"AMD Moves Away From PCs Amid Steep Losses"

[pelov]

Graphics IP cant be worth much really. And you most likely dont want to compete against nVidia either in a marketspace that is slowly going out. (Discrete cards.)

There is no GPU/CPU integration. Its hard bolted on.

OpenCL was initially developed by Apple Inc., which holds trademark rights...

They're still using PowerVR and they haven't been too happy with nVidia either.

 

[ShintaiDK]

So basically we are stuck with i7-3960x as the top CPU for the next 20 years.

Only if you failed at business 101.

 

[ShintaiDK]

They're still using PowerVR and they haven't been too happy with nVidia either.

Do nVidia hold any advantage over PowerVR in the SoC area? The answer is most likely none at all.

If Apple wanted graphics, they would basicly buy something else. Merging large parts are always bad. ATI/AMD merger also failed down to the employee level.

Even Intel didnt want ATI or nVidia.

 

[v8envy]

It would cost billions to do that. You wouldnt be able to make x86 CPUs. So we are talking about a monthly expense of around 500mio$.

Then you got 10000 employees you more or less want to get rid of. That would also quickly cost a billion or 2. Then there are contracts and obligations, unknown amount there. But anything than pennies.
Lastly you got about 3.6billion in liabilities that also needs to be payed.

Compare that to maybe buy the IPs you want and hire the engineers and other people for maybe 100-250mio later on.

I´m kind enough to say AMDs cash=stock price in the overhead.

This, this, a thousand times this. People forget that buying assets of a company does not equate to getting the liabilities, senior management, IT, software licenses, etc. Does a successful Korean company like Samsung really want the vision and track record of AMD leadership so strongly they'd be willing to pay the costs of the baggage that comes with them? I'm going out on a limb here to say "hell no." Looking at the stock's performance today fewer people than ever want those guys on their team.

AMD brand is an asset. AMD IP is an asset. AMD's phone system, ERP system, employees, facilities are also assets. If nobody is willing to buy them as a bundle deal they won't be sold as a bundle deal.

I don't have any skin in the game, but if I did I'd be betting on Chapter 11 best case.

 

[Idontcare]

So basically we are stuck with i7-3960x as the top CPU for the next 20 years.

Nah, Intel needs people to upgrade their computers otherwise its own revenue shrivels.

Intel has little choice but to compete with itself, or rather the Intel it was about 2yrs ago.

They have to compete for price/performance to make a compelling upgrade argument or people will just hang onto their existing computers.

 

[mrmt]

Trinity shipments shows that the product is doing fine, they just need to sell more not only in Mobile.

Is it? Trinity is a chip bigger than Llano, so unless yields are drastically different, it needs better ASP than Llano to break even. The opposite is just happening. ASP is going down, but this isn't everything.

I haven't seen the Q3 numbers but Im betting they lost more Sales/Revenue from the desktop and Server segments than Mobile.

CFO commentary states that APU >>shipments<< were flat as a percentage of the Client products, so they are losing sales across the board, APU (Trinity) included.

You may argue that Trinity should not be suffering from losses in ASP as, for example, Bulldozer for desktop is, but that's another story.

The fact that they decided to withhold old Llano inventory instead of fire sale it says a lot about how strong the demand for AMD chips is in the channel. Keep in mind that it will be a lot harder to sell those chips in the future, and in this case, Llano sales at a much lower prices are going to compete with Trinity, further impairing the product.

Keep in mind that AMD is already on its bottom limit on gross margins. The lower they go now, the faster their cash burns. To make things worse, they have nothing in the pipeline to change this scenario.

 

[pelov]

Do nVidia hold any advantage over PowerVR in the SoC area? The answer is most likely none at all.

If Apple wanted graphics, they would basicly but something else. merging large parts are always bad. ATI/AMD merger also failed down to the employee level.

They *need* the graphics. They're still relying on PowerVR on the low end and nVidia on the high end. ATi offers very potent graphics IP for their desktops/laptops and the flexibility of integrating it with a CPU for added performance via openCL. The only gap it doesn't fill is the ubermobile space, but if they're able to trim their current approach to ARM-SoC level then it would offer that as well.

openCL is Apple's baby, not AMD's. It just so happens that, at least for the moment, AMD does the most with it. If Apple can take that they will absolutely make a killing

 

[mrmt]

They *need* the graphics. They're still relying on PowerVR on the low end and nVidia on the high end. ATi offers very potent graphics IP for their desktops/laptops and the flexibility of integrating it with a CPU for added performance via openCL. The only gap it doesn't fill is the ubermobile space, but if they're able to trim their current approach to ARM-SoC level then it would offer that as well.

AMD brings a lot of liabilities with it. The debt, management, obscure agreements with globalfoundries and a CPU business that today is entangled with the GPU business. Plus it takes time to integrate the IP in the product line.

It will be easier to bring aboard AMD/ATI teams on board for a fraction of the price with no overhead at all and develop your own technology.

 

[ultimatebob]

Lol, you don't get how business works, do you? Intel would not sell CPUs for $700 because nobody would buy them. Period. Intel has a lot of competition at the low end much more potent than AMD. Their names are Apple, Samsung, Qualcomm, and Nvidia.

I would think that Intel's CPU prices would slowly creep upward without any competition from AMD. I don't see $700 price tags happening for low end processors, but I wouldn't be surprised if the next generation Core iX processors end up costing 10-15% more instead of prices staying flat or going down as they have before.

I could also see Intel stretching out their release cycles even more than they already have. We may have to wait 18-24 months for new processor generations instead of the 12-18 months as we do now.

 

[mrmt]

I would think that Intel's CPU prices would slowly creep upward without any competition from AMD. I don't see $700 price tags happening for low end processors, but I wouldn't be surprised if the next generation Core iX processors end up costing 10-15% more instead of prices staying flat or going down as they have before

Not gonna happen. When AMD fans see a monopoly, Intel sees ARM around the corner. Intel has been shrinking x86 cores and investing in very power efficient chips thinking on this clash.

When Intel started Ultrabooks they just didn't want a sleek laptop, they were thinking in convertible tablets, combining the convenience of the tablet with the power of a full fledged PC. It is in this form factor and smartphones that matters now. To charge 700 USD for a chip would simply kill the desktop/conventional laptop market faster, and this is every that Intel does not want.

And on this scenario, AMD is largely irrelevant.

I could also see Intel stretching out their release cycles even more than they already have. We may have to wait 18-24 months for new processor generations instead of the 12-18 months as we do now.

Not gonna happen, for the same reason stated above. Intel is simply accelerating their road maps.

 

[BenchPress]

I read the transcript i don't see anything new in it - APUs, APUs, APUs.

I think they have no new strategy right now. [...] AMD has no clue how their graphics IP will help their business

+1

 

[pelov]

When Intel started Ultrabooks they just didn't want a sleek laptop, they were thinking in convertible tablets, combining the convenience of the tablet with the power of a full fledged PC. It is in this form factor and smartphones that matters now. To charge 700 USD for a chip would simply kill the desktop/conventional laptop market faster, and this is every that Intel does not want.

Although I agree with you that Intel had something up their sleeves with Ultrabooks, they didn't play along with the pricing. Their ULV chips still cost a fortune and Ultrabooks in turn cost far too much to be attractive, hence the very disappointing sales figures. If you're paying $300 for a processor, and I do mean just the processor, how is that going to compare to a $299 Google Nexus where you get the entire package? Intel wanted Ultrabooks to compete with tablets and offer something ultramobile, but thus far they haven't been playing ball with the pricing. These new hybrid devices are going to cost even more than Ultrabooks and things don't look to be changing with Haswell either.

 

[Olikan]

CFO commentary states that APU >>shipments<< were flat as a percentage of the Client products, so they are losing sales across the board, APU (Trinity) included.

flat shipments in a shrinking market is a good thing
ok, ok not so much...
it just mean that you have more time to run away 😛

 

[mrmt]

Intel wanted Ultrabooks to compete with tablets and offer something ultramobile, but thus far they haven't been playing ball with the pricing. These new hybrid devices are going to cost even more than Ultrabooks and things don't look to be changing with Haswell either.

They did want both to compete but on Intel terms, not ARM terms. Intel thinks it can make the consumer pay more for a single device that can do what both a PC and a Tablet do, so it's justifiable the price disparity.

But in the medium term, this is also going to change. Intel has been shrinking the size of its processors generation to generation since Conroe, to the point that the smallest Ivy Bridge processer isn't really bigger than the biggest ARM counterpart. A single core IVB or Haswell should have comparable size to most ARM SoCs, and don't forget Atom. Silvermont might surprise us. So if Intel can reach the right volumes, they will gladly sell smaller chips for less money.

 

[mrmt]

flat shipments in a shrinking market is a good thing
ok, ok not so much...
it just mean that you have more time to run away 😛

That's not what he said. He said that shipments went down but APU percentage of shipments was the same.

 

[v8envy]

If you mean the Google Nexus 7, it's $199, not $299. Including a decent quality IPS display.

I don't think Intel has much wiggle room to play around with ULV pricing, not without losing the margins on the same chips being sold to Apple. Why would apple pay $300 for a CPU when Acer is getting the same thing for $100?

The problem lies in PC manufacturers not being able to command the same price premiums Apple does. This has nothing to do with Intel and everything to do with consumers realizing that buying the single luxury model from a company optimized to produce and service budget products is not going to result in the same ownership experience as buying a luxury model from a company withing a lineup of nothing but luxury products.

 

[pelov]

They did want both to compete but on Intel terms, not ARM terms. Intel thinks it can make the consumer pay more for a single device that can do what both a PC and a Tablet do, so it's justifiable the price disparity.

But in the medium term, this is also going to change. Intel has been shrinking the size of its processors generation to generation since Conroe, to the point that the smallest Ivy Bridge processer isn't really bigger than the biggest ARM counterpart. A single core IVB or Haswell should have comparable size to most ARM SoCs, and don't forget Atom. Silvermont might surprise us. So if Intel can reach the right volumes, they will gladly sell smaller chips for less money.

I wish you were right, but if we take a look at the actual die sizes
http://www.anandtech.com/show/5876/the-rest-of-the-ivy-bridge-die-sizes
http://www.anandtech.com/show/6330/the-iphone-5-review/4

You can see that just isn't the case. Meanwhile, you're getting the entire package with the SoC while you only get the on-die GPU + CPU for the Ivy H-2/M-2.

The issue here isn't die size, it's the cost attributed to it. For Intel, they still believe that charging consumers an x86 tax is going to work yet they're neglecting their sales figures which show that, while it may have worked before, it's changing.

If Intel really meant to take on ARM tablets with Ultrabooks, they'd have priced them accordingly. OEMs are operating at sub-7% margins for Ultrabooks. While they offer better performance, people don't care for them nor will pay the inflated price tags.

Instead of directly decreasing prices for the chips:

- Intel attempted to standardize plastics for the chassis so OEMs don't have to use magnesium/aluminum allows. That failed miserably
- standardizing an mSATA form factor
- allowed a very vague description of Ultrabooks (5lb weight, 15.6" size, 5k RPM drives with small cache, etc.)
- standardizing batteries
- marketing

Only the marketing affects Intel's pockets directly, while the others are meant to alleviate the very low margins on OEMs. Basically, Intel is willing to decrease the quality of an Ultrabook and make it less appealing to consumers so long as they get to keep their sky high prices on their chips for said Ultrabooks.

Haswell changes only some of this, in that it's made specifically for that TDP range. This should mean we see far less throttling under load and better performance, but if you believe the rumors, the prices have also gone up. Although throttling was clearly an issue, the biggest detractor is still the price and that's looking like it's going to be even worse.

 

[Genx87]

I wish you were right, but if we take a look at the actual die sizes
http://www.anandtech.com/show/5876/the-rest-of-the-ivy-bridge-die-sizes
http://www.anandtech.com/show/6330/the-iphone-5-review/4

You can see that just isn't the case. Meanwhile, you're getting the entire package with the SoC while you only get the on-die GPU + CPU for the Ivy H-2/M-2.

The issue here isn't die size, it's the cost attributed to it. For Intel, they still believe that charging consumers an x86 tax is going to work yet they're neglecting their sales figures which show that, while it may have worked before, it's changing.

If Intel really meant to take on ARM tablets with Ultrabooks, they'd have priced them accordingly. OEMs are operating at sub-7% margins for Ultrabooks. While they offer better performance, people don't care for them nor will pay the inflated price tags.

Instead of directly decreasing prices for the chips:

- Intel attempted to standardize plastics for the chassis so OEMs don't have to use magnesium/aluminum allows. That failed miserably
- standardizing an mSATA form factor
- allowed a very vague description of Ultrabooks (5lb weight, 15.6" size, 5k RPM drives with small cache, etc.)
- standardizing batteries
- marketing

Only the marketing affects Intel's pockets directly, while the others are meant to alleviate the very low margins on OEMs. Basically, Intel is willing to decrease the quality of an Ultrabook and make it less appealing to consumers so long as they get to keep their sky high prices on their chips for said Ultrabooks.

Haswell changes only some of this, in that it's made specifically for that TDP range. This should mean we see far less throttling under load and better performance, but if you believe the rumors, the prices have also gone up. Although throttling was clearly an issue, the biggest detractor is still the price and that's looking like it's going to be even worse.

Makes you wonder how hard Wintel is going to fall in the next decade. Android tablets with ARM processors can be had for 200 bucks or less now. And of course smartphones are pretty much 100% non-intel and about 5-8% microsoft. They may not outright crash and burn but will become increasinging irrlevant.

 

[dguy6789]

Sooner or later the only people that remember when AMD made competitive and even superior products to Intel will be old. I'll feel old then.

 

[mrmt]

You can see that just isn't the case. Meanwhile, you're getting the entire package with the SoC while you only get the on-die GPU + CPU for the Ivy H-2/M-2.

I agree with you. Die size advantage is still with ARM. But sizes are at least in the same order of magnitude. What about a single core IVB core, don't you think that it could make a SoC smaller than A6? That would be my guess.

Just to keep things on topic, AMD is selling APUs that are almost three times the size of IVB H2M2.

Haswell changes only some of this, in that it's made specifically for that TDP range. This should mean we see far less throttling under load and better performance, but if you believe the rumors, the prices have also gone up. Although throttling was clearly an issue, the biggest detractor is still the price and that's looking like it's going to be even worse.

I understand that Haswell GT3 is more expensive, but offers more capabilities as well. As I'm not in the OEM channel and have not any kind of inside information, I cannot really say anything.

Anyway, let's get back to AMD because we are already derailing the thread. We can take this subject in a propper thread.

 

[pelov]

Makes you wonder how hard Wintel is going to fall in the next decade. Android tablets with ARM processors can be had for 200 bucks or less now. They may not outright crash and burn but will become increasinging irrlevant.

Imo, it's Intel's most catastrophic failure since Netburst or Larrabee.

Ultrabooks are a fantastic idea in revitalizing that once vibrant netbook area. People can replace their desktop with a very sleek, pretty looking laptop that can do everything their desktop did. Why buy a tablet if you can get the same and more with an Ultrabook for only $100/$150 more?

Instead, Intel approached it like Apple, thinking people would pay the extra money for the "Inspired by Intel" sticker... and no f**ks were given.

For the price of an i7 ULV you can buy 1.5 Google Nexus tablets and do almost everything the Ultrabook can do. You even get a more vibrant ecosystem along with it! Better battery life, a better display, lighter weight too. So what's the point of an Ultrabook?

So instead of attacking the ARM problem head on with actual competition and attractive products at close-enough-to-matter prices, Intel instead decided that they'll take their 35-45W chips (which won't even fit properly), stick them in a metallic shell and charge an arm and a leg for them.

I hope AMD can learn from the above. In fact, if they wish to remain alive then they should. Intel was lazy with their Atom development, letting it lag behind, and then responded to the ARM surge by using a product that doesn't fit with a price tag that doesn't make sense.

 

[frozentundra123456]

Although I agree with you that Intel had something up their sleeves with Ultrabooks, they didn't play along with the pricing. Their ULV chips still cost a fortune and Ultrabooks in turn cost far too much to be attractive, hence the very disappointing sales figures. If you're paying $300 for a processor, and I do mean just the processor, how is that going to compare to a $299 Google Nexus where you get the entire package? Intel wanted Ultrabooks to compete with tablets and offer something ultramobile, but thus far they haven't been playing ball with the pricing. These new hybrid devices are going to cost even more than Ultrabooks and things don't look to be changing with Haswell either.

I do agree that ultrabook prices are way to high. But I also dont think you can compare it to a low end tablet. I have a 7" acer tablet, and I guarantee you that even if I did not game, I would hate to have that tablet as my only computing device. It tries to do everything, but is really good at nothing other than reading books and watching netflix. It is so slow that half the time you cant even watch you tube without lags, battery life is 4 hrs, wi-fi connectivity is terrible, and no matter how much I use it, I still hate the touch interface with a passion. Perhaps if you spent more for a bigger tablet with 3g connectivity and a dock, it would be much more satisfactory, but then you are getting close to low end ultrabook prices. I probably would not buy an ultrabook, because they are too expensive for the performance you get, but I would sorely hate to rely on a tablet for my only computing device.

 

[cytg111]

stupid question : couldnt amd license the arm tech and glue an x86 decoder on it for a quick win?

 

[ShintaiDK]

Although I agree with you that Intel had something up their sleeves with Ultrabooks, they didn't play along with the pricing. Their ULV chips still cost a fortune and Ultrabooks in turn cost far too much to be attractive, hence the very disappointing sales figures. If you're paying $300 for a processor, and I do mean just the processor, how is that going to compare to a $299 Google Nexus where you get the entire package? Intel wanted Ultrabooks to compete with tablets and offer something ultramobile, but thus far they haven't been playing ball with the pricing. These new hybrid devices are going to cost even more than Ultrabooks and things don't look to be changing with Haswell either.

Surface is 499 and ARM...

You tend to get what you pay for.

Ultrabooks are still in their infant stage, where the production lines are getting developed and optimized with standardlizing of components.

And for that google chromebook.
http://www.google.com/intl/en/chrome/devices/samsung-chromebook.html#specs

2GB memory, 16GB NAND, 1.7Ghz Exynos 5 dualcore CPU.

And you can get the chromebook with more memory, Intel Core CPU etc for 449$.
http://www.google.com/intl/en/chrome/devices/chromebook-samsung-550.html#specs

4GB memory, 16GB NAND, 1.3Ghz dualcore Celeron 867.

Even google knows ARM is too slow.

The 867 is also replaced by the 887 at 1.5Ghz. 17W TDP and 86$.

 

[pelov]

I do agree that ultrabook prices are way to high. But I also dont think you can compare it to a low end tablet. I have a 7" acer tablet, and I guarantee you that even if I did not game, I would hate to have that tablet as my only computing device. It tries to do everything, but is really good at nothing other than reading books and watching netflix. It is so slow that half the time you cant even watch you tube without lags, battery life is 4 hrs, wi-fi connectivity is terrible, and no matter how much I use it, I still hate the touch interface with a passion. Perhaps if you spent more for a bigger tablet with 3g connectivity and a dock, it would be much more satisfactory, but then you are getting close to low end ultrabook prices. I probably would not buy an ultrabook, because they are too expensive for the performance you get, but I would sorely hate to rely on a tablet for my only computing device.

They may not be for you and me, but many people have already begun replacing their laptops and desktops with tablets. Because people are also inclined to spend more money on tablets/smartphones and replace them more often than PCs, that means Intel/AMD have to offer something else, and preferably in the same price range, to compete. It's the first time PC sales have slipped since 2001 and most analysts believe they are going to lose more and more ground going forward.

Think about it,

Can you really make a convincing argument for the average user to buy the new Haswell chip if they've already bought a PC in the last 4-5 years? What about a shiny new tablet/smartphone?

Ultrabooks are still in their infant stage, where the production lines are getting developed and optimized with standardlizing of components.

That's not going to impact the price much, if at all. It might give OEMs a bit more wiggle room, but when the processor costs more than an entire tablet you have to ask where the savings are going to come from. If you honestly think standardizing batteries and mSATA form factors is going to cut the cost by $400, you're absolutely insane.

And for that google chromebook.
http://www.google.com/intl/en/chrome...ook.html#specs

Tell me whats missing.

And you can get the chromebook with more memory, Intel Core CPU etc for 449$.
http://www.google.com/intl/en/chrome...550.html#specs

The ARM-based Chromebook is $200 cheaper. Is that what you're getting at?