A lot of things change in a decade. At the end of the '90s, it seemed clear that Intel's Itanium was going to dominate the UNIX world and that the proprietary processors would give way to the inevitable. Itanium slipped badly and, while most of the other custom processors went by the wayside, IBM stuck with Power. Today, it announced Power7, a significant improvement on its platform and one that continues to differentiate it well at the high end.
At the other end of the spectrum, and effectively bracketing Intel, is ARM. Here, Apple recently announced its own version of this processor called the A4. This appears to be less of an advantage for Apple initially because, unlike where Power plays in the low-volume, high-cost space, ARM plays in the low-cost, high-volume space that would typically benefit a common provider.
Power7: The Sun Killer
Power7 is IBM's SPARC killer product and was well positioned to go after the UNIX opportunity that Sun lost largely through mismanagement. I could go on at length on the subject of mismanagement, but in the end IBM, by having solutions that could easily move on the opportunity created by Sun's collapse, was able to get more of the resulting benefit than HP was. Granted this wasn't just hardware. It included a tighter initial focus on the opportunity and the fact that IBM appeared to be co-resident in more Sun accounts because Sun had made massive inroads into IBM's accounts in the '90s when IBM was having issues. A lot of payback came to IBM this past decade.
In addition, even though most thought mainframes would collapse in the '80s, they actually went through somewhat of a resurgence because the platform had advantages in reliability and scalability in an ever-more-connected world, and proved to be reasonably successful as a consolidation platform during the past decade. It also turned out to be vastly easier to move older legacy mainframe applications to newer mainframes than to rewrite them. This, too, helped assure that this platform survived.
By focusing on energy-efficient performance for this latest class of processor, companies like the Australian Broadcasting Corp. have reported a 25 percent reduction in total cost of ownership, according to IBM numbers. Additional benefits include dropping the server footprint 35 percent and an 80 percent improvement in the time it takes to set up new customer environments. This all works because the space that Power7 occupies tends to be relatively custom and favors unique advantages in power savings and I/O, both of which IBM is expert at. Core to the success is that the kind of system IBM builds favors a low-volume unique part because of the unique nature of ever-larger enterprises that would be likely to favor these very large systems. In short, this is a market with only two very different power players right now: IBM with a vertically integrated solution and Intel, which largely relies on HP to build around its platform. The IBM approach seems to have been the more successful.
Apple A4: Keeping a Secret
The ARM space is made up of technology specialist companies that design chips and then share fabrication facilities to manufacture them, spreading out the cost. Intel doesn't yet play in this space in volume, though it hopes to enter it with a future version of the Atom processor. Each of the companies, which include Qualcomm, Marvell, Freescale and others, competes rather aggressively on price, performance and overall solution. In fact, the ARM space is less about processor than it is about the SOC (System On Chip) nature of the solution.
The products that play in this space tend to be subsidized by carrier charges, have incredibly thin margins, and compete on price as much as on user experience. The buyers tend to look at the overall solution and don't seem to care that much about the technology that surrounds it. In short, this is a market that would seem to favor the non-vertically integrated approach that appears to surround PCs and consumer appliances, making Apple's move to do its own processor seem foolish. However, while IBM's products are often heavily tested against their unique uses, Apple's products are largely defined by Apple itself, and the user experience defines the offering, not specific benchmarks. In addition, with Apple, the big problem has been leaks. The more people involved in a design, the more likely that design will leak out and be copied in China (working copies of the iPhone were available there within days of its first release). By doing its own part, it can substantially reduce the opportunity for a technology leak on a new product and have the option of switching over to a more common ARM part for future versions of a new product to contain costs as the product matures.
Apple will sacrifice some performance to gain better secrecy, but given that it is involved in a market that is less measurement oriented than IBM's, this is likely a good tradeoff. The iPhone, which was successful despite performance and carrier shortcomings, is proof of this. Granted the release of a potential iPad-killing iTablet in April for Europe (where Apple is weaker), based on Intel's new Atom part, likely will have us revisit this decision later in the year. Intel can keep a secret.
Wrapping Up: IBM and Apple-Similar Destination, Different Path
Both companies got to the conclusion of having a specialized part along different paths and both decisions can be justified competitively. IBM's decision is justified by the unique nature of the market that Power7 occupies, which favors very power-efficient, scalable platforms that can take on the massive virtualized loads of today and tomorrow. Apple's decision is justified by the need for absolute secrecy and the reality that a leak could turn an incredibly successful product into an also-ran. Both companies stand as testament to their ability to make unique decisions, which allow them to lead their respective industries even though those industries are vastly different.