This would be a recurring theme with subsequent sockets and CPUs from AMD after that point. Those who bought AM2 systems could later make the jump to Phenom X4 and join in on the quad-core bandwagon. Even AM3 processors whose claim to fame was support for the new DDR3 standard could be retrofitted to AM2+ systems thanks to the retained support for DDR2 memory. That means a single motherboard purchased in 2007 could run the gamut from single-core Semprons all the way up to six-core Phenom IIs. It was all simply contingent on a BIOS update. That's what you call giving your customers value.
I tried to make a similar transition to hexa-core recently, but found that DFI in their absconding of the enthusiast consumer market left much of their customers between 2009 and 2010 up the creek. Despite producing high-quality boards that were more than capable of it, they neglected to add X6 support to the BIOS for all but the lucky few that owned AM3 boards. I had 8GB of DDR2 memory that should have had plenty of mileage left, but must now be chucked along with the motherboard because I didn't buy from the right brand. That really sucks.
But you know what, I'm not really alone anymore. The fact is that anyone who bought an AMD platform within the last year or so is feeling the same hurt, but not because a few board manufacturers are irresponsible. It's because the CPU and chipset maker that started the idea of unified platforms with long legs has since lost that edge. With the release of Bulldozer, suddenly the vast majority of AM3 owners find themselves with an obsolete socket. In fact it's due to the diligence of a couple of highly regarded board makers -- Asus and MSI -- that anyone has a chance at extending the life of their PCs with a Zambezi processor, despite not having AM3+. And do I need to mention the pissed off people that bought into Socket FM1?
AMD is losing their long-held advantages, and it only had a few to begin with. When I first wrote about the apparent breakdown of the company's sanity, I mostly talked about GPUs. But let's face it: GPUs are the least of their worries. They may have lost their GPU product planner, but the effects of that won't be felt until 2014, and it may be moot by then anyway. Their real problem is Rory Read.
Late last year news was circulating of a 10% hatcheting of their workforce, mostly involving PR and marketing people. The idea behind it was that a) it was a cost-cutting measure, and b) it was the first step towards AMD's big new strategy. Most people were more than a little concerned (it wasn't the only bad press AMD was generating at the time), and the only insight we'd get into their thought-process would come shortly after New Years via their Financial Analysts Day conference. That's pretty much where my last rambling left off.
Well that press event came and went a long time ago, and we're not feeling very reassured right now. In it they harped on a lot of the same taglines and marketing speak we've heard from them for several years now, such as heterogeneous computing and whatnot. However, we did hear more emphasis on mobile and SoC development. The new AMD is to focus on flexible, reusable designs, time-to-market, increased agility and ODM/OEM partnerships that offer tailored solutions. And we heard about the console market as an example of the latter. Here's a good slide showing how they see the changes that are being made:
I like some of what I'm hearing. I like that they acknowledge mistakes made in creating complex designs that had issues executing, lagging process technology, and trying to tackle the performance problem with more cores. Again, the new AMD is supposed to focus on lower-power designs, flexible SoCs that can be brought to market quickly, and getting partnerships with companies so they can make an impact in datacenters and HPC markets with heterogeneous computing. The problem is they have some serious set-backs so far delivering on those promises.
For starters, they've lost a lot of their core talent. A LOT. These aren't just superfluous pencil-pushers either. As this article points out, they are the people responsible for some of AMD's strongest offerings, particularly their APUs. Worse, their headliner product for pushing into the low-power and mobile market, Brazos, has lost its important drivers -- the men that designed and produced the thing. Guys such as Jim Mergard who just made the news recently with his move to Apple; and Brag Burgess who left AMD around the same time and headed the Bobcat architecture.
One result of this is that in the nearly two years since Bobcat debuted, there has not been a significant revamp in the design, or even a process shrink. We're finally getting Hondo by the end of this year, the long-awaited "Brazos 2.0", but disappointingly it will not be a change in architecture, and no partners have been announced alongside it. Bobcat, while a good chip, had a similar problem of winning designs, and it seems like Hondo's hopes of getting into Windows 8 tablets against Atom are slim to the point of anorexia.
AMD can't afford to lose any more of its idea men, but in a shocking new rumor it seems like the company is planning layoffs that make the last one seem minuscule. Numbers being floated around suggest that 20-30% of their engineering and sales people may be filing for unemployment soon -- people that the company probably shouldn't be losing. The new CEO would like you to believe that cuts like these are a result of revenue loss and a continuing march towards a slimmer and more dexterous company, but stockholders (those dumb enough to still own some) are finding that explanation disagreeable. Share prices are falling like shell casings at a redneck family reunion, and if this is part of some grand new strategy for the company, the mastermind behind it might find himself shuffling for a new job before his plan has a chance at doing more damage.
AMD doesn't have much of a grip left on the desktop market, and if the new Trinity launch fiasco didn't hurt their image enough, the mediocre Piledriver launch should do plenty to further marginalize their presence in the minds of systems builders. By the time Steamroller comes out, they'll be up against Intel's aggressive new architecture, leaving them in the dust regarding performance and power efficiency. Hondo won't likely gain them share of the tablet market, Haswell will take what remains of the desktop and laptop market, and AMD will have to scale back their operation with all the reductions they incurred in their workforce. Their recent re-hiring of K8 designer Jim Keller is small consolation there.
Their only hope is to score some major wins in the next console race. Undoubtedly they'll be pushing a whole platform solution per their new ODM partnership approach, and an AMD CPU in a next-generation game system might actually help get games optimized for their architecture. PC gamers often lead the charge for high performance adoption, and scoring with them might help energize their standings on the desktop. Short of that, they might have to take more and more drastic cuts to their size in order to stay profitable, and that means Intel will run rampant without a competitor. We're already seeing what problems arise when that happens.
No comments:
Post a Comment