[Canard PC Hardware] Intel prepares Ryzen's response behind the scenes

[Phynaz]

Intel is still pursuing full node shrinks, unlike its competitors at GlobalFoundries, Samsung, and TSMC. These firms won’t offer equivalent gate lengths or feature sizes to Intel until they deploy their 10nm nodes and aren’t expected to match Intel’s 10nm until they hit 7nm.

https://www.extremetech.com/computi...l-still-use-14nm-will-arrive-earlier-expected

 

[raghu78]

Phynax I will wait to see actual 10nm Cannonlake launch. As for real Intel 10nm ramp it will be known based on product launch and availability for the first few months.

 

[krumme]

I kind of get that at Intel you design for the process. Makes sense and is all night and dandy but when your neighbour is process designing for apple iphone and samsung galaxy it kind of shows how old school and reactionary it is. 5 years ago it was super. Today its simply a no go. The world have changed.
Show me where process for the major part of the server market is so radical different from low power phones that it can pay for 5nm dev cost?
Look at zen clocks and efficiency - the 180w tdp 32c will be a slamdunk. If you can share that kind of cost with iphone it looks like an uphill battle to me.

 

[Andrei.]

FYI, Intel is manufacturing 10nm CPUs now.

Hardly.

 

[Phynaz]

Hardly.

Next time you should Google it before you deny somebody's claims. Keeps you from looking silly.

 

[tential]

Will intel have something price competitive when Zen 2 comes around? I mean they may not make a major move now, but not doing anything when Zen 2 comes out would seem odd right? Unless intel intends to move the hexacore into mainstream.
Still though, that octocore not being mainstream makes it Zen 2 and beyond look strong.

 

[Arachnotronic]

Will intel have something price competitive when Zen 2 comes around? I mean they may not make a major move now, but not doing anything when Zen 2 comes out would seem odd right? Unless intel intends to move the hexacore into mainstream.
Still though, that octocore not being mainstream makes it Zen 2 and beyond look strong.

Mainstream hex comes in early 2018 with Coffee Lake.

 

[Arachnotronic]

Phynax I will wait to see actual 10nm Cannonlake launch. As for real Intel 10nm ramp it will be known based on product launch and availability for the first few months.

10nm CNL will launch at the end of 2017 with serious volumes in 1H 2018. If I would be so bold a to make a prediction, I think Cannon Lake is gonna get "Broadwell'd" and will be replaced by Ice Lake-Y/U in the second half of 2018 on 10nm+, then if the normal pattern follows, we get desktop/high perf notebook goodies in early 2019. Then I think we'll get "the bomb" of Ice Lake-X around Computex/Gamescom 2019 on 10nm++ -- I'm predicting a mess o' cores and some really nice IPC/clock frequency to boot.

Then in mid/late 2020, Intel should HEDT buyers some 7nm Sapphire Rapid goodness

 

[waffleironhead]

Next time you should Google it before you deny somebody's claims. Keeps you from looking silly.

Do you have links to the quantity of 10nm being manufactured?

 

[Sven_eng]

Do you have links to the quantity of 10nm being manufactured?

No, he doesn't.

 

[CHADBOGA]

10nm CNL will launch at the end of 2017 with serious volumes in 1H 2018. If I would be so bold a to make a prediction, I think Cannon Lake is gonna get "Broadwell'd" and will be replaced by Ice Lake-Y/U in the second half of 2018 on 10nm+, then if the normal pattern follows, we get desktop/high perf notebook goodies in early 2019. Then I think we'll get "the bomb" of Ice Lake-X around Computex/Gamescom 2019 on 10nm++ -- I'm predicting a mess o' cores and some really nice IPC/clock frequency to boot.

So approx. 2 years before Intel offers anything more compelling than their current range for desktop users?

 

[krumme]

10nm CNL will launch at the end of 2017 with serious volumes in 1H 2018. If I would be so bold a to make a prediction, I think Cannon Lake is gonna get "Broadwell'd" and will be replaced by Ice Lake-Y/U in the second half of 2018 on 10nm+, then if the normal pattern follows, we get desktop/high perf notebook goodies in early 2019. Then I think we'll get "the bomb" of Ice Lake-X around Computex/Gamescom 2019 on 10nm++ -- I'm predicting a mess o' cores and some really nice IPC/clock frequency to boot.

Then in mid/late 2020, Intel should HEDT buyers some 7nm Sapphire Rapid goodness

What info points to ice lake having really nice ipc and freq - to make it "a bomb"? - and what is the purpose?

Do we get the pleassure of a 35 stage pipeline and a 6 wide design in a 300w tdp for a 4c consumer cpu in 10nm - while the st perf priests is chanting "st perf is what matters for consumers" repeatedly? (Its soooo good for gaming and it got avx512 !)

I dont think the future is fatter cpu than we have now. Its futile to go for radically more ipc and freq besides the usual aprox 7% ipc uplift.

 

[lopri]

We have this gem, too. I would not have trusted a PR graph wholeheartedly but even so I was caught off-guard when someone pointed me to the precision required to read this mundane-looking graph. Apparently you should give nuance to every word, especially how the wording has changed in comparison to past graphs of the same kind. And when it comes to the graph, you had better look at it in full screen mode with your eyes wide open, because every millimeter count. The variables as well as the distances between coordinates on X-axis and Y-axis are carefully chosen for maximum visual impact, and every shape and size of dots or circles has purposes. Where a line begins or ends have often huge implications, but if you do not pay attention you might only remember the angles or overlaps. Talking about overlaps, any misalignment of coordinates that appears a simple graphical error is never an error. That is what they want you to think it is, not what it really is. Arrows should be carefully inspected to see where exactly the tips are, and whether they are perpendicular to an axis or almost perpendicular.

Basically you have to read this slide as if it were a legal document, keeping in mind that its purpose is to deceive while leaving a plausible deniability for the future. That, in my opinion, not how much Intel is really ahead, is what is really telling. (and yay, Intel is will have enjoyed ~3 years lead ahead when Samsung launches Galaxy S8 next month! Totally amazing!)

 

[Crumpet]

We won't see cannonlake 10nm on the desktop until 2018...

 

[Lodix]

We have this gem, too. I would not have trusted a PR graph wholeheartedly but even so I was caught off-guard when someone pointed me to the precision required to read this mundane-looking graph. Apparently you should give nuance to every word, especially how the wording has changed in comparison to past graphs of the same kind. And when it comes to the graph, you had better look at it in full screen mode with your eyes wide open, because every millimeter count. The variables as well as the distances between coordinates on X-axis and Y-axis are carefully chosen for maximum visual impact, and every shape and size of dots or circles has purposes. Where a line begins or ends have often huge implications, but if you do not pay attention you might only remember the angles or overlaps. Talking about overlaps, any misalignment of coordinates that appears a simple graphical error is never an error. That is what they want you to think it is, not what it really is. Arrows should be carefully inspected to see where exactly the tips are, and whether they are perpendicular to an axis or almost perpendicular.

Basically you have to read this slide as if it were a legal document, keeping in mind that its purpose is to deceive while leaving a plausible deniability for the future. That, in my opinion, not how much Intel is really ahead, is what is really telling. (and yay, Intel is will have enjoyed ~3 years lead ahead when Samsung launches Galaxy S8 next month! Totally amazing!)

https://www.semiwiki.com/forum/f2/intels-fake-news-8953.html

 

[Nothingness]

https://www.semiwiki.com/forum/f2/intels-fake-news-8953.html

A thing that one can be sure of is that the data on the slide can be proven correct in the proper context given that it was presented at an investor meeting

 

[Phynaz]

Do you have links to the quantity of 10nm being manufactured?

Of course not. It's not relevant to a true/false argument anyway.

 

[unseenmorbidity]

This isn't a real response, it's pure marketing. They are trying to draw the spotlight away from amd with the same old overpriced crap.

A real response would be slashing their prices, but they aren't going to do that. Instead, they will do stuff like this and basically buy their way out of this hole, just as they have in the past.

Who in their right mind would want an i5 over a ~4790k in 2017?! Scratch that, the mb+cpu will likely be far more expensive than that... It would be like an i5 vs 5960x.

 

[Magic Hate Ball]

Mainstream hex comes in early 2018 with Coffee Lake.

*Mainstream hex comes March 2nd.

All joking aside, Intel will do fine.

Their stock may take some hits though if AMD gains a server foothold with Naples and a decent laptop presence with Raven Ridge however.

 

[Arachnotronic]

We won't see cannonlake 10nm on the desktop until 2018...

IMHO you'll never see Cannon Lake 10nm on the desktop.

After Coffee Lake should come Ice Lake for desktop.

 

[Phynaz]

A real response would be slashing their prices, but they aren't going to do that.

Please tell me why Intel (or any company) would want to compete on price.

How has competing on price worked out for AMD the last decade or so?

 

[Phynaz]

IMHO you'll never see Cannon Lake 10nm on the desktop.

After Coffee Lake should come Ice Lake for desktop.

I believe this to be true. As of now 10nm Cannon Lake is mobile only.

 

[realibrad]

Please tell me why Intel (or any company) would want to compete on price.

How has competing on price worked out for AMD the last decade or so?

Basic economic principle. If your product costs less than a competitor, then you can reduce your price while having equal profit margins. Intel could run a simple cost benefit analysis and see that the total lost sales would reduce revenue than reducing price and accepting lower margins.

So, if Intel were keep the same prices they could lose 50% in a segment resulting in X amount less. If Intel were to reduce their price by 20%, it could be that their net reduction would be less than the previous X amount. In the short term, it could be very useful to do a price reduction.

 

[waffleironhead]

Of course not. It's not relevant to a true/false argument anyway.

I must be reading "hardly" to mean something different then.

 

[waffleironhead]

No, he doesn't.

Then how is he refuting "hardly", which in english means "just barely". Its not a true/false statement.