AMD back in gear, Centurion FX

[AnandThenMan]

The market leader generally has more money and more push on developing new innovations and these innovations are more likely to be embraced by the industry (having a greater marketshare)

Embraced by what industry? In x86 Intel is effectively a monopoly. Not to mention Intel has copied AMD in several key respects, and did so free of charge no less. You would have a valid point if it was similar to the auto industry where there is intense competition from multiple players in every price segment.

But as it stands, there is Intel gobbling up the vast majority of profit share, and AMD gets the scraps. This forces AMD to differentiate as much as possible because they are held hostage by the x86 license. And let's not forget that the only reason AMD can even make x86 silicon for Sony and Microsoft is because Intel was forced into it as part of the out of court cash settlement.

x86 processor space is fairly unique IMO. If x86 was something any company could freely license, this would be a completely different discussion. Notice how in several other market segments that are not locked down (meaning x86) Intel competes poorly or not at all. It's easy to be the best at something when you have no competition.

On AMD pushing HSA, Intel can freely join and field competing products, nothing is stopping them. In fact x86 in consoles will help Intel because the gaming codebase will be very similar and able to leverage Intel processors.

 

[Sleepingforest]

Not exactly on topic, but I'd like to contribute some background about why Intel is an effective monopoly (and it's not going to be RAH RAH Intel's the BEST!)

The biggest problem is the huge financial barrier to entry. A single fab (which is the building essential to producing semiconductors and thus CPUs) is going to cost upwards of a billion dollars, and that's just the physical equipment. The incredibly high precision required of these instruments and the need for perfect purity within the clean room is what contributes to these costs.

Even if your company is going to be fabless (contracts the actual production out to another company, like GlobalFoundries or TSMC), this does not even count in the costs of researching and developing the technology necessary to get down to a competitive manufacturing node in the CPU space (28nm to 22nm right now). Intel spends about $2.5 billion a year trying to move to the next process node, and it still takes them over a year each time.

Can you imagine going to investors and saying "I'm going to need a couple of years to begin producing, and it's going to take more than 2.5 billion dollars a year just to be able to compete with the existing companies. In fact, if we want to have vertical integration, it's going to take another several billion dollars up front to match the production rate of our intended rivals. Furthermore, that rival company will be continually improving, and has already hired most of the talent in this field. We're not going to see any revenue at all for the first several years, and we'll almost certainly end up having less than 20% market share in a shrinking market." Not a great sales pitch.

 

[galego]

Not exactly on topic, but I'd like to contribute some background about why Intel is an effective monopoly (and it's not going to be RAH RAH Intel's the BEST!)

The biggest problem is the huge financial barrier to entry. A single fab (which is the building essential to producing semiconductors and thus CPUs) is going to cost upwards of a billion dollars, and that's just the physical equipment. The incredibly high precision required of these instruments and the need for perfect purity within the clean room is what contributes to these costs.

Even if your company is going to be fabless (contracts the actual production out to another company, like GlobalFoundries or TSMC), this does not even count in the costs of researching and developing the technology necessary to get down to a competitive manufacturing node in the CPU space (28nm to 22nm right now). Intel spends about $2.5 billion a year trying to move to the next process node, and it still takes them over a year each time.

Can you imagine going to investors and saying "I'm going to need a couple of years to begin producing, and it's going to take more than 2.5 billion dollars a year just to be able to compete with the existing companies. In fact, if we want to have vertical integration, it's going to take another several billion dollars up front to match the production rate of our intended rivals. Furthermore, that rival company will be continually improving, and has already hired most of the talent in this field. We're not going to see any revenue at all for the first several years, and we'll almost certainly end up having less than 20% market share in a shrinking market." Not a great sales pitch.

AMD gained market share in Q4 2012. Intel has dropped its profits by a 25% in Q1 2013.

 

[ShintaiDK]

AMD gained market share in Q4 2012. Intel has dropped its profits by a 25% in Q1 2013.

AMDs CPU division shrinked 10% in Q1. And AMD lost massively the entire 2012 with -37% YoY. The CPU division is on life support and essentially dead. It went from ~1200mio revenue per quarter to ~750mio in a single year. And from ~1300mio to ~750mio the last 15 months.

 

[Cloudfire777]

Since we are talking mobile processors:

AMD Trinity APUs are way ahead of Intel`s Ivy Bridge in games.

Intel`s Ivy Bridge is way ahead of AMD Trinity APUs with CPU bound tasks. Which also is reflected in some games where the HD4000 comes ahead even though the HD4000 have less raw power

Trinity 35W Quad cores is good enough for discrete GPUs up to GTX 660M. Higher than that, it will bottleneck the GPU. 7970M paired with Trinity A10-4600M resulted in the APU chocking the GPU to perform like GTX 660M. Those two GPUs are miles apart-

Ivy Bridge Quad cores does not bottleneck the high end GPUs. Neither does the 35W Dual cores.

If your goal is to get the highest FPS with the lowest amount of money, buy a notebook with a AMD APU.

If you are willing to pay slightly extra money, buy a notebook with i5 + GT 640M. It runs circles around any IGP out there and you can buy one for $800.

If you don`t care about gaming, but want to use the notebook to surf: Ivy Bridge and Trinity are both good choice.

-----------------------------------------------------------------------------------

Question is: How much will Intel gain on Richland when Haswell with GT3 (HD5200) with its own dedicated eDRAM on a die, is released?

 

[Idontcare]

AMDs CPU division shrinked 10% in Q1. And AMD lost massively the entire 2012 with -37% YoY. The CPU division is on life support and essentially dead. It went from ~1200mio revenue per quarter to ~750mio in a single year. And from ~1300mio to ~750mio the last 15 months.

I remember when TI, and then Via (Cyrix @ Nat Semi then) underwent similar market revenue contractions (relative-percentage wise to the overall TAM of course)...TI threw in the towel, as did many others, as the ROI just wasn't there to keep developing future x86-based cores. Via stuck it out, but contracted and contracted until they found a niche they could serve.

That is the cusp of where AMD is at right now in x86. And without a future beyond 28nm at the moment, it looks like everyone in AMD's management (but not their fanboys) has recognized this and planned (internally at least) accordingly.

Given that 14XM is specifically being developed for mobile-only ICs, and that AMD is bound to keep production at GF or pay unacceptably high exclusivity-waiver fees, AMD's future in both discrete GPU and CPU is drawing to a conclusion that is clearly written on the wall (if only people would open their eyes).

 

[Idontcare]

AMD gained market share in Q4 2012. Intel has dropped its profits by a 25% in Q1 2013.

Yes, it is clearly just a matter of time now before AMD will truly come to dominate Intel in ways that many have long dreamed would happen.

 

[jhu]

Not exactly on topic, but I'd like to contribute some background about why Intel is an effective monopoly (and it's not going to be RAH RAH Intel's the BEST!)

The biggest problem is the huge financial barrier to entry. A single fab (which is the building essential to producing semiconductors and thus CPUs) is going to cost upwards of a billion dollars, and that's just the physical equipment. The incredibly high precision required of these instruments and the need for perfect purity within the clean room is what contributes to these costs.

Even if your company is going to be fabless (contracts the actual production out to another company, like GlobalFoundries or TSMC), this does not even count in the costs of researching and developing the technology necessary to get down to a competitive manufacturing node in the CPU space (28nm to 22nm right now). Intel spends about $2.5 billion a year trying to move to the next process node, and it still takes them over a year each time.

Can you imagine going to investors and saying "I'm going to need a couple of years to begin producing, and it's going to take more than 2.5 billion dollars a year just to be able to compete with the existing companies. In fact, if we want to have vertical integration, it's going to take another several billion dollars up front to match the production rate of our intended rivals. Furthermore, that rival company will be continually improving, and has already hired most of the talent in this field. We're not going to see any revenue at all for the first several years, and we'll almost certainly end up having less than 20% market share in a shrinking market." Not a great sales pitch.

One thing your summary doesn't include is that there were many companies making x86 processors in the 80s-90s including IBM, Cyrix, AMD, TI, NEC, etc. Other than AMD, none of them could design a better overall performing x86 processor than Intel.

 

[guskline]

:biggrin:Wait a minute IDC, how can all of this AMD gloom be true ?

Why I'm SURE Centurion will turn it all around (tongue-in-cheek):awe:

 

[Cloudfire777]

AMDs CPU Division (they call it Computing Solution Division):
Net income:
Q1 2012: $124 Millions
Q2 2012: $82 Millions
Q3 2012: -$150 Millions
Q4 2012: -$102 Millions

AMDs GPU Division:
Net income:
Q1 2012: $34 Millions
Q2 2012: $31 Millions
Q3 2012: $18 Millions
Q4 2012: $22 Millions

Q1 Results
Q2+Q3 Results
Q4 Results

 

[AnandThenMan]

...AMD's future in both discrete GPU and CPU is drawing to a conclusion that is clearly written on the wall (if only people would open their eyes).

:hmm:

 

[Charles Kozierok]

On the bolded, what? D: Why does value have to come from the market leader?

It doesn't. But all else equal, most people prefer to go with the market leader, perceiving it as a safer choice. That's just human nature.

Do you want to see Intel with no competition in x86? Because if AMD is a "failing chip company" I take you expect them to soon be no longer in business, so this is exactly what will happen.

No, I very much wish AMD was competing solidly with Intel. But they aren't, and what I want simply does not matter.

AMD gained market share in Q4 2012. Intel has dropped its profits by a 25% in Q1 2013.

With this utterly ridiculous apples-and-oranges comparison, you've proven beyond any doubt that you are not here to discuss this issue seriously, just to rah-rah for AMD.

You cannot talk about market share for one company without looking at market share for the other. You cannot talk about net income for one company without looking at the other.

AMD gained market share? Not that I can see:

Intel's (ticker: INTC) total fourth-quarter unit share increased 150 basis points sequentially to 84.8%, its highest since third-quarter 2002, with gains across the server, desktop and notebook segments.
...
Intel share reaches another 10-year high. On a total microprocessor unit (MPU) unit basis, Intel gained roughly 150 basis points of unit share from 83.3% in third-quarter 2012 to 84.8% in fourth-quarter 2012, its largest share since reaching 86.8% in third-quarter 2002, due to gains across the server, desktop and notebook microprocessor segments.

Even if AMD did slightly increase its market share compared to Intel, it would still be trailing behind massively.

And you have the audacity to bring up Intel's profits declining as a response to claims that it is dominating the industry? Yeah, it declined 25%, to $2 billion in one quarter.

What did AMD do in the first quarter? Lost $146 million.

This sort of intellectual dishonesty makes reasonable discussion impossible.

 

[CHADBOGA]

AMD gained market share in Q4 2012. Intel has dropped its profits by a 25% in Q1 2013.

So when do you see the crossover point in their share prices?

 

[Exophase]

One thing your summary doesn't include is that there were many companies making x86 processors in the 80s-90s including IBM, Cyrix, AMD, TI, NEC, etc. Other than AMD, none of them could design a better overall performing x86 processor than Intel.

Yeah but for most of this time they were making either legally licensed or reverse engineered clones. And while the result was often better than the original they were released much later. If you're in the market of cloning CPUs then competing with the original company's bleeding edge is never even really on the table.

Back then (say, 486 and earlier, although many CPUs called 486 were 386-based) the designs were simple enough and the datasheets described them thoroughly enough that you could get a pretty close clone design through reverse engineering. Later on this would become less and less practical. And a lot of these companies probably became burned out on legal action from Intel.

It wasn't until the mid-90s that you really began to see new uarchs designed from the ground up to compete with Intel. I'd say Cyrix and what effectively became the combined efforts of NexGen + AMD were the only ones seriously trying. IDT (Centaur) was clearly going for a totally different niche and not trying for the performance crown for general purpose code, AMD and Cyrix actually had pretty good designs (K6 line, M2), but were hobbled by two problems: bad FPU performance and being stuck on Socket 7. AMD attempted to mend the former with 3DNow!, technically (much IMO) superior but not viable for market adoption. Socket 7 was a big problem because it meant they were limited to slow on-motherboard L2 cache at a time when on-die cache was too expensive, and AFAIK there were legal barriers to making Slot 1 CPUs. AMD corrected both of those problems with K7, giving it a fast scalar FPU and an Alpha derived FSB and new physical socket + motherboard, and hence started doing pretty well against Intel. IMO, having a separate slot/socket was the final big step in the evolution of being more than just a clone.. but of course the market never would have had room for a bunch of different sockets.

 

[Sleepingforest]

One thing your summary doesn't include is that there were many companies making x86 processors in the 80s-90s including IBM, Cyrix, AMD, TI, NEC, etc. Other than AMD, none of them could design a better overall performing x86 processor than Intel.

This thread is pretty touchy, so I tried to avoid specific names as much as possible. (People still find reason to complain, of course, but I want to avoid lighting a fuse if I can).

 

[Idontcare]

...AMD's future in both discrete GPU and CPU is drawing to a conclusion that is clearly written on the wall (if only people would open their eyes).

:hmm:

The future is fusion, right? Not discrete CPUs and discrete GPUs. If the future wasn't fusion then I'd expect them to have claimed as much.

 

[galego]

I find interesting that the Centurion FX chip is marketed with an "I will be back" motto

http://uk.hardware.info/news/34463/amd-developing-centurion-fx-processor

because AMD is offering some good news:

How Did AMD Garner 20 Percent Market Share?

http://www.crn.in/news/hardware/2013/03/18/how-did-amd-garner-20-percent-market-share-

AMD results beat expectations in declining PC market

http://www.reuters.com/article/2013/04/18/us-amd-results-idUSBRE93H16E20130418

AMD’s Result Confirm Improving Operational Efficiency

http://www.trefis.com/stock/amd/art...m-improving-operational-efficiency/2013-04-19

Whereas Intel continue to feel pain, quarterly profits drop 25% for Q1 2013

http://www.tweaktown.com/news/29763...rterly-profits-drop-25-for-q1-2013/index.html

 

[galego]

So when do you see the crossover point in their share prices?

Who said you there is one?

 

[inf64]

That motto is from 2011 and original BD launch. I guess they didn't use it back then but slides (one above included) leaked to the net.
As for this new chip, I heard this might actually be real. I still think they should have better opted for straight 8350 follow up model, say 8370 or whatever and called it a day.

 

[Phynaz]

How Did AMD Garner 20 Percent Market Share?

http://www.crn.in/news/hardware/2013/03/18/how-did-amd-garner-20-percent-market-share-

I don't know, since the article says they are at 14%.

 

[Sleepingforest]

I'd also say good and bad are relative. AMD is still suffering loses, it's just less than anticipated. Intel actually managed the same feat; here's the first headline I say in Google News after searching "Intel": Intel's Q1: Better than Expected

Personally, I find it riddiculous that random analysts are making predictions and then the stock moves. The scenario, too often, is like this:
Company: We expect $1 billion revenue this quarter!
Analysts: We know better than the company, and predict $1.2 billion!

*company makes $1.1 billion*

Company: Yay!
Analysts: you have failed us for the last time... *stock prices tumble*

 

[guskline]

galego, I'm still waiting for you to post your machine specs!

 

[Mallibu]

I find interesting that the Centurion FX chip is marketed with an "I will be back" motto

http://uk.hardware.info/news/34463/amd-developing-centurion-fx-processor

because AMD is offering some good news:

How Did AMD Garner 20 Percent Market Share?

http://www.crn.in/news/hardware/2013/03/18/how-did-amd-garner-20-percent-market-share-

AMD results beat expectations in declining PC market

http://www.reuters.com/article/2013/04/18/us-amd-results-idUSBRE93H16E20130418

AMD’s Result Confirm Improving Operational Efficiency

http://www.trefis.com/stock/amd/art...m-improving-operational-efficiency/2013-04-19

Whereas Intel continue to feel pain, quarterly profits drop 25% for Q1 2013

http://www.tweaktown.com/news/29763...rterly-profits-drop-25-for-q1-2013/index.html

Look at Charles Kozierok post above, to learn how a proper argument is made, then take a deep breath while self-reflecting on all misinformated/semi-advertising trash you're spreading all over the place. At the moment you're just testing everyones patience.

 

[galego]

That motto is from 2011 and original BD launch. I guess they didn't use it back then but slides (one above included) leaked to the net.

I didn't know.

 

[AnandThenMan]

The future is fusion, right? Not discrete CPUs and discrete GPUs. If the future wasn't fusion then I'd expect them to have claimed as much.

Then why didn't you just say, the future of discrete CPUs/GPUs is coming to an end. Not that it actually is, there is room for both, but having more silicon dedicated to GPGPU functions is certainly something we will all benefit from in the traditional CPU perf/watt/price points.