Who Needs 64-Bit CPUs?

 
 
By Peter Coffee  |  Posted 2003-10-06 Email Print this article Print
 
 
 
 
 
 
 

New crop of PCs offer major improvements.

As Advanced Micro Devices Inc. rolls out its much-anticipated Athlon 64 desktop processors, enterprise system buyers should avoid the tunnel-vision error of looking only at the 64-bit capabilities of these and other next-generation PC chips. This begs the question, "Who needs 64 bits?"—and thats the wrong question to ask.

When people use the size of the data word, or the size of the system address space, as the measure of a machine, theyre seeking a single, simple measure of capability. Thats the same urge that has led to overemphasis on processor clock rate, for example, as a single number that fails to fully represent what a machine offers.

The size of the digital word, like the number of clock cycles per second, is an input to a task, not an output. Extending either attribute past its previous limits enables new capabilities but also increases the cost of every computer subsystem. Proponents of 64-bit CPUs must answer, instead, the rightly skeptical buyers larger question: "Whats in it for me?"

The answer is that the year-end crop of 64-bit PCs—Apple Computer Inc.s already-shipping G5 machines and the systems soon to be built around AMDs Athlon 64 CPUs—merely begin with a greatly expanded address space. They build on that foundation with other technologies, notably major improvements in memory use and I/O capability, to make that expansion much more useful.

AMDs label of "cinematic computing" is an appropriate description of the resulting leap in overall processing power.

At the risk of raining on a few parades, it should be noted that 64-bit processors have been available for years. In the hoopla of claims such as "first 64-bit personal computer," its easy to forget that a trial crop of single-user 64-bit desktop systems were built several years ago by several vendors using Digital Equipment Corp.s Alpha CPU—back when Microsoft Corp.s Windows NT was briefly being positioned as an operating system for more than just x86 machines. Market reaction was tepid, as it has also been to personal systems using Intel Corp.s 64-bit Itanium. Proponents of both the G5 and the Athlon 64 must therefore be prepared to answer the additional question: "Why now?"

Both Alpha and Itanium committed a common yet cardinal sin: Their blank-paper designs faced prospective buyers with costly new hardware that reduced, rather than increased, the value of the x86-compatible software (as well as associated tools and skills) that they had acquired. Alpha failed to address this issue with its binary translation technology, which was technically impressive but cumbersome and unexciting in practice; Itaniums initial promise of a smooth transition from the x86 software base looked better on paper than in silicon.

Buyers were also unimpressed by the argument that future software, written for these brave, new architectures without the burden of backward compatibility, would give them much-improved performance. Buyers knew that if they refused to bite, the hardware engineers would grumblingly go back to their labs and obey Moores Law, wringing more speed from binary-compatible designs.

This gave buyers a more predictable path to higher performance, in a time frame no worse than likely to be needed for a more risky software replacement.

Apple faces only trivial barriers in moving to 64 bits, since this capability was built into its PowerPC architecture from the beginning. AMD had a tougher row to hoe in extending the x86 instruction set and architecture to 64-bit capability, but the result gives buyers the option on which they insist: the ability to choose their own balance between preserving old software and exploiting new capabilities.

In its "legacy" mode, an AMD64 CPU will run 16-bit and 32-bit x86 code as readily as an Intel Pentium-class chip does today. In "long" mode, an AMD64 processor with a 64-bit operating system can run 32-bit processes in separate, 4GB spaces. It can also run 64-bit applications in a vastly expanded address space and can give them twice as many on-chip, general-purpose registers for holding data and intermediate results.

Combined with drastically improved on-chip hardware for predicting software branches, based on past behavior, the result is that an AMD64 processor doesnt just have more power; it also has better handling and braking, putting its resources into useful work instead of wasteful recovery from over-exuberant errors.

The argument that capability creates its own demand was once invoked to justify building the Concorde supersonic airliner—a design thats soon to be retired, having failed to make economic sense. Will users quickly find compelling value in 64-bit designs? Or will the 32-bit CPU prove to be the Boeing 747 of computer architectures, with at least another decade of life in the PC mainstream?

Given the rapid expansion of digital media, data mining, and immersive simulation and entertainment applications, it seems the better bet is that software developers will find things to do with the 64-bit expansion of address space and usable memory and that system buyers, when not forced to give up their current software as the price of transition, will move with alacrity to 64-bit platforms.

64-bit PCs: Why now?

  • Previous 64-bit desktop offerings imposed unacceptable software costs

  • Moores Law hardware improvement rates matched or outpaced promised benefits of 64-bit software conversion

  • New 64-bit processors (Athlon 64, Apple G5) provide smooth software transition

  • Advanced memory and I/O enhancements offer broad, balanced growth of system performance

    Technology Editor Peter Coffee can be reached at peter_coffee@ziffdavis.com.

  •  
     
     
     
    Peter Coffee is Director of Platform Research at salesforce.com, where he serves as a liaison with the developer community to define the opportunity and clarify developers' technical requirements on the company's evolving Apex Platform. Peter previously spent 18 years with eWEEK (formerly PC Week), the national news magazine of enterprise technology practice, where he reviewed software development tools and methods and wrote regular columns on emerging technologies and professional community issues.Before he began writing full-time in 1989, Peter spent eleven years in technical and management positions at Exxon and The Aerospace Corporation, including management of the latter company's first desktop computing planning team and applied research in applications of artificial intelligence techniques. He holds an engineering degree from MIT and an MBA from Pepperdine University, he has held teaching appointments in computer science, business analytics and information systems management at Pepperdine, UCLA, and Chapman College.
     
     
     
     
     
     
     

    Submit a Comment

    Loading Comments...

     
    Manage your Newsletters: Login   Register My Newsletters























     
     
     
     
     
     
     
     
     
     
     
    Rocket Fuel