64-Bit Computing Gets More Interesting, More Complicated

Last week, at the Intel Developer Forum (IDF) Spring 2004 in San Francisco, microprocessor giant Intel revealed what CEO Craig Barrett called the "worst kept secret" in the IT industry: Intel is now mounting a two-pronged attack on 64-bit computing and will complement its Itanium product family with a quickly growing line of 32-bit chips that will include 64-bit extensions. This move basically means that Intel will retroactively adopt the strategy innovated by its competitor AMD and add 64-bit capabilities to its popular 32-bit server, workstation, and, eventually, desktop chips. For businesses starting to hit the limits of 32-bit computing--especially the 4GB memory constraint--Intel's move is good news. But the change isn't going to happen overnight, and it raises some serious concerns about compatibility, performance, and choice. Let's look at what's happening and examine how Intel's plans will affect your hardware purchases over

Intel's original Itanium wasn't the first 64-bit platform that Microsoft supported, but it was the first one that lasted any appreciable amount of time. Today's Itanium 2 products are backed by workstation and server versions of Windows, as well as by other OSs, including Linux. The Itanium is a "from scratch" 64-bit microprocessor, an original design that Intel co-developed with hardware giant HP in a bid to spearhead enterprise-class 64-bit computing. From a technological standpoint, the Itanium was the right idea: Break from the technical limits of the past and start fresh, addressing the unique concerns (e.g., increased memory space and bandwidth) that 64-bit computing enables. From a marketing perspective, however, the Itanium has proven to be something of a disaster for Intel, which typically sees its products sell in mass-market numbers. Incompatible with 32-bit software, except through a slow-performing emulation environment, the Itanium hasn't sold well. Further hobbling its growth is the megahertz myth that Intel helped to propagate as it moved its desktop chips to ever-faster speeds. Because Itanium chips run at much slower clock speeds than its Xeon and Pentium 4 chips, many customers perceive that the Itanium chips perform poorly as well, which isn't the case.

Meanwhile, one of Intel's few remaining competitors was working on an ingenious idea: Instead of wasting the time and expense of developing a completely new and incompatible microprocessor platform, tiny AMD decided to bring customers the best of both worlds and started a project, code-named Sledgehammer, that would bind a set of 64-bit extensions to a 32-bit Intel x86-compatible line of processors. The new processors would be backward-compatible with the Intel Pentium 4 family of chips but also offer the benefits of 64-bit computing. Released in mid-2003 as the Opteron line and followed by a line of desktop-compatible chips named Athlon 64, these innovative server-oriented processors were not only fully compatible with existing systems, they were also much cheaper than Itanium-based systems. Collectively, these chips make up the AMD64 platform.

Regardless, AMD's solution would have been dead on arrival if Microsoft hadn't cooperated by announcing plans to support AMD64 with special 64-bit versions of Windows Server 2003 and Windows XP. Various Linux makers, including Sun Microsystems, also support the chips. Today, customers can run 32-bit Windows versions on these systems or run a beta version of the 64-bit OSs, which, according to Microsoft, will be finalized by late 2004.

Back at Intel, the company was still pushing the Itanium 2 to a reluctant market, but it was also secretly pursuing a way to add 64-bit capabilities to its high-volume 32-bit processors. Last week, the company officially unveiled these plans, and in a stunning moment of humility, the company announced that it plans to follow AMD's lead by offering 64-bit extensions to its chips, extensions that will be 100 percent compatible with AMD's offerings. That move means that Intel's upcoming hybrid chips will run the same 64-bit Windows and Linux versions now supporting the AMD64 chips, a boon to customers, who will now be able to make purchase decisions based on price and performance rather than on compatibility.

Intel will phase in the chips over time. By mid-2004, Intel will issue next-generation, dual-processor-capable Xeon chips with 64-bit extensions aimed at server and workstation markets. These chips will compete with the AMD Opteron family of processors. By late-2004, Intel will ship 64-bit Pentium 4 chips, aimed at the workstation and desktop markets and targeting AMD Athlon 64 processors. In early 2005, Intel will ship 4-processor-capable Xeon versions. The company will continue to develop and market the Itanium family for high-end server needs, and it recently released a software update that lets 32-bit applications run faster on those systems, although those applications will never run as quickly on Itanium as native applications do.

From what I can tell, the Itanium will always have a place, but the new 64-bit-compatible Xeon and Pentium 4 chips will be mass-market successes that will bring 64-bit computing to a much larger audience than would previously have been possible, although the availability of 64-bit applications, servers, and services will be key. In some ways, I feel for AMD, which will likely suffer the innovator's dilemma and see its market-leading moves rewarded with a thorough beating at the hands of Intel. That's too bad, because AMD had the right idea. Although adding 64-bit capabilities to an aging 32-bit platform might seem kludgy, I've always advocated customer needs over technical excellence; heck, that explains the success of the wider PC industry better than just about any other argument.

Looking ahead to 64-bit computing, you should be aware of a few factors that can affect any rollout strategy. First is the aforementioned application compatibility problem. Second is cost: If the 64-bit-compatible Xeon and Pentium 4 systems are price-compatible with today's chips, buying them as they come out will be an interesting way to future-proof your purchases and make them compatible with tomorrow's 64-bit applications if the need arises. Third, you'll now have a choice when selecting your next-generation processors. With both Intel and AMD plying compatible wares, enterprises will soon have the same type of platform choice to make that consumers now enjoy. I don't have any experience with AMD's server chips, but its desktop chips have consistently outperformed Intel's at the same clock speed, leading AMD to market its products against those Intel chips that the AMD chips outperform, and not by actual clock speed. For example, the Athlon 64 3400+ performs at the level of a 3.4GHz Intel Pentium 4, a chip that doesn't even exist yet. If the competitive nature of the desktop market moves into the workstation and server markets, we're all going to benefit. And really, that's the most exciting aspect of Intel's recent 64-bit moves.