AMD Migrates 20nm Chips In Development to FinFET – Pays $33 Million Charge

[desprado]

SUNNYVALE, CA — (Marketwired) — 07/06/15 — AMD (NASDAQ: AMD) today announced that revenue for the second quarter ended June 27, 2015 is expected to be lower than previously guided. The company now expects second quarter revenue to decrease approximately 8 percent sequentially, compared to the previous guidance of down 3 percent, plus or minus 3 percent. The sequential decrease is primarily due to weaker than expected consumer PC demand impacting the company’s Original Equipment Manufacturer (OEM) APU sales. The company expects second quarter channel sales and channel inventory reduction efforts to be in-line with the company’s plans.
The company anticipates non-GAAP gross margin to be approximately 28 percent, compared to the previous non-GAAP guidance of approximately 32 percent primarily due to a higher mix of Enterprise, Embedded and Semi-Custom segment sales and lower than anticipated Computing and Graphics segment APU unit volumes due to weaker than expected OEM PC product demand. Additionally the company anticipates GAAP gross margin to be further impacted by a one-time charge of approximately $33 million associated with a technology node transition from 20 nanometer (nm) to FinFET. The company started several product designs in 20nm that will instead transition to the leading-edge FinFET node.
Cash and cash equivalents at the end of the second quarter are expected to be approximately $830 million, in line with expectations.
AMD will report second quarter 2015 results after market close on Thursday, July 16, 2015. AMD will hold a conference call for the financial community at 2:30 p.m. PT (5:30 p.m. ET) that day to discuss second quarter financial results and to provide information regarding expected third quarter results. AMD will provide a real-time audio broadcast of the teleconference on the Investor Relations page at http://www.amd.com. The webcast will be available for 10 days after the conference call.
​

http://wccftech.com/amd-transitions-20nm-designs-finfet/
Although after this announcement AMD stock hit by a lot.
Stock price: AMD (NASDAQ) $2.02 -0.06 (-3.11%)

 

[ShintaiDK]

It was hit with a 15% drop due to the Q2 warning where this info was part of.

 

[desprado]

It was hit with a 15% drop due to the Q2 warning where this info was part of.

No wonder.

 

[Fjodor2001]

$33 million is actually not much in the grand scheme of things.

It was probably a wise move to convert the designs from 20 to 14 nm, since 20 nm was available quite late and 14 nm quite early compared to the competition.

 

[Rvenger]

They are taking the expense up front to try to be competitive. A bold move I must say, but if they miss the mark this last time, they are probably screwed.

Are they making sure Zen gets on 14nm I am guessing? They must be confident in the performance if so.

 

[Stuka87]

I am sure they already have had chips fabbed on 14nm in order for them to feel good about the switch.

But I feel its an important move. If they want to be competitive in the mobile space, they pretty much have to be on 14nm. It will be the first time in what, 6-7 years, since Intel and AMD have both been on the same process node?

 

[ShintaiDK]

I am sure they already have had chips fabbed on 14nm in order for them to feel good about the switch.

But I feel its an important move. If they want to be competitive in the mobile space, they pretty much have to be on 14nm. It will be the first time in what, 6-7 years, since Intel and AMD have both been on the same process node?

Its not the same node tho. PR took over and gave "inflation".

 

[Fjodor2001]

Its not the same node tho. PR took over and gave "inflation".

Still very close to Intel's 14 nm process tech, much closer than AMD has been for a long time.

 

[frozentundra123456]

Its not the same node tho. PR took over and gave "inflation".

Not to mention, by the time Zen appears in mobile (sometime in 2017 at best, assuming no delays) Intel will probably be close to or starting to transition to 10nm for mobile.

 

[ShintaiDK]

Still very close to Intel's 14 nm process tech, much closer than AMD has been for a long time.

Still comparing apples and oranges after multipatterning. Its not going to be fun if one company got a 50% gate utilization and the other 70% for example. Not to mention the obvious size differences in the process nodes or parametric yields.

 

[ShintaiDK]

Not to mention, by the time Zen appears in mobile (sometime in 2017 at best, assuming no delays) Intel will probably be close to or starting to transition to 10nm for mobile.

Even more important will be what 14nm fabless costs are. They will most likely be at a huge price disadvantage. Cost will most likely be similar to 28nm on a transistor basis.

 

[Sweepr]

Assuming the leak is correct and Summit Ridge arrives in October 2016 (Q4-2016) it will compete with Intel's 3rd generation 14nm processors (Kabylake) in desktops. Also in mobile where new nodes matter the most there's no 14nm FinFET from AMD till 2017 (Raven Ridge), by the time they launch (or shortly before/after) 10nm Cannonlake will be on the market.

 

[Fjodor2001]

Even more important will be what 14nm fabless costs are. They will most likely be at a huge price disadvantage. Cost will most likely be similar to 28nm on a transistor basis.

How is that different from all other previous nodes? They always demand a price premium when new.

Also, note that Zen in expected to arrive in late 2016. By then 14 nm prices have likely cooled down quite a bit compared to now.

 

[ShintaiDK]

How is that different from all other previous nodes? They always demand a price premium when new.

Also, note that Zen in expected to arrive in late 2016. By then 14 nm prices have likely cooled down quite a bit compared to now.

The difference is previous nodes have always been cheaper.

We are not just talking about new, we are talking about cost issues that only EUV can possible solve for the fabless.

Remember these 14nm costs are based on the projected cost in Q4 2016.

 

[JDG1980]

Assuming the leak is correct and Summit Ridge arrives in October 2016 (Q4-2016) it will compete with Intel's 3rd generation 14nm processors (Kabylake) in desktops. Also in mobile where new nodes matter the most there's no 14nm FinFET from AMD till 2017 (Raven Ridge), by the time they launch (or shortly before/after) 10nm Cannonlake will be on the market.

Summit Ridge is a server/HEDT product (8 cores/16 threads, no iGPU). Therefore it will likely be positioned to compete with Broadwell-E, not Kabylake. (Though it's possible it could wind up competing with Skylake-E if this story is true.)

As for the 2017 APUs... well, this quarter's results show how poor a strategy it is to compete on price alone in the low to mid-range OEM market when the products are clearly inferior to the competition's best. If AMD wants to stand out and not just be known as the "cheap" alternative (and this is one of the stated goals of the 2016 Financial Analyst Day), then they need to do something unique that Intel can't match, at least not immediately. This would mean making Raven Ridge a strong APU with Tonga-level graphics performance, four Zen cores, and shared HBM2 (8GB or so) on an interposer. It would be more expensive to manufacture, but it would be competing not against lone Intel CPUs but against Intel CPU /Nvidia mobile GPU combos (in laptops, AIOs, Steamboxes, and so forth), where it could provide better overall performance at a lower price and still make a profit.

 

[Fjodor2001]

The difference is previous nodes have always been cheaper.

We are not just talking about new, we are talking about cost issues that only EUV can possible solve for the fabless.

Not sure what you are talking about? No one will be using EUV for 14 nm as far as I'm aware.

Remember these 14nm costs are based on the projected cost in Q4 2016.

Sure, but that goes for everyone in the Industry, not just AMD.

Also, looking at that slide, 14 nm does not seem to be much more expensive than 28 nm in 2016Q4 (assuming that is the time that the slide estimate actually is for - not sure where that can be seen?).

 

[Fjodor2001]

Summit Ridge is a server/HEDT product (8 cores/16 threads, no iGPU). Therefore it will likely be positioned to compete with Broadwell-E, not Kabylake. (Though it's possible it could wind up competing with Skylake-E if this story is true.)

As for the 2017 APUs... well, this quarter's results show how poor a strategy it is to compete on price alone in the low to mid-range OEM market when the products are clearly inferior to the competition's best. If AMD wants to stand out and not just be known as the "cheap" alternative (and this is one of the stated goals of the 2016 Financial Analyst Day), then they need to do something unique that Intel can't match, at least not immediately. This would mean making Raven Ridge a strong APU with Tonga-level graphics performance, four Zen cores, and shared HBM2 (8GB or so) on an interposer. It would be more expensive to manufacture, but it would be competing not against lone Intel CPUs but against Intel CPU /Nvidia mobile GPU combos (in laptops, AIOs, Steamboxes, and so forth), where it could provide better overall performance at a lower price and still make a profit.

+1

"Tonga-level graphics performance, four Zen cores, and shared HBM2 (8GB or so) on an interposer"

That can actually be a very nice product. Also, Intel has no corresponding product in that segment so the competition is less fierce. And since Intel have worse GPU cores, it's not likely whey will be able to enter that market segment later either in any meaningful way.

 

[AnitaPeterson]

Ballsy move for AMD. Fingers crossed the gamble pays off...

 

[ShintaiDK]

Not sure what you are talking about? No one will be using EUV for 14 nm as far as I'm aware.


Sure, but that goes for everyone in the Industry, not just AMD.

Also, looking at that slide, 14 nm does not seem to be much more expensive than 28 nm in 2016Q4 (assuming that is the time that the slide estimate actually is for - not sure where that can be seen?).

Currently only Intel gets cheaper transistor cost.

Remember the chart is per transistor cost. So if you take a 28nm chip and shrink it to 14nm without changes. Then it will cost about the same.

When you try to actually make a better chip, cost goes up. This is also why this is an outright disaster for GPUs.

 

[LTC8K6]

+1

"Tonga-level graphics performance, four Zen cores, and shared HBM2 (8GB or so) on an interposer"

That can actually be a very nice product. Also, Intel has no corresponding product in that segment so the competition is less fierce. And since Intel have worse GPU cores, it's not likely whey will be able to enter that market segment later either in any meaningful way.

How big of a socket would that require? Sounds like it would be a rather large package.

 

[ShintaiDK]

Summit Ridge is a server/HEDT product (8 cores/16 threads, no iGPU). Therefore it will likely be positioned to compete with Broadwell-E, not Kabylake. (Though it's possible it could wind up competing with Skylake-E if this story is true.)

As for the 2017 APUs... well, this quarter's results show how poor a strategy it is to compete on price alone in the low to mid-range OEM market when the products are clearly inferior to the competition's best. If AMD wants to stand out and not just be known as the "cheap" alternative (and this is one of the stated goals of the 2016 Financial Analyst Day), then they need to do something unique that Intel can't match, at least not immediately. This would mean making Raven Ridge a strong APU with Tonga-level graphics performance, four Zen cores, and shared HBM2 (8GB or so) on an interposer. It would be more expensive to manufacture, but it would be competing not against lone Intel CPUs but against Intel CPU /Nvidia mobile GPU combos (in laptops, AIOs, Steamboxes, and so forth), where it could provide better overall performance at a lower price and still make a profit.

Better be more than 8GB in 2017. And did you consider the actual cost and target audience of such a chip? Not to mention what marketshare AMD may have left by the time.

APUs have been a massive flop for AMD in the PC segment. And sales is getting close to 0. And we are reaching a point where the ultimate question may be if Zen launches at all. Q2 will show a semicustom company with a PC division that is tagged on.

 

[erunion]

Ballsy move for AMD. Fingers crossed the gamble pays off...

GF 20nm was cancelled and replaced with Samsung 14nm. They had no choice but to port those designs.

 

[Sweepr]

Summit Ridge is a server/HEDT product (8 cores/16 threads, no iGPU). Therefore it will likely be positioned to compete with Broadwell-E, not Kabylake. (Though it's possible it could wind up competing with Skylake-E if this story is true.

It's a desktop chip. Did they mention anything about servers yet?

http://benchlife.info/wp-content/uploads/2015/06/amd-client-micro-pga-socket-roadmap.jpg

As for the 2017 APUs... well, this quarter's results show how poor a strategy it is to compete on price alone in the low to mid-range OEM market when the products are clearly inferior to the competition's best. If AMD wants to stand out and not just be known as the "cheap" alternative (and this is one of the stated goals of the 2016 Financial Analyst Day), then they need to do something unique that Intel can't match, at least not immediately. This would mean making Raven Ridge a strong APU with Tonga-level graphics performance, four Zen cores, and shared HBM2 (8GB or so) on an interposer. It would be more expensive to manufacture, but it would be competing not against lone Intel CPUs but against Intel CPU /Nvidia mobile GPU combos (in laptops, AIOs, Steamboxes, and so forth), where it could provide better overall performance at a lower price and still make a profit.

Quite a long road till 2017, just like AMD could deliver something interesting with their Zen x86 cores Intel could also manage to greatly improve their graphics performance per watt/mm² (there will be Gen 9, Gen 10 and possibly Gen 11 by then). Such product would likely be expensive to manufacture, they would have to price it cheaper than competing CPU+dGPU solutions in desktops and deliver notable efficiency/integration gains in laptops compared to the 2-chip approach (gaming laptops niche too, otherwise you wouldn't need a huge iGPU like that). Albeit interesting from a technical point of view I'm not sure how such product would fit with the market.

 

[frozentundra123456]

If AMD can incorporate HBM into their APUs, I think they for sure could make a gaming type apu with much better performance than Intel. I have given up on igpu improvements from Intel TBH, except for just throwing more EUs at the problem. First it was wait for Gen 8, now it is Gen 9, and on and on. The problem for AMD is exactly where that theoretical APU will fit and what the cost to manufacture and sell it will be.

I still am not convinced an APU is going to be anything more than a low/mid range solution for desktops. They have to be able to get good performance into a mobile package for it to make sense.

Another problem (for both AMD and Intel) is that 14nm dgpus should be out by then with HBM as well, giving much better performance (and performance/watt) across the board.

 

[Fjodor2001]

Currently only Intel gets cheaper transistor cost.

Why would that be the case? If the cost per transistor for 14 nm is higher when buying from the foundries, it means their investments for 14 nm process tech has also been higher. Intel has had to absorb that process tech cost too for 14 nm.

Also, as could be seen in your graph, the difference in cost between 28 nm vs 14 nm was not that big anyway.

Remember the chart is per transistor cost. So if you take a 28nm chip and shrink it to 14nm without changes. Then it will cost about the same.

When you try to actually make a better chip, cost goes up. This is also why this is an outright disaster for GPUs.

Of course. But that's the case for all manufacturers of discrete GFX cards (effectively AMD and nVidia), not just AMD. Also, the cost in your graph is just at one point in time. The cost per transistor for new nodes goes town over time, just as it does for all new nodes. That's the whole deal with Moore's law. Relax.

Finally, you never answered my question regarding your statement that all companies buying from foundries at 14 nm would have to use EUV due to unspecified "cost issues". What was that all about?