Troubled microprocessor firm banks on embedded and heterogeneous computing shift as it sees PC market fade.
Two challenges immediately present themselves. First off, the design
cycles in the embedded space are longer than in the traditional
PC/consumer area and ASPs can be lower. Secondly, AMD's third targeted embedded market, gaming, left some scratching their heads.
"They already have Nintendo and rumors about them getting next Xbox
720 and PS4," said Patrick Moorehead, president and principal
analyst with Insights & Strategy (Austin, Texas). "They already
have graphics for Wii and Xbox 360. Maybe by gaming, they mean Las
Vegas gaming where AMD does a lot of business."
Of AMD's three turn-around keys, Moorehead sees embedded as the one with the least potential long-term impact, but also the fastest potential for a short-term revenue boost. The architectural push toward heteregenous
computing and the SeaMicro acquisition loom as bigger impacts for
the company, he argues.
"They've agreed on a new memory architecture which uses the GPU
(graphics processing unit) and CPU as equal citizens in getting
access to memory," Moorehead said. "Inside of the SoC it would make
the CPU a lot less important where Intel has a dramatic lead over
AMD and the ARM guys."
AMD has a leg up on graphics processing technology (through its 2006
acquisition of ATI), where Intel needs to "pay the Nvidia tax every
year," Moorehead said.
In September, the AMD SeaMicro
group announced Intel- and AMD-based systems extending
the link to up to 1,408 hard drives or five petabytes of external
storage using AMD's Freedom Fabric interconnect technology. SeaMicro
has gained some traction with large enterprise data centers
interested in the significant power savings the systems offer.
"I see Sea as a disruptive play in exa-scale data
centers," Moorehead said. "As traditional data centers move to these
football-field sized data centers, the industry moves to specialized servers from homogenous servers.
That's the Sea play."
AMD against the world
Strategy on a conference call is one thing; making it happen
another, as AMD executives acknowledge. "We do have to build our
execution credibility," Su said.
The company has lost some key executives in recent weeks, but
inside, there remains the OK Corral mentality that's marked the AMD
culture since the days of Jerry Sanders.
"There is still optimism at AMD, even through there are challenges,"
said Moorehead, himself an AMD alumnus. "People buy into those big
bets and kind of see where AMD is and where it could be. There are
people who still get excited about that stuff. You have to have a
different kind of personality to work at AMD."
The best predictor of future performance is past performance. Can anyone point to anything successful that the band of ex-IBM thieves has done that would predict success for AMD? I think that they've all left a trail of fail where they've gone.
As to embedded... let's just assume that we rule out network infrastructure for the moment and count the number of double-digit ASP sockets out there? Buehler? Anyone? How about development cycles? Time to revenue?
While blowing a distant-second position in the compute world, AMD is moving towards a zero-crossing - not just in profit (in which they have proven skills) but in revenue.
Sad. And a lot of good engineers are going to get crushed while the band of thieves get wealthy. Or more wealthy anyway. Hector was just the first raider of the AMD treasury.
Margins in the embedded space will be far lower. It would have been better for them to target Mobility space. But they probably realize that Investors may not be patient; If we remove Debt from the picture this may be an acquisition/merger candidate for either NVIDIA, Qualcomm or Broadcom or even Samsung;
I have been watching AMD over many many years. It appears that the number of years in which AMD lost money is more than the number of profitable years. I always wonder why the investors were willing to put up with the management of AMD. Now, we are probably seeing the end of AMD.
This still leaves one remaining and major question unresolved. We could all easily say "the PC is dead" but there's still the IT and replacement markets. So with AMD totally out of the PC CPU market, will Intel be able to still avoid being tagged as an uncompetitive monopolist, both here and in Europe? The left (represented by the current US administration) is usually the clear opponent of trusts, although the opposition will also be on the hot seat to rectify unfair competitive advantage, especially if the upcoming election results in a reversal of fortune. It should sure be interesting to watch this space for the next few months!
At EE Times we have seen many chip makers turn to the diverse and fragmented set of "embedded" markets when they could not make a go of it in the big markets such as PC or smartphones.
TI could not carve out a top tier position in mobile for Omap and so is taking it to embedded, i.e. all other markets.
Now AMD is not making enough headway in PCs and tablets so is turning to embedded, a collection of markets it has worked on for many years with marginal results.
But --O, oh-- most "embedded" markets don't need as much graphics umph as AMD has in its IP. What's more, AMD lacks the breadth of experience and contacts to drive its business in medical, industrial, aerospace, military, networking and the dozen other "embedded" markets.
Key issue: Are they going to make it until long term strategy pay's off?
BTW I do not buy there embedded story with APU, this is yet another new product looking for market, that will take 3 ~4 years to happen in volume, once again they do not have time.
Lets see they want to go after Comm's , Gaming, and Industrial.
Seems like Industrial is Embedded Single Board Computer, they have been there since they bought Cyrix from National whats new about Mkt.
Gaming : This yet another mature Mkt? Consoles are all SOC with Graphics IP on it, they have MSFT.
Comm's: This new Mkt where everyone who sells Embedded has presence, what are they going to do differently, for customers, that will make customer say forget Cavium and lets go with AMD? Lets ask Mr Cisco, forget all that custom ASIC, just buy my AMD APU, Let's see Mr Cisco has say about all this.
Seems like Chop shop approach will be needed sooner versus day dreaming, they are headed the CSR way, CSR cut there limb recently to survive Zoran acquisition, by selling there Combo radio with GPS to Samsung.