AMD Q414 results

[ClockHound]

Hey IDC, I'm going to put you on the spot. I'm really interested in what you think the answer is for AMD.

Monday morning you're President, CEO and Chairman of AMD. What's your strategy?

Thanks!

I'm not IDC and I don't even play him on internet TV, but if he's as shrewd as he seems, he would devote most of that substantial/moderate/insignificant R&D budget to developing a negentropy generator - then complete his thermodynamic trifecta metaphor and ask the BoD for a huge bonus. Then quit and retire to his overclocking negentropy lab where the higher the clock, the lower the temps - without delidding!

 

[mrmt]

Not only has AMD's R&D contracted, but they investing in future products to a lesser extent than even Nvidia

There was a small snafu in the last Q&A that some investors took note:

Lisa Su implied that R&D would be mostly spared of the cuts, but what we saw in Q414 is completely the opposite, SG&A was more or less intact while R&D fell 15%. This was completely the opposite of what everyone was expecting. The last round of cuts will bring their R&D expenses to a range of 900-950MM, or almost 50% smaller than Nvidia R&D budget. Not a good prospect for the execution of the previous pipeline.

On top of that Lisa raised the issue of the channel problems they were facing, which will take two quarters to fix up. It seems that AMD was reaching its sales forecast at the expense of stuffing the channel with products they knew they wouldn't be able to sell at all, and now the channel is throwing up everything back to AMD. It is no small problem, that's were the 200MM in WSA chips not ordered went and AMD will need two quarters to clean up the mess, half a year worth of chips flying around. No wonder Rory was fired.

It also changes everything regarding their discourse of improving the product mix: AMD wasn't improving its product mix, it wasn't unable to sell to the channel, hence the business shrank and the OEM share (which has higher ASP) became more representative. A failure being portrayed as a success.

Those snafus also raises questions about the credibility of the BoD. Is what they saying really what they mean? Do they know what they are saying or are they deliberately misleading them? Not a good start for Lisa Su.

 

[el etro]

Not only has AMD's R&D contracted, but they investing in future products to a lesser extent than even Nvidia

But Nvidia have a larger number of assets, and their R&D money still cover the SoCs/ARM lines(and these SoC products compete with competitors as tough as Intel). So, both Nvidia and AMD spent the about the same money on both developing CPUs and GPUs.

 

[el etro]

You mean the 3/4 module APU with tiny iGPU NTMBK mentioned in post #301?

Here are a few factors to consider:

1.) Will Zen be late or on time?
2.) Will Zen be a true desktop performance CPU or something optimized for mobile?
3.) What die configurations will Zen come in? (iGPU less, with tiny iGPU or with large iGPU?)
4.) How soon could AMD ready a 3/4 module APU with tiny iGPU? (I personally would want the die to be based on 3 module APU with tiny iGPU for desktop.)
5.) How does AMD handle the WSA commitment in order to make big cores based on smaller die (ie, tiny iGPU) an incentive? Right now, as has been mentioned in previous posts AMD is paying for wafers they don't need anyway.

My bets:

1) Zen will be more or less on time, can be delayed a little or coming early a little. AMD CPU division spent a big part of its resources fixing the original Bulldozer performance, increasing the IPC of the processors three times over the five years of Bulldozer cycle(2011-2015);

2) I bet AMD is making something that can substitute both Jaguar/Puma and K10/Bulldozer line of processors with Zen, i expect a core that is a mid-term on this(With decent IPC this time);

3) The same die configurations as Kaveri is being sold. Maybe a MOOOOOAR CORZZ version can land at servers;

4) Six cores will not help. The best approach for 100 percent(yes, i'm including enthusiast users) of desktop PCs is 4 cores at max(The Intel's approach);

5) I pass on this, i don't have a well developed opinion on this.

 

[mrmt]

But Nvidia have a larger number of assets, and their R&D money still cover the SoCs/ARM lines(and these SoC products compete with competitors as tough as Intel). So, both Nvidia and AMD spent the about the same money on both developing CPUs and GPUs.

What larger assets whoud impact R&D spending? And how Nvidia SoCs are more complex than the SoCs AMD is developing? In fact, K12 is aiming to be more ambitious and more complex than anything Nvidia has on its pipeline. And AMD still has to pay for Zen.

AMD is trying to develop more product lines with less money than Nvidia.

 

[III-V]

Again, they can't (at least now) because they would lose the x86 license.

I don't think the FTC would allow that.

 

[el etro]

AMD is trying to develop more product lines with less money than Nvidia.

Is exactly what i said.

:thumbsdown:


But a CPU architecture is a CPU architecture anyway, and both Nvidia and AMD is developing one. With the word i get K12 and Zen becoming very similar, so AMD will have virtually only one processor arch in the market, like Nvidia.

 

[jpiniero]

I don't think the FTC would allow that.

Why not? What is so necessary about x86 at this point? Windows? The ARM companies are really Intel's competition these days.

 

[svenge]

Monday morning you're President, CEO and Chairman of AMD. What's your strategy?

Take the "golden parachute" and resign.

 

[mrmt]

Is exactly what i said.

:thumbsdown:


But a CPU architecture is a CPU architecture anyway, and both Nvidia and AMD is developing one. With the word i get K12 and Zen becoming very similar, so AMD will have virtually only one processor arch in the market, like Nvidia.

The uncore should be shared but the core design should not. It is still a massive undertaking to develop two custom cores of two different instruction sets like AMD is doing.

 

[III-V]

Is exactly what i said.

:thumbsdown:


But a CPU architecture is a CPU architecture anyway, and both Nvidia and AMD is developing one. With the word i get K12 and Zen becoming very similar, so AMD will have virtually only one processor arch in the market, like Nvidia.

Layout is expensive, though, and they have more designs.

Although at this point, AMD has fallen so far behind that that's barely the case anymore.

Take the "golden parachute" and resign.

No kidding!

 

[ShintaiDK]

AMDs biggest problem as I see it is, that they refuse toa ccept their position. And instead trying to go into segments they dont belong at, at the expense of the segments they can master.

Look at the interview with Huang from nVidia. He knows exactly as a CEO what segments nVidia can and cant compete in. And if they cant compete, they withdraw and try somewhere else.

When we look on AMD, its almost the exact opposite case. AMD sniffed the server segment in a short period, and they ahve wanted to get it back ever since. And it affects their entire product stack. Even with complete failure, they still launch a project like K12 that shouldnt exist at all. Again in a desperate attempt to get back into the server space in the hope of feeling the past glory. K12 is a prime example of whats wrong with AMD, if Bulldozer wasnt enough to show it.

 

[jpiniero]

When we look on AMD, its almost the exact opposite case. AMD sniffed the server segment in a short period, and they ahve wanted to get it back ever since. And it affects their entire product stack. Even with complete failure, they still launch a project like K12 that shouldnt exist at all. Again in a desperate attempt to get back into the server space in the hope of feeling the past glory. K12 is a prime example of whats wrong with AMD, if Bulldozer wasnt enough to show it.

Except Bay Trail has totally destroyed the Cat cores, and that was the one place where AMD was having some success. And with Intel selling it for basically nothing or less there's no real avenue to fight back. They don't really have a future in x86.

 

[III-V]

Why not? What is so necessary about x86 at this point? Windows? The ARM companies are really Intel's competition these days.

Well if that's the case, why does it matter if they lose their license?

 

[jpiniero]

Well if that's the case, why does it matter if they lose their license?

Because the existing x86 products are still the main source of their revenue, not to mention the console deals. It doesn't cost them very much to continue to milk what they have now even as sales decline.

 

[Arachnotronic]

Except Bay Trail has totally destroyed the Cat cores, and that was the one place where AMD was having some success. And with Intel selling it for basically nothing or less there's no real avenue to fight back. They don't really have a future in x86.

Intel is likely getting 60%+ gross profit margins on these chips. Remember, approximately 15% of all of Intel's PC chip sales are Bay Trail based, so they need to be making good money on these.

I think where things get more difficult for AMD is that Intel, in optimizing its platform BoM costs for tablets, will feed right back into their PC platforms.

AMD is too busy spending its R&D on K12, micro-servers, and dGPUs to be able to afford the tight platform level optimization that Intel is doing, IMO, and this is why it won't be able to compete as effectively as some might hope.

 

[III-V]

Except Bay Trail has totally destroyed the Cat cores, and that was the one place where AMD was having some success. And with Intel selling it for basically nothing or less there's no real avenue to fight back. They don't really have a future in x86.

Contra revenue only applies to tablets. If AMD's lost market share in notebooks and inexpensive desktops, it's not contra revenue that's to blame.

I actually just had an interesting thought... if Intel manages to get their graphics game together in the next few years, they could have a product capable of winning the next consoles. Sure, Intel might not be interested in the margins, but it's fun to think about.

 

[ShintaiDK]

Except Bay Trail has totally destroyed the Cat cores, and that was the one place where AMD was having some success. And with Intel selling it for basically nothing or less there's no real avenue to fight back. They don't really have a future in x86.

Do you think they have it easier in the ARM segment where they are a nobody, with more players and even lower prices making a server oriented CPU?

 

[Arachnotronic]

Contra revenue only applies to tablets. If AMD's lost market share in notebooks and inexpensive desktops, it's not contra revenue that's to blame.

I actually just had an interesting thought... if Intel manages to get their graphics game together in the next few years, they could have a product capable of winning the next consoles. Sure, Intel might not be interested in the margins, but it's fun to think about.

I'm sure Intel would be interested in the volumes that those would bring into the fabs.

 

[III-V]

I'm sure Intel would be interested in the volumes that those would bring into the fabs.

That's very true.

 

[jpiniero]

Do you think they have it easier in the ARM segment where they are a nobody, with more players and even lower prices making a server oriented CPU?

Well, I won't say it would be successful. It is AMD after all. But the idea is that they would be able to build a more competitive server product (from a MT perf/watt perspective) since they don't have to pay the x86 tax.

 

[Arachnotronic]

Well, I won't say it would be successful. It is AMD after all. But the idea is that they would be able to build a more competitive server product (from a MT perf/watt perspective) since they don't have to pay the x86 tax.

Just how large do you think the "x86 tax" is in the scheme of a complex server processor?

 

[ShintaiDK]

Well, I won't say it would be successful. It is AMD after all. But the idea is that they would be able to build a more competitive server product (from a MT perf/watt perspective) since they don't have to pay the x86 tax.

The issue I got is the server part. The company simply dont belong there. And they are trying to catch a ghost while they negligate the rest of their products. There is no success stories with ARM in servers either.

To put it bluntly, K12 is a complete flop of a product if it doesnt create some kind of server ARM revolution that can get billion $ revenue figures.

 

[monstercameron]

The issue I got is the server part. The company simply dont belong there. And they are trying to catch a ghost while they negligate the rest of their products. There is no success stories with ARM in servers either.

To put it bluntly, K12 is a complete flop of a product if it doesnt create some kind of server ARM revolution that can get billion $ revenue figures.

would they need billions of revenue to be profitable and in the game?

 

[III-V]

would they need billions of revenue to be profitable and in the game?

Annually, yes. Quarterly, no.