Intel Skylake / Kaby Lake

[Phynaz]

Not problem like a fault but problem of manufacturing roadmaps/timetables and end-user not having smaller process earlier.

Why do end users need smaller process nodes?

Manufacturing process is an enabler, not a feature of an end user device.

 

[Phynaz]

I wonder if it was Intel or ASML who was overly optimistic.

Intel most likely. They have been stocking their labs with EUV equipment for years and I'm sure they expected to get them into production much sooner.

 

[LTC8K6]

7800x only has 28 lanes. Which is more than the 8700k will likely have but hardly way more.

You don't think 28 is way more than 16?

 

[Dayman1225]

Intel most likely. They have been stocking their labs with EUV equipment for years and I'm sure they expected to get them into production much sooner.

Doesn't matter now that Intel have denied that Cannonlake is delayed,


https://www.extremetech.com/computing/256127-rumors-imply-intel-pushed-10nm-cannon-lake-back-2018

"Update: Intel has reached out to ET with the following comment:" “We’ll be shipping our first 10-nanometer products near the end of the year beginning with a lower volume SKU followed by a volume ramp in the first half of 2018.

 

[AtenRa]

Well you're the one who said it was a problem, so now it's not problem? It okay to correct yourself, just making sure I'm reading your posts correctly.

So why do end users need smaller process nodes anyway?

Manufacturing process is an enabler, not a feature of an end user device.

Its a problem for the end user, every two years end user would get a new much better mobile chip. Now it will be three years since the first 14nm mobile chips were launch.
This is a problem because end user will be stuck to Intel's HD6xx series of iGPUs for more than two years when 10nm mobile chips will launch.
It is also a huge problem for Intel because it cannot compete with latest AMD mobile iGPUs.

 

[zinfamous]

Doesn't matter now that Intel have denied that Cannonlake is delayed,


https://www.extremetech.com/computing/256127-rumors-imply-intel-pushed-10nm-cannon-lake-back-2018

"Update: Intel has reached out to ET with the following comment:" “We’ll be shipping our first 10-nanometer products near the end of the year beginning with a lower volume SKU followed by a volume ramp in the first half of 2018.

first half of 2018 for real, consumer volume...so ~June 2018.

 

[Dayman1225]

first half of 2018 for real, consumer volume...so ~June 2018.

Perfect timing for Icelake, eh?

 

[elhefegaming]

Doesn't matter now that Intel have denied that Cannonlake is delayed,


https://www.extremetech.com/computing/256127-rumors-imply-intel-pushed-10nm-cannon-lake-back-2018

"Update: Intel has reached out to ET with the following comment:" “We’ll be shipping our first 10-nanometer products near the end of the year beginning with a lower volume SKU followed by a volume ramp in the first half of 2018.

So release the 10nm (that will likely need new mobos) just a couple of months after releasing a new 14nm CPU that requires new mobos? Really?

8xxx on Oct 5
9xxx by Dec 31? wtf? Why do they announce stuff they can fulfill

 

[Dayman1225]

So release the 10nm (that will likely need new mobos) just a couple of months after releasing a new 14nm CPU that requires new mobos? Really?

8xxx on Oct 5
9xxx by Dec 31? wtf? Why do they announce stuff they can fulfill

These first 10nm chips are U/Y Mobile chips only called Cannonlake, it will not affect Coffee lake in the slightest.

Now with Icelake in H2 2018~, its possible it will work on Z390 (CNL PCH), Z370 is dead end IMO.

 

[LTC8K6]

What would a "low volume SKU" be?

I don't think "mobile" when I hear that.

 

[jpiniero]

Digitimes seems to have rewritten the article, implying that the Cannonlake models that are coming will have the IGP fused off. So... yeah that would be low volume and would explain why the OEMs are upset.

 

[Dayman1225]

What would a "low volume SKU" be?

I don't think "mobile" when I hear that.

Probably means only few OEMs will implement them, or it's just low volume in general

 

[Dayman1225]

Digitimes seems to have rewritten the article, implying that the Cannonlake models that are coming will have the IGP fused off. So... yeah that would be low volume and would explain why the OEMs are upset.

If that's true, Cannonlake is definitely a pipe cleaner in Intel's eyes, but I prefer this than a full year delay.

I do wonder if this means Intel is struggling to shrink their IGPU Cores though.

EDIT: Didn't Ian Cutress just take a photo of a Cannonlake wafer and estimated 70mm^2~ 2+2 part, so IDK if I believe DigiTimes on this.

 

[jpiniero]

EDIT: Didn't Ian Cutress just take a photo of a Cannonlake wafer and estimated 70mm^2~ 2+2 part, so IDK if I believe DigiTimes on this.

The IGP would still be there, just fused off.

 

[Dayman1225]

The IGP would still be there, just fused off.

So I assume this is due to yields then? Wasn't one of the biggest things about Cannonlake its IGPU? 40EUs vs 24/8 (Forget exact)

 

[mikk]

Digitimes seems to have rewritten the article, implying that the Cannonlake models that are coming will have the IGP fused off. So... yeah that would be low volume and would explain why the OEMs are upset.

Interesting, that means the rumour was right about the culprit towards GPU for Cannonlake. For mobile OEMs that would be useless, a Dualcore without even its integrated GPU, there is no need for it. It's more a paper product in order to give Intel the right to say Cannonlake is/was on track. Intels answer to extremetech is just a standard answer from the marketing, they won't admit Cannonlake/10nm is a mess. I wonder if these GPU issues are caused by poor yields of 10nm or some unforeseen technical design bugs in Gen10.

 

[Phynaz]

Its a problem for the end user, every two years end user would get a new much better mobile chip. Now it will be three years since the first 14nm mobile chips were launch.
This is a problem because end user will be stuck to Intel's HD6xx series of iGPUs for more than two years when 10nm mobile chips will launch.
It is also a huge problem for Intel because it cannot compete with latest AMD mobile iGPUs.

Intel just released new, much better mobile chips. Please pay better attention.

Intel is competing just fine with AMD's mobile parts, as evidenced by market share. Your broken record of iGPU performance doom and gloom for Intel is really, really, old. During the entire time that AMD had a faster iGPU than Intel, AMD's market share dropped to essentially zero. That shows how much customers value AMD's iGPUs.

Next, you'll say this time is different because of Zen. But is it really? Where is mobile Zen? AMD hasn't said a word about it. My money is on the fact that Vega architecture is a very big problem for AMD mobile parts, and they are screwed by Vega's power consumption.

You still seem to think manufacturing process technology is a product feature that you can wave around to try to make Intel look inferior to AMD. It's not.

 

[Ajay]

Digitimes seems to have rewritten the article, implying that the Cannonlake models that are coming will have the IGP fused off. So... yeah that would be low volume and would explain why the OEMs are upset.

Wow, that's sad - stick a fork in it. Time for 10nm+.

 

[Ajay]

Interesting, that means the rumour was right about the culprit towards GPU for Cannonlake. For mobile OEMs that would be useless, a Dualcore without even its integrated GPU, there is no need for it. It's more a paper product in order to give Intel the right to say Cannonlake is/was on track. Intels answer to extremetech is just a standard answer from the marketing, they won't admit Cannonlake/10nm is a mess. I wonder if these GPU issues are caused by poor yields of 10nm or some unforeseen technical design bugs in Gen10.

Intel's iGPU have been pretty conservative - so I doubt it's a technical issue, though one never knows.

 

[Phynaz]

What would a "low volume SKU" be?

I don't think "mobile" when I hear that.

Maybe the 7.5w chips. Remember, the definition of low volume is relative. A low volume product for Intel could sell more than AMD's entire product line.

 

[Phynaz]

Digitimes seems to have rewritten the article, implying that the Cannonlake models that are coming will have the IGP fused off. So... yeah that would be low volume and would explain why the OEMs are upset.

Which makes the "article" no more credible than it was to begin with.

 

[DrMrLordX]

This is a problem because end user will be stuck to Intel's HD6xx series of iGPUs for more than two years when 10nm mobile chips will launch.
It is also a huge problem for Intel because it cannot compete with latest AMD mobile iGPUs.

Intel really hasn't been competing with AMD's mobile iGPUs. The last mobile product AMD launched was in 2015 (Carrizo). It's quite difficult to actually find Carrizo products, and the rest of the APU is weak enough that most consumers wouldn't miss them anyway (vs. Intel products).

You might be thinking of mobile Raven Ridge. That doesn't exist yet, at least not in the consumer space. If Intel is trying to bring mobile Coffeelake against Raven Ridge then yes, they may be in a little trouble . . . maybe. If OEMs actually adopt AMD solutions with sensible configurations. You know how THAT goes.

 

[Dayman1225]

Which makes the article no more credible than it was to begin with.

I agree, I think DT are wrong and are trying to save their bacon at this point...

 

[Ajay]

Which makes the "article" no more credible than it was to begin with.

Good point.

 

[mikk]

I agree, I think DT are wrong and are trying to save their bacon at this point...

I think they are correct because this is the only explanation how Intel is able to release a Cannonlake SKU in early 2018.