What About Intel?

The King of CISC needs to come up with an adequate response to the RISC phenomenon

Dennis Allen

With all the talk about the benefits of RISC microprocessors, you have to wonder about the King of CISC, Intel. The RISC architecture has long promised lower cost and higher performance than the CISC architecture. What CISC has going for it, though, is Intel. Specifically, the Intel 80x86 family of CPUs has been the cornerstone of software compatibility.

But a growing number of analysts--including the editors of this magazine--believe that RISC is the future. Higher performance and lower costs count for a lot in this industry. So does compatibility. However, what if software compatibility weren't an issue? In other words, what if you could buy a system with, say, either a Pentium or a PowerPC and be able to run any Windows a pplications? That's the question you'll face with Windows NT running in native code on PowerPC systems.

Much of the compatibility question simply evaporates because Windows NT will run on PowerPC systems. The point is, if a system can run Windows, it doesn't matter if it's "Intel Inside"--especially if the RISC-based alternative delivers better performance.

That's the scenario using today's paradigm. Let's look beyond that. Apple is shifting its entire line of computers to incorporate the PowerPC, and as a result, the Apple platforms will take on more of the PReP (PowerPC Reference Platform) attributes. IBM will roll out its Power Personal systems later this year, and they will be extensions of PReP, too. We are also likely to see a good number of so-called clone makers roll out their PowerPC systems based on PReP later this year. Pretty soon, the standard of compatibility will be whether a system is PReP-based--not whether it runs Windows.

That has to be troubling to Intel. After the com pany has poured buckets of money into promoting "Intel Inside" as an assurance of compatibility, the whole compatibility premise as it has applied to Intel may become moot.

And it's not like Intel has made a lot of loyal PC manufacturing friends over the years, either. In Asia, for example, makers of systems and motherboards are anxiously awaiting an opportunity to end their dependency on Intel. Largely because of Intel's very unpopular chip-allocation policy, Asian PC makers have often found competing difficult because they could not get enough Intel CPUs. An added slap on the face for Asian PC makers was Intel's recent attempt to demand royalties on PCs that used Intel-compatible chips from AMD.

Things may be a little more friendly in the U.S., but the situation is still not good. IBM sells an incredible number of PCs, and IBM is leading the move to the PowerPC. Also, Apple's move to evolve its line into a higher-performance arena will bolster the success of the PowerPC.

These are formi dable challenges for Intel. We are on the verge of a mass exodus from the CISC architecture to the RISC architecture. Motorola made the move by getting on the RISC bandwagon with IBM and Apple as part of the PowerPC Consortium. Traditional workstation CPU makers made the commitment to RISC long ago with Alpha, Mips, and SPARC. That leaves Intel and compatible makers Cyrix and AMD as the remaining CISC players. Of those three, shifting gears into RISC will probably be easier for Cyrix and AMD because they have proven themselves to be excellent niche players in the CPU market.

For Intel, on the other hand, it's a different story. Intel has based its CPU business on all the compatibility baggage it has accumulated over the last decade or so. That wasn't the wrong thing to do; we demanded compatibility, and Intel, with the help of Microsoft, supplied it. But that was then, and this is now: The compatibility baggage just weighs too much. To get the performance gains we need in computing, we have to shed som e of that chip-level compatibility. Let the operating systems' microkernel take over some of that burden, and let the CPUs run faster.

That means that Intel has to find a way out of its 80x86 debacle. Sure, millions of 80x86 and Pentium systems will still be sold, but more and more of the market--and eventually, nearly all of it--will move to RISC. So far, Intel hasn't given any real hints as to what its response will be. The P54C--and, based on sketchy preliminary reports, the P6--are not an adequate response to the RISC phenomenon. Unless Intel mounts a more meaningful response, the King of CISC might become nothing more than king of the hill after all the other players move to another hill.