Does that mean that we can expect 3 products on a node from now on (Tick, Tock, Tock), since that's what they will do on 14nm? Or is 14nm the exception?
Looking at the absence of a Tick in 2014 on desktop (we're still on HW), I feel completely in the dark about what we can expect, and this only made things more confusing.
Any info out on what improvements the 2016 product will feature?
Given that we haven't got any solid info on the Zen architecture yet, I'm not sure how you can comment on it
Yes, it's clearly going to be an uphill battle, but at least it will be against 14nm as opposed to 10nm.
For those who have not read the conference call,Intel is uncertain about the 7nm cycle. There's lots of unanswered questions like EUV.
If EUV is available, Intel might go back to 2 years for 7nm if possible, that's what they strive for.
The multiple patterning really is the big culprit here (along with the usual increase in difficulty).
Sure, but we're talking about Intel here. Intel talked about 14nm like it was this one exception, that is not nice. Enjoy these quotes from November:
For those who have not read the conference call,Intel is uncertain about the 7nm cycle. There's lots of unanswered questions like EUV.
If EUV is available, Intel might go back to 2 years for 7nm if possible, that's what they strive for.
The multiple patterning really is the big culprit here (along with the usual increase in difficulty).
Well, in hindsight it's easy. But fact is that ALL information said there would be no delay, eg even in February: http://www.anandtech.com/show/8991/intel-at-isscc-2015-reaping-the-benefits-of-14nm-and-going-beyond-10nm.As far as I remember you were most vocal supporter of Intel's chips in 10nm in 2H 2016.
Looks like in best scenario we will see them in late/very late 2017 - how you feel now with your prediction?
We'll see. What I get from Brian's comments is that they might be able to get yields healthy earlier, but they want to have this nice, predictable 3 year cadence. Remember Intel's been working on 10nm for many years, even while 14 was delayed, so a 7nm 2 years after 10 seems plausible.They might be striving for it, but I don't think it's going to happen. IMHO the 2 year node cadence is dead.
We'll see. What I get from Brian's comments is that they might be able to get yields healthy earlier, but they want to have this nice, predictable 3 year cadence. Remember Intel's been working on 10nm for many years, even while 14 was delayed, so a 7nm 2 years after 10 seems plausible.
But no one knows at this point. EUV is still far from ready.
Or, considering 14nm got later and later with every prediction, 7nm could be even more delayed, nobody knows at this point. And what about after that??
But really, Intel has to implement some sort of alternative strategy other than node shrinks to drive performance and new products. Maybe the "second tock" on 14nm is the start of such a new strategy, or maybe it is just some minor tweak to cover up the lack of progress. We will just have to wait and see.
It won't be getting any better, ever.
Adjust your expectations to the new normal, be you customer, employee, or shareholder.
If AMD dies, forget that. No company would dare to defy Intel. And obviously the cost will skyrocket.Well, personally, and I have been saying it in the forums, I think Intel needs to bring out a mainstream hex core and make hyperthreading available in lower lines, like the pentium.
Something like i7=hex core+HT, i5=quad core+Ht, i3=quad, no HT, pentium= dual core+Ht, celeron= dual core no HT. Basically increase the threads/cores at every level except the absolute bottom end celeron. They sure seem to have hit the wall as far as per core performance increases go. At least such a line-up would give higher end users/enthusiasts some reason to upgrade. And at 14nm they could surely afford the die space.
If AMD dies, forget that. No company would dare to defy Intel. And obviously the cost will skyrocket.
Imagine the Core Celeron at Core i3 cost and forcing us to buy the Craptom Celeron.
Oh it does.In an imaginary world perhaps. But not this one.
AMD dont have any influence on Intels prices.
RIP Tick-Tock. Brian Krzanich is treating dangerous waters here...
I hope he follows Intel's legacy of investing more instead of less in difficult times. I wonder what will happen to Intel's momentous process lead. My first guess is it should stay flat.
Could it be Intel's waiting for EUV?
But yeah, we had some bad news in the last couple of years. The cost per foundry transistor first remained flat and 28nm was a very long nodes and still is for GPUs. We got the FinFET transition on the 20nm BEOL, 14 got delayed by 9 months, and now, after all the optimism from the investor meeting, Intel declares Moore's Law death.
Happy 50th anniversary, Moore's Law.
I don't think it's *that* bad, Witeken. Intel already had its share of problematic nodes in the pasts, like 90nm and they recovered from that.
I think we can expect 3 products from now on.
If Zen makes it on time and AMD doesn't bungle it they might have an opportunity. Unfortunately, AMD nearly guarantees over promised and under delivered.
From what I can tell, AMD's Zen projections are rather modest. An IPC somewhere between SB and IB seems fairly easy. I'm guessing they are still stinging from the great Bulldozer belly flop.
If Zen makes it on time and AMD doesn't bungle it they might have an opportunity. Unfortunately, AMD nearly guarantees over promised and under delivered.
Sadly AMD is dead and nVIDIA will suffer the same destiny at max on 2 Intel generations more.
The era of the x86 computing is about to end and seems that ARM will be the future
Oh it does.
While AMD is alive, Intel wouldn't dare to increase the Celeron and Pentium prices due the Athlon and the A8. If AMD dies, Intel could easily increase the prices of Core products and promote more aggresively the Atom products (who are in critical loses) since common people won't realize the difference of the Atom and the Core.
Well, personally, and I have been saying it in the forums, I think Intel needs to bring out a mainstream hex core and make hyperthreading available in lower lines, like the pentium.
Something like i7=hex core+HT, i5=quad core+Ht, i3=quad, no HT, pentium= dual core+Ht, celeron= dual core no HT. Basically increase the threads/cores at every level except the absolute bottom end celeron. They sure seem to have hit the wall as far as per core performance increases go. At least such a line-up would give higher end users/enthusiasts some reason to upgrade. And at 14nm they could surely afford the die space.