Intel Skylake / Kaby Lake

[Intervenator]

Anyone have a good source to comprehensively and simply sum up everything we know about the differences between all these upcoming processor generations and chipsets? My head hurts... "Tick tock" is dead, "process, architecture, optimization" is dead, Moore's Law is dead, release dates are unpredictable, certain generations skip certain platforms, small performance improvements makes each generation less notable, and the similar sounding "Lake" nomenclature definitely does not help. Please tell me I am not alone?

 

[Bouowmx]

See:

• https://en.wikipedia.org/wiki/List_of_Intel_CPU_microarchitectures#Pentium_4_.2F_Core_Lines
  
• https://en.wikipedia.org/wiki/List_of_Intel_chipsets#100.2F200_Series_chipsets

S-series processor and chipset compatibility:

| Compatibility                | 100 series | 200 series | 300 series |
|------------------------------|------------|------------|------------|
| 6th generation               | TRUE       | TRUE       | TRUE (?)   |
| 7th generation               | TRUE       | TRUE       | TRUE (?)   |
| 8th generation (Coffee Lake) | FALSE      | FALSE      | TRUE       |

Spoiler: Kaby Lake-R Z370 vs Cannon Lake 300 series PCH

 

[jpiniero]

As mentioned, the 300 series in theory can have backward compatibility with Sky and Kaby, but it's apparently optional.

 

[jpiniero]

Anyone have a good source to comprehensively and simply sum up everything we know about the differences between all these upcoming processor generations and chipsets? My head hurts... "Tick tock" is dead, "process, architecture, optimization" is dead, Moore's Law is dead, release dates are unpredictable, certain generations skip certain platforms, small performance improvements makes each generation less notable, and the similar sounding "Lake" nomenclature definitely does not help. Please tell me I am not alone?

I would not be surprised if Cannonlake gets completely cancelled. The only real improvement with Coffee over Kaby is the additional cores although it might overclock an extra 100 mhz or two.

 

[CakeMonster]

I would not be surprised if Cannonlake gets completely cancelled. The only real improvement with Coffee over Kaby is the additional cores although it might overclock an extra 100 mhz or two.

Not counting on CNL and who knows when/if it is released. CFL is the safe bet for everyone who wants more cores NOW.

I'm going to be very disappointed if it can't match the single thread performance and clocks of 7700K though.

 

[Drazick]

Yes, we have seen the mobile chips' ability to boost higher, and for longer periods, within the same TDP envelope for 14+ over 14. This was a real-world indication that 14+ was better than 14 as far as overall efficiency. We have also seen the 14+ KL desktop chips have similar efficiency gains over the 14 SL chips. All of this is well known, so I am not sure why it keeps coming up. It was well reported at the time the 14+ mobile chips were released, and at the time the 14+ desktop chips were released.

There's no reason to think that 14++ is not more efficient again than 14+, regardless of marketing talk or fanboy talk.

Did you read what I wrote?
I didn't say anything about whether there is an improvement or not.
I just said that most probably what we see is a dicretization of points in time of process improvements that would have happened anyway and happened before (We didn't see them cause they were masked by stronger marketing features of the products).

So basically nothing special happened here besides that the same product is produced for longer time on the same process and hence experiences all improvements during the life time of any process.

The 14++ is just marketing discretization of this process.
The funny thing all others (Other manufacturing companies) also think this a good marketing idea and embraced it.

Had Intel original plans worked as expected this improvements would have happened without this marketing thing.
Just other products would benefited from it as the main product would move forward to the next process.

So we have an improvement of process as any other process before it and we have the same product using it for long hence marketing guys said, stick to it cool name and move on.

 

[maddie]

You do know that you still have X16 available for one slot, for the GPU if it becomes an absolute benefit.

At some point you have to stop making mountains out of motes of dust (this isn't even a molehill size issue).

Maybe you should look at the post I was replying to before becoming defensive.

 

[mikk]

So basically nothing special happened here besides that the same product is produced for longer time on the same process and hence experiences all improvements during the life time of any process.

There is a contradiction in your posting. In the first part you are in agreement but in the second you are denying what you said before. In the past Intel did not release a third and fourth generation of the same node, so Intel didn't need a distinction on a same node before. For 22nm Intel did release only Ivy Bridge and Haswell. That would be Broadwell and Skylake with 14nm, both are 14nm. No distinction beside that Intel further improved 14nm for Skylake especially the yields. With Kabylake (14nm+) they made further enhancements and for Coffeelake (14++) even more on top of that. This has never happened in the past, you are wrong in that case. Also for potential customers it's a welcomed differentiation that potential customers get an idea of the difference. I mean Kabylake was all about the improved 14nm with its improved clock speeds. These clock speed improvements were real despite people like you told from the beginning 14nm+ is marketing only. Same happens for 14nm++ now. I'm not sure if you are serious or just an Intel hater given that you are denying the improvements Intel made for 14nm+ and obviously for 14nm++.

 

[Drazick]

There is a contradiction in your posting. In the first part you are in agreement but in the second you are denying what you said before. In the past Intel did not release a third and fourth generation of the same node, so Intel didn't need a distinction on a same node before. For 22nm Intel did release only Ivy Bridge and Haswell. That would be Broadwell and Skylake with 14nm, both are 14nm. No distinction beside that Intel further improved 14nm for Skylake especially the yields. With Kabylake (14nm+) they made further enhancements and for Coffeelake (14++) even more on top of that. This has never happened in the past, you are wrong in that case. Also for potential customers it's a welcomed differentiation that potential customers get an idea of the difference. I mean Kabylake was all about the improved 14nm with its improved clock speeds. These clock speed improvements were real despite people like you told from the beginning 14nm+ is marketing only. Same happens for 14nm++ now. I'm not sure if you are serious or just an Intel hater given that you are denying the improvements Intel made for 14nm+ and obviously for 14nm++.

In the past the improvement of what we get as 14++ in the life cycle of a product was a new smaller process.
Now, instead of getting new and smaller (Considerably better) we get marginally improvement marketed as 14++.

I'm saying that only because Intel said 14++ is much better doesn't mean something out of the usual happened.
It is the same improvement cycle any process of Intel has went through.
We just never saw this cycle since the products of interest always moved to a completely new process at this time (Which is much better improvement than any xx+ / xx++ or even xx+++++++ process).

Anyhow, let's just Intel will bring much better innovations down the road.

 

[krumme]

Actually it would be interesting to know the difference from a late 14+ process to a new 14++
Not from priduct names or similar discrete and constructed tdp brackets but eg actual measured power for same freq or top freq for same power consumption.

 

[Excessi0n]

So basically nothing special happened here besides that the same product is produced for longer time on the same process and hence experiences all improvements during the life time of any process.

They aren't the same products, though. Kaby Lake (14nm+) involved new die designs and there's no indication that the quad-core Coffee Lake (14nm++) die won't be a new die as well, even though they're all the same architecture with the same features and the same core design. That points to a fundamentally different process rather than the usual slight enhancements to an existing process.

You can still buy newly-manufactured 6700Ks, you know. They won't hit the same frequencies at the same voltages as a 7700K.

 

[Arachnotronic]

They aren't the same products, though. Kaby Lake (14nm+) involved new die designs and there's no indication that the quad-core Coffee Lake (14nm++) die won't be a new die as well, even though they're all the same architecture with the same features and the same core design. That points to a fundamentally different process rather than the usual slight enhancements to an existing process.

You can still buy newly-manufactured 6700Ks, you know. They won't hit the same frequencies at the same voltages as a 7700K.

Yup. One of the improvements in 14nm+ was taller fins (improves drive current). No matter how long you produce 6700K chips, the fin height isn't going to change, and you are never going to get the increase in drive current/frequency potential that the enhanced process will give.

It's really amazing how much unsubstantiated FUD gets thrown around.

 

[jpiniero]

Yup. One of the improvements in 14nm+ was taller fins (improves drive current). No matter how long you produce 6700K chips, the fin height isn't going to change, and you are never going to get the increase in drive current/frequency potential that the enhanced process will give.

The max OC frequency potential really isn't that much different though.

 

[PeterScott]

Maybe you should look at the post I was replying to before becoming defensive.

I did. Why would you think otherwise? The post was pointing out that 8X slot allocation, has no real impact on GPU performance.

But I am not seeing what would stop your from using x16 GPU slot allocation.

It looks like you are combining two imaginary situations to have something to complain about. Some unknown future where having x16 PCIe slot allocation is critical for GPU performance, while some unknown factor is preventing you from allocating your board with x16 mode for your GPU slot.

 

[Arachnotronic]

The max OC frequency potential really isn't that much different though.

It's a 200-300MHz improvement.

 

[jpiniero]

It's a 200-300MHz improvement.

Which percentage wise isn't much.

 

[Arachnotronic]

Which percentage wise isn't much.

That's true. The impact from 14nm+ was most seen in the mobile/ULV SKUs.

 

[maddie]

I did. Why would you think otherwise? The post was pointing out that 8X slot allocation, has no real impact on GPU performance.

But I am not seeing what would stop your from using x16 GPU slot allocation.

It looks like you are combining two imaginary situations to have something to complain about. Some unknown future where having x16 PCIe slot allocation is critical for GPU performance, while some unknown factor is preventing you from allocating your board with x16 mode for your GPU slot.

The original poster was obviously referring to another PCIe device using some of the 16 available lanes leaving only X8 for the GPU. HE was replying to an earlier pessimistic post about only 16 lanes direct to CPU and saying that X8 is enough for now. You're jumping to defend without reading all, or are you saying that no one will use anything but a single GPU?

 

[PeterScott]

The original poster was obviously referring to another PCIe device using some of the 16 available lanes leaving only X8 for the GPU. HE was replying to an earlier pessimistic post about only 16 lanes direct to CPU and saying that X8 is enough for now. You're jumping to defend without reading all, or are you saying that no one will use anything but a single GPU?

You can still use 16x on the GPU and chipset lanes on other devices.

You are just manufacturing drama/issues, where none really exist.

 

[frozentundra123456]

In the past the improvement of what we get as 14++ in the life cycle of a product was a new smaller process.
Now, instead of getting new and smaller (Considerably better) we get marginally improvement marketed as 14++.

I'm saying that only because Intel said 14++ is much better doesn't mean something out of the usual happened.
It is the same improvement cycle any process of Intel has went through.
We just never saw this cycle since the products of interest always moved to a completely new process at this time (Which is much better improvement than any xx+ / xx++ or even xx+++++++ process).

Anyhow, let's just Intel will bring much better innovations down the road.

In the past the improvement of what we get as 14++ in the life cycle of a product was a new smaller process.
Now, instead of getting new and smaller (Considerably better) we get marginally improvement marketed as 14++.

I'm saying that only because Intel said 14++ is much better doesn't mean something out of the usual happened.
It is the same improvement cycle any process of Intel has went through.
We just never saw this cycle since the products of interest always moved to a completely new process at this time (Which is much better improvement than any xx+ / xx++ or even xx+++++++ process).

Anyhow, let's just Intel will bring much better innovations down the road.

Kaby Lake brought a 10 to 15 % improvement in mobile due to higher clocks and ability to maintain turbo longer. Recent ticks and tocks after sandy bridge brought very similar increases (sometimes probably smaller) in performance. When I recently bought a laptop for my wife, I made sure it was KL due to the higher clocks and overall better performance. I think I am also intelligent enough to make a decision based on actual performance rather than based on marketing.

 

[tamz_msc]

I thought it was pretty clear why Intel labels it's slowed-down node improvements the way it does - because it has been spending a longer time getting to the next "numbered" node with lesser relative size reduction compared to the competition. Thus they need to label existing improvements in order to market themselves in a better position than they are in reality.

https://www.semiwiki.com/forum/content/6895-standard-node-trend.html

From the above figure we can see that once again Intel takes the lead in 2014 with their 14nm process with a standard node value of 12.1. Samsung and then TSMC take the lead in 2017 with their 10nm processes having standard node values of 11.2 and 10.3 respectively. Intel takes the lead back in early 2017 with their 10nm process with a new standard node value of 8.3. In late 2017 TSMC takes the lead back with their 7nm with a standard node of 7.9 before GLOBALFOUNDRIES takes the lead in early 2018 with their 7nm process with a standard node value of 7.8.

Conclusion
By either the old or the new standard node values Intel has lost their multiyear density lead over the foundries. Based on the new more accurate standard node value the average node value for Intel's 10nm and the foundry 7nm processes is 8.05nm and all four companies are within a 0.5nm standard node value of each other.

 

[Dayman1225]

http://www.guru3d.com/news-story/european-core-i7-8700k-coffee-lake-prices-spotted-in-germany.html?

 

[Dayman1225]

https://twitter.com/AorusOfficial/status/906679000650493953/

5th confirmed it seems

 

[Drazick]

Kaby Lake brought a 10 to 15 % improvement in mobile due to higher clocks and ability to maintain turbo longer. Recent ticks and tocks after sandy bridge brought very similar increases (sometimes probably smaller) in performance. When I recently bought a laptop for my wife, I made sure it was KL due to the higher clocks and overall better performance. I think I am also intelligent enough to make a decision based on actual performance rather than based on marketing.

Again, It's like you are not reading what I'm writing.

Let's do a virtual experiment.
Let's say you're part of the team in Intel which tests CPU's for the binning process (There is a team like, pretty big actually, they write the software to automate the process).

Let's say you are working on a CPU manufactured by Intel in the xx Process (The name doesn't matter).

Now, you build histogram of the speed and power consumed by each CPU out of the Fab per each month.

You do it for 4 years.
What do you expect to see?

I will tell you what you will see.
Let's assume the distribution is approximately Gaussian.
Hence you define it by Variance and the Mean.

With time the Mean of the speed will rise, the mean of the power will go down (Per speed) and the variance will get smaller.
Till now it happened (I can promise you that) on each and every process Intel used, Each of Them!
You saw that once with Haswell but in small marketing scale.

Till now, Intel has never used that as a "feature" since they had better things to show.
But now, when they have little to show (Process wise) they use it as a marketing feature.

That's what I'm saying, the improvement is there.
It has always been there in some ways.

When you play cards you can only play with the ones you have.
Intel, currently, doesn't have new process to show, so they market the regular over time improvements as a feature.

Funny thing is some users here sell it even better than Intel as this is the best thing ever.
The next step is, we never want to move to 10nm, we want 14+++ and then 14++++ .
This is just silly.
Intel markets what it can at this time.
It wished it could market better things like a real new process with much better performance.

 

[Lodix]

Again, It's like you are not reading what I'm writing.

Let's do a virtual experiment.
Let's say you're part of the team in Intel which tests CPU's for the binning process (There is a team like, pretty big actually, they write the software to automate the process).

Let's say you are working on a CPU manufactured by Intel in the xx Process (The name doesn't matter).

Now, you build histogram of the speed and power consumed by each CPU out of the Fab per each month.

You do it for 4 years.
What do you expect to see?

I will tell you what you will see.
Let's assume the distribution is approximately Gaussian.
Hence you define it by Variance and the Mean.

With time the Mean of the speed will rise, the mean of the power will go down (Per speed) and the variance will get smaller.
Till now it happened (I can promise you that) on each and every process Intel used, Each of Them!
You saw that once with Haswell but in small marketing scale.

Till now, Intel has never used that as a "feature" since they had better things to show.
But now, when they have little to show (Process wise) they use it as a marketing feature.

That's what I'm saying, the improvement is there.
It has always been there in some ways.

When you play cards you can only play with the ones you have.
Intel, currently, doesn't have new process to show, so they market the regular over time improvements as a feature.

Funny thing is some users here sell it even better than Intel as this is the best thing ever.
The next step is, we never want to move to 10nm, we want 14+++ and then 14++++ .
This is just silly.
Intel markets what it can at this time.
It wished it could market better things like a real new process with much better performance.

Nobody denied that process and yields gets better with time. But the improvement is there and it is noticeable. So I don't see the problem in using a different name to clarify and differentiate this improvements. And Intel is tweaking the process to get that extra performance it is not just the maturing of it. You won't get the same product from 14nm++ with the initial 14nm.