off. Of course it always differs a bit from benchmark parcours to benchmark parcours. With AVX2 the difference would be much bigger than 10% nevertheless.

It certainly isnt "complete BS". There is some variability depending on what benchmarks you look at, but I would say it is generally accurate. The generally accepted increase for haswell is 8 to 10 %. That said, I am getting tired of seeing it, I have to admit.

 

[RampantAndroid]

Mmm.. Let me check this...
The improvement of IPC is:
From Nehalem from SB - 25%
From SB to Ivy Bridge - 15%
From Ib to Haswell - 10%
From Haswell to Broadwell - 10%
From Broadwell to Skylake - 15% (estimated and TBD)
So in total it could get up to 75% at minimun.

Not that at all. I've just forgot to add some more factors who nerf that percentages to the real performance. That's all.

Nehalem = 100
SB = 125
IB = 143.75
Haswell = 158.13
Broadwell = 173.94
Thus, Skylake = 200.03, or a 100% increase over Nehalem?

 

[Dresdenboy]

Why don't you use data provided by Intel? Skylake has to be added of course.

(removed, didn't check the history far enough while on mobile)

 

[Sweepr]

Skylake motherboards price
http://www.compsource.com/searches/search.asp?keyword=Skylake
http://us.directdial.com/vs/go/go.asp?SearchString=Skylake&Search=GO

Nice to see H170 and B150 options listed so early. Some people will pair Core i7 6700K with the cheapest motherboard + RAM kit + cooler available.

 

[Sweepr]

Core i5 6200U (Skylake-U, 2.4GHz base clock) Geekbench Score
Single-thread: 2928
Multi-thread: 6195

http://browser.primatelabs.com/geekbench3/3017270

To put things in perspective, the first leaked score is already faster than the fastest Core i5 5200U score listed.
Average Core i5 5200U (Broadwell-U, 2.2GHz base clock, 2.7GHz Turbo) Geekbench score is ~2500/5000.
Average Core i5 4200U (Haswell-U, 1.6GHz base clock, 2.6GHz Turbo) Geekbench score is ~2500/4800.

This will be the main processor of choice of many Windows 10 notebooks, including (probably) the Surface Pro 4.

 

[JDG1980]

It certainly isnt "complete BS". There is some variability depending on what benchmarks you look at, but I would say it is generally accurate. The generally accepted increase for haswell is 8 to 10 %. That said, I am getting tired of seeing it, I have to admit.

I'm skeptical of that chart because it seems to indicate that SB->Haswell is a bigger jump than Nehalem->SB, which is an assessment few enthusiasts would agree with. Sandy Bridge was seen as a major leap forward, while Haswell was largely greeted with "meh" on the desktop.

 

[dark zero]

Nehalem = 100
SB = 125
IB = 143.75
Haswell = 158.13
Broadwell = 173.94
Thus, Skylake = 200.03, or a 100% increase over Nehalem?

Well... If is in theory yeah. But in Real Life Apps it could go down to 80% increase.

 

[stag3]

so is the best bang for the buck still a 4790k or the 5820k?
i'm fine waiting a few weeks, my pc's generally last 5 years or so with
a video card upgrade tossed somewhere in that time

 

[dark zero]

so is the best bang for the buck still a 4790k or the 5820k?
i'm fine waiting a few weeks, my pc's generally last 5 years or so with
a video card upgrade tossed somewhere in that time

Wait for Skylake. Is the only option avaliable...

And I am starting to doubt if Intel is doing the lowest tier because Mobile is filling that spot

 

[tential]

so is the best bang for the buck still a 4790k or the 5820k?
i'm fine waiting a few weeks, my pc's generally last 5 years or so with
a video card upgrade tossed somewhere in that time

Yes the best bang for the buck is still those 2 processors.
But like you said, Skylake is RIGHT around the corner lol. You're not upgrading an intel CPU every couple of years you're holding onto it. You might as well get skylake. Older processors don't get discounted either so don't bother waiting for that lol. Just get skylake.

Short Answer: You're getting skylake.

 

[frozentundra123456]

I'm skeptical of that chart because it seems to indicate that SB->Haswell is a bigger jump than Nehalem->SB, which is an assessment few enthusiasts would agree with. Sandy Bridge was seen as a major leap forward, while Haswell was largely greeted with "meh" on the desktop.

Partially because there was a pretty big clockspeed increase as well, and SB also overclocked like crazy. So the absolute performance increase was very good, and huge if you overclocked. The chart shows performance per clock, which obviously is not the whole story, but nevertheless seems about right for what it is.

 

[JM Popaleetus]

As someone who literally just bought a Haswell-E setup a couple days ago. I am personally not opening anything until Skylake's numbers come out on August 5th.

That said, I highly doubt being an early adopter of Skylake is going to show any significant performance advantage over Haswell. Skylake's focus is mobile, expect to be amazed there...not on the desktop.

Cliffs: Wait the two weeks. But don't expect a Sandy Bridge-esque improvement.

 

[Zodiark1593]

Yes the best bang for the buck is still those 2 processors.
But like you said, Skylake is RIGHT around the corner lol. You're not upgrading an intel CPU every couple of years you're holding onto it. You might as well get skylake. Older processors don't get discounted either so don't bother waiting for that lol. Just get skylake.

Short Answer: You're getting skylake.

This, seriously.

At least if retailers discounted stock within weeks of a new release, the choice would be much easier, but unless you absolutely need to buy now (as in you'll lose money if you don't), you will potentially get a (almost) free 10%-20% performance boost out a few weeks wait. You otherwise don't benefit any whatsoever from buying so close. Tis the harsh reality of the tech world.

 

[frozentundra123456]

On the desktop at least, I would not count on 10 to 20 percent increase in absolute cpu performance. In performance per watt or igp performance, maybe, but in absolute performance we haven't seen that except for a very few early leaks, which haven't seemed to hold up.

Edit: plus if going to skylake you probably will have to buy new ram, and motherboards may be expensive initially. But haswell is not going away for quite a while either, so I suppose you might as well wait. Just dont count on any big price drops.

 

[JM Popaleetus]

This, seriously.

At least if retailers discounted stock within weeks of a new release, the choice would be much easier, but unless you absolutely need to buy now (as in you'll lose money if you don't), you will basically get a (almost) free 10%-20% performance boost out a few weeks wait. You otherwise don't benefit any whatsoever from buying so close. Tis the harsh reality of the tech world.

10-20% is extremely optimistic I think.

Definitely wait. But I expect a ~14% AT THE MOST improvement. Skylake lacks the free performance that Broadwell-C's eDRAM brought, and is a completely new platform. Meaning, expect to see some growing pains with Z170, especially with the lack of FIVR (this means motherboard choices will matter more again) and immaturity of DDR4.

Again. Skylake is going to shine in the mobile market. But I predict, for desktops it's going to be another 7-14% release.

 

[Zodiark1593]

Yeah, that's where my ninja edit comes into play, switching basically for potentially. Didn't expect such fast responses.

 

[RampantAndroid]

10-20% is extremely optimistic I think.

Definitely wait. But I expect a ~14% AT THE MOST improvement. Skylake lacks the free performance that Broadwell-C's eDRAM brought, and is a completely new platform. Meaning, expect to see some growing pains with Z170, especially with the lack of FIVR (this means motherboard choices will matter more again) and immaturity of DDR4.

Again. Skylake is going to shine in the mobile market. But I predict, for desktops it's going to be another 7-14% release.

7-14% over Broadwell? That means it's a sizable jump over the previous desktop chip consumers could get - which was Haswell. Right?

 

[ShintaiDK]

I'm skeptical of that chart because it seems to indicate that SB->Haswell is a bigger jump than Nehalem->SB, which is an assessment few enthusiasts would agree with. Sandy Bridge was seen as a major leap forward, while Haswell was largely greeted with "meh" on the desktop.

Remember frequency.

 

[Nothingness]

Remember frequency.

The graph deals with IPC, not performance.

FWIW I went from i920 to i4770K (both not OC) and saw some of my programs getting twice faster, but these are very specific (using GMP lib which uses new instructions).