AMD Q313 Results

[krumme]

I don't think it's high expectations. They have a 450MM OPEX bill to foot, the business that pays it is dying and the new business only pays 11% of it. Another 15% drop in their CPU business is enough to wipe out the console gains and put the company back into negative cash flow territory, and we know that it isn't a too distant scenario for 2014.

Yes and thats why i wrote in the same post there had to be more reductions if Kaveri will not deliver big time.

What you are writing is just proof of the new strategy as the only viable alternative. What do you have them to do else?

 

[mrmt]

What you are writing is just proof of the new strategy as the only viable alternative. What do you have them to do else?

I would have rebuilt the product line around Jaguar products, and not around a mix of cheap x86 and vanilla ARM chips as they are doing now. But they either didn't like this course of action or saw things that I do not see right now, so this question is mainly rethoric now. AMD is fully committed to their current strategy, and that's it.

But look at some deleterious effects here:

Now turning to the business segments. Computing Solutions segment revenue was $790 million, down 6% sequentially, due to lower notebook and chipset unit shipments, partially offset by higher desktop unit shipments. Computing Solutions operating income was $22 million, up from $2 million in the second quarter despite the sequential 6% decline in revenue, driven by lower operating expenses.

(...)

We have had, just like we have to diversify our portfolio across high-growth segment, we need to diversify this core business. We need to move stronger into desktop and as we talked about a year ago, we worked on the inventory on the desktop segment and we built and repaired that and we have seen two quarters of consistent revenue growth in that segment and we believe that we have the right product stack to continue to make progress and that’s part of our business as well.

The desktop market is shrinking, everything is moving up to mobile. Desktop is *not* high growth. Even the x86 market is going mobile and AMD is losing the healthiest PC market out there, and focusing on the shrinking, lower ASP desktop market. This means that AMD is being pushed lower and lower to the bottom of the value chain in the consumer market, and this does not end well. It is a repeat of what happened in the server market, where years of stagnant AMD product lines against Intel solid improvements ended with AMD server line up being completely outclassed by Intel offers. This is AMD retreating to the last trench, a move akin to them offering 4P servers cheaper than Intel 2P westmere servers in 2010. once year later, once SNB-EP arrived AMD kissed the server market goodbye.

FWIW I don't see how AMD will be able to replace OPEX from this dying business, and I don't see them investing enough to keep this business healthy in the future. I see AMD stable for the next 2-3 quarters, but I can't really be sure about Q314 and beyond, especially when we'll have the introduction of Intel new node, which usually wreaks havoc in AMD income statement.

 

[SiliconWars]

Yeah we remember last year what you said about AMD being bankrupt in Q3 this year mrmt, about how they couldn't pay the wafers, how there was no money in consoles etc etc.

With your record, I'm really glad to see you predicting more doom for them.

 

[mrmt]

Yeah we remember last year what you said about AMD being bankrupt in Q3 this year mrmt, about how they couldn't pay the wafers, how there was no money in consoles etc etc.

If their CPU market share kept contracting at Q312/Q412 pace, yes, they would be bankrupt by now. For example, AMD would be earning just 550MM in CPU sales and losing some 150-200MM in cash per quarter if they did, as you can see they do not have the balance sheet to support this for 2-3 quarters without selling assets or filling for protection as soon as possible. The fact that the market priced AMD stock at liquidation tells me that a few people agreed with this evaluation at the time. And I don't call 50MM/quarter real money for a company of AMD's size. It's just better than nothing, but it's not what will save the company.

I can look back and explain why I reached those conclusions. I can even look back and tell you which premises made some of my conclusions wrong, and I wouldn't mind doing this with you if you weren't the way you are. Unfortunately I, or anyone else for that matter, cannot expect the same from you, as you refuse to tell where's the dark hole you pulled your numbers on AMD gaining market share in Q3, or AMD console margins, or about Intel atrocious Q3 numbers and what not.

Btw, did you start reading the book I suggested you? I'm sure it will be a lot easier to talk to you once you finish it.

 

[SiliconWars]

I already explained the margins and it's clear I was far closer to the mark than you were. How can they be 15% gross when they are "mid teens" (you realise that mid teens is between 15 and 17 right?) operating? Let me guess, you believe there is absolutely zero additional cost involved and you're taking the 15 as the "mid teens". What absolute garbage.

What else did you say, consoles would be worth about $20 million a quarter?

 

[AtenRa]

Can someone ask mrmt how did he calculated or come to a conclusion of 15% Gross Margins without knowing the COGs (Cost Of Goods) of the Consoles ??

Also, AMDs strategy is to have an OPEX of 450M per quarter, that includes R&D, Marketing and Administrative. They are targeting 50% of their Revenue to be from Semi-Custome and Embedded products. So even if CPU Revenue continue to decline, they will compensate with Custom/Embedded.

 

[SiliconWars]

Here is what he said 2 months ago - http://forums.anandtech.com/showpost.php?p=35418952&postcount=29

Here is what I said in response - http://forums.anandtech.com/showpost.php?p=35420275&postcount=59

It's pretty clear who is closer to the mark, with console revenue more than double what he said it would be and the 30% of revenue mark is almost 3 times higher than his 12% mark. Also profit is more than 3x higher than his estimate of $25 million.

 

[Abwx]

Despite their reduced means they are moving forward...

Advanced Micro Devices said Thursday that it would tape out the first products to be manufactured using 14nm FinFET and 20nm planar process technologies in the coming quarters.

Given the timing, Ms. Su referred to TSMC’s 20nm process technology that is going online in February ’14 and that is likely to be used to manufacture AMD’s next-generation graphics processing units. In addition, AMD will use 14nm-XM FinFET process technology from GlobalFoundries to make certain low-power products. 14nm-XM should enter mass production stage sometimes in calendar 2014.

http://www.xbitlabs.com/news/other/...nm_FinFET_Chips_Within_Next_Two_Quarters.html

 

[Phynaz]

So they haven't reached tape out of their next products yet?

Ugh. That puts them over a year away.

 

[rainy]

So they haven't reached tape out of their next products yet?

Ugh. That puts them over a year away.

I don't know why are you suprised - it's quite obvious that AMD will use 20nm and below starting from 2015 (except GPU).

While AMD did mention 20nm and 14nm FinFET process technologies, it does not mean that products made using both will actually be available in 2014. Most likely the company will only tape-out new products that will reach the market in late 2014 or early 2015. Virtually all AMD roadmaps indicate that the vast majority of AMD 2014 products will be made using 28nm and 32nm SOI process technologies. Still, the company will likely introduce new graphics chips made using 20nm fabrication process.

 

[Abwx]

I don't know why are you suprised - it's quite obvious that AMD will use 20nm and below starting from 2015 (except GPU).

Surely but what wasnt obvious was that they ll use
STMicro 14nm XM wich is licenced to GF, likely for their
ARM based offering but Jaguar could well also be used.

 

[ShintaiDK]

So nextgen AMD 20nm GPUs are CPUs are already into the 2015 timeframe. Ouch.

If nVidia suffers the same on the GPU front, then we talk serious GPU stagnation.

And that 14nm finfet is really a 20nm finfet. Just a question on how many years GloFo will add to its delays.

 

[SiliconWars]

It's not, it's a 14nm FinFET with other 20nm elements.

 

[Phynaz]

It's not, it's a 14nm FinFET with other 20nm elements.

Really? What's the metal pitch of both nodes?

Hint: It's the same.

How about contacted gate pitch? Is it scaling by 0.7?


There's my technical contribution that you wanted.

 

[Abwx]

So nextgen AMD 20nm GPUs are CPUs are already into the 2015 timeframe. Ouch.

If nVidia suffers the same on the GPU front, then we talk serious GPU stagnation.

And that 14nm finfet is really a 20nm finfet. Just a question on how many years GloFo will add to its delays.

It s the products that are taped out , not the processes...

On the GPU front expect nexts gens to be more or less
die shrinks or extensions of the currents offerings , nothing
that would require revolutionary implementations.

 

[rainy]

So nextgen AMD 20nm GPUs are CPUs are already into the 2015 timeframe. Ouch.

I expect new generation of Radeons would be released late 2014 - quite possible it could be early 2015 as well.

And that 14nm finfet is really a 20nm finfet. Just a question on how many years GloFo will add to its delays.

I think no one knows that - GlobalFoundries for sure is a a big mystery.

 

[Abwx]

Really? What's the metal pitch of both nodes?

Hint: It's the same.

Other hint : ASML s higher precision offering has
18nm max resolution for xtors size...

No one will offer real 14nm processes given the state
of the available technology , and that include Intel
and TSMC.

 

[Phynaz]

I think no one knows that - GlobalFoundries for sure is a a big mystery.

GloFo announced their 14nm node over a year ago. I'll try to find the details, but I recall it being a 64nm metal pitch and 80nm poly pitch.

 

[SiliconWars]

@ Abwx Yep that's correct, GF actually states "14nm class" instead of saying 14nm though.

 

[Abwx]

Indeed but it s funny that the same people never questionned
Intel alleged capability to do actual 14nm despite they re using
the same lithographic machinery as other foundries and that
such tools are incapable of real 14nm.

 

[Khato]

Indeed but it s funny that the same people never questionned
Intel alleged capability to do actual 14nm despite they re using
the same lithographic machinery as other foundries and that
such tools are incapable of real 14nm.

Good thing then that Intel's 22nm isn't 'actually' 22nm eh? Since that means they can achieve the areal scaling of 22nm to 14nm without having to pattern any 14nm features.

 

[Abwx]

Good thing then that Intel's 22nm isn't 'actually' 22nm eh? Since that means they can achieve the areal scaling of 22nm to 14nm without having to pattern any 14nm features.

That s quite possible , their current node is at 25.8nm ,
the usual 1.4 shrinking would involve going to 18.4nm ,
right at the 18nm max resolution allowed by their equipement.

 

[Khato]

That s quite possible , their current node is at 25.8nm ,
the usual 1.4 shrinking would involve going to 18.4nm ,
right at the 18nm max resolution allowed by their equipement.

While I have no idea where you came up with those numbers it doesn't matter that they're off by a bit as the effect is the same. Of course as usual Intel will fall slightly short of the 'perfect' scaling that 22nm -> 14nm would imply, but they're going to be pretty close.

 

[Abwx]

While I have no idea where you came up with those numbers it doesn't matter that they're off by a bit as the effect is the same. Of course as usual Intel will fall slightly short of the 'perfect' scaling that 22nm -> 14nm would imply, but they're going to be pretty close.

They will do "14nm" lithography with this :

The NXE:3300B is the successor to the NXE:3100, offering 22 nm resolution with conventional illumination and 18 nm with off-axis illumination

http://www.asml.com/asml/show.do?lang=EN&ctx=46772&dfp_product_id=842

So if i m off it s forcibly that my estimation was still too
optimist in respect of actual possibilities...

 

[Khato]

They will do "14nm" lithography with this :

http://www.asml.com/asml/show.do?lang=EN&ctx=46772&dfp_product_id=842

So if i m off it s forcibly that my estimation was still too
optimist in respect of actual possibilities...

That's a good laugh. Considering the fact that Intel has stated that they don't believe EUV will be ready for use on the 10nm node, much less 14nm...