Stay in touch…

Blog

Read the latest Bitstream

RSS Feed

LinkedIn

Look for us at LinkedIn

Twitter

Follow us on Twitter

Mix Magazine

This installment of The Bitstream column appeared in the September 2003 issue of Mix Magazine.

The Bitstream

This column discusses the introduction of 64 bit CPUs for desktop use…

Man, And I Thought The Salsa Was Hot!

Chips…they’re everywhere; In your toaster, in your car (quite a few, in fact) and, of course, in your DAW. This month’s Bitstream is the first installment of a two part series looking at the age old PPC versus Intel debate and also the new 64 bit crop of CPUs, which will have a significant impact on host–based DAWs.

To begin our journey, we must go…Forward, Into The Past! Yes, cast your mind back, back to last March, when the Bitstream discussed single versus double precision DSP processing. If you remember, run–of–the–mill DSPs typically store numbers as a floating point, 24 bit mantissa and an 8 bit exponent. Fancy pants DAWs double up the data and perform double precision calculations, with a 48 bit mantissa and an 16 bit exponent. This larger number of significant digits allows more signal processing to be performed before rounding errors are introduced.

OK, fair enough but, to generalize, most any arithmetic process is more precise with double the word length and this includes the general purpose CPUs that are the soul of Win and Mac computers. In the Windows world, those general purpose Intel CPUs we take for granted started life as their 4004, introduced in 1971. The 4 bit 4004 wasn’t much fun but it’s children, the 8 bit 8008 introduced the following year and the 8080 in 1974, paved the way for the 8086, which arguably spawned the whole personal computer industry.

Intel 8086-8088 CPU
Intel’s 8086-8088 Central Processing Unit, circa 1978

The 8086 ran its 16 bit words at a blazing 5 MHz, could address a fat 1 MB of memory and formed the computing core of IBM’s first PC. MITS’ Altair, the first personal computer, used a 8080 and came to market in 1975 for $395. Notice the trend: 4, 8 then 16 bit computers. By the mid ’80s, Intel had a 32 bit processor, the 386.

By the early ’90s, Intel was working on the first Pentium and users of the other common flavor of microprocessors, Motorola’s 68000 series, had realized that they also needed a next generation shot in the arm to compete. In 1991, Motorola, IBM and Apple Computer formed the PowerPC (PPC) Alliance and so was born the PPC’s RISCy yin to Intel’s CISCy yang. Though the first product of the PPC Alliance, the 601, was a 32 bit microprocessor, it had both 32 and 64 bit registers and a 64 bit FPU (Floating Point Unit). Four years later, the PPC Alliance released the second generation 620. With 64 bit internal data paths and 32 bit I/O, it was the first, though partial, 64 bit implementation of their PowerPC architecture and set the stage for things to come. In a very smart move, the PPC family was designed as a 64 bit cruncher with backwards binary compatibility for 32 bit applications. It would take a while for Intel to learn the backward compatibility lesson but, with the ability to chop their spectacular profit margins and still make a tidy penny on their CPUs, Intel countered the PPC threat with lower prices while also starting on a dizzying spiral of ever increasing clock rates to compensate for inherent CISC performance limitations.

A big physical difference between the PPC and Pentium is the PPC’s Reduced Instruction Set Computing character. Relative to CISC or Complex Instruction Set Computers like the original Pentium, RISC architecture, with fewer instructions baked into silicon, sacrifices complexity for increased speed - the lean and mean approach. Another factor, according to IBM, was that “…the POWER Architecture was unique among the (six) existing RISC architectures in that it was functionally partitioned, separating the functions of program flow control, fixed-point computation, and floating-point computation.” The architecture’s partitioning facilitated the implementation of a “superscalar design,” a now ubiquitous feature, which makes it possible to execute multiple instructions during a single clock cycle. Intel, while clinging to their x86 architecture, has their own superscalar family, the i960, which is used for embedded application, not general purpose computers like the Pentium. Not to be outdone by the competition, the Pentium has increasingly cribbed pages from its i960 brother, adding a RISC core with CISC trimmings to make it backwards compatible with older x86 processors.

In the past decade however, there is one metric that CPU powerhouse Intel has emphasized in their saturation bombing ad campaign. Clock speed is an easy measurement for your average punter to understand — faster is better! — so Intel has latched onto that dimension as the defining selling point for their products. Unfortunately, clock speed alone is a spurious measure of real world performance and can only get you so far, since architectural decisions by the Intel design team has forced them to ramp their clock to the limits of heat dissipation in an effort to improve execution. Indeed, current draw and subsequent heat generation is the limiting factor in the go–fast world of Intel. It’s not uncommon to find all sorts of exotic cooling methods used by manufacturers and hobbyists in their attempt to keep their Pentiums from doing a Three Mile Island on the motherboard: heat pipes, active refrigeration (both solid state Peltier heat pumps and old school compressed gas versions), even water cooling has been pressed into service to chill those big, power hungry chips.

NEC VX9006F
NEC’s quiet, cool VX9006F ValueSTAR, the world’s first desktop personal computer with factory–equipped water cooling

When asked about the current range of 32 bit Intel processors and how they hold up in the real world to current PowerPCs, digital audio geek James A. Moorer of Adobe Systems doesn’t have any qualms about stating that , “…the PPC is light-years better and faster than any of the others. The P4 (Pentium 4) is pretty close, but still quite a ways off. The big thing is that the PPC chip has 64 bit data paths, and all the Pentium chips are 32 bit data paths only. That makes the PPC chip more than twice as fast - it yields more than twice the data rate. Every benchmark I have done comes out with a difference, on my personal benchmark tests, of a factor of 4x over the fastest single-processor Intel processor I could find. My benchmark is pure double-precision floating point crunch-power with big sprawling data structures. My FFT routine (the Fast Fourier Transform is the basic math trick that powers all audio DAWs - OM) generally excites the worst-case behavior of most processor’s address prediction hardware.”

Address prediction…the whole PPC vs Intel debate, in a way, boils down to prediction and how designers go about auguring upcoming processing requests. However, that’ll have to wait ’till next month, now that we’ve set the stage for the discussion of a feature that mainframe, scientific workstation and Apple customers have long enjoyed; 64 bit and longer data words. Longer is better for “sprawling data structures,” and such big league computing is starting to make an impact on the Windows desktop as well.

Next month, we’ll delve further into the Intel versus AMD and Itanium versus Opteron tussle while providing some perspective with a look at IBM’s 970, the fifth generation PowerPC destined for Big Blue’s servers and new PowerMacs from Apple.

Bio

OMas just came back from speaking at this year’s TapeOpCon 2003, where he suppressed the urge to vent his spleen too vigorously. All the while, this month’s Bitstream was written while under the influence of the Yellow Bird and the camp blacksploitation of super bad and beautiful Cleopatra Jones. Drop by www.seneschal.net for deeper threads and illustrations concerning this month’s topic.