Future of the PC: Intel & AMD October 28, 2004 Jakub Wojnarowicz

The Past and Present

Page:: ( 1 / 3 )

A few years ago I wrote a State of the Industry mini-report, detailing where AMD, Intel, NVIDIA, 3dfx, Creative Labs and Aureal were going. Since then, two of those companies have ceased to exist and the scene has changed dramatically.

In the vein of Japanese car companies that started by providing low-cost alternatives to American cars, AMD cut itself a niche in the market in the same fashion. Its K6 line of processors were never the performers that Intel's Pentiums were, but AMD managed to milk them for all they were worth with 3DNow! SIMD extensions, and later an on-chip cache in the K6+3D. AMD subsisted on the K6 long enough to research, develop and put into production its Athlon.

The Athlon has been a breakthrough for AMD. For the first time in their history, they were able to deliver a superior platform ahead of Intel. Though the various incarnations of the Athlon have traded performance crowns back and forth with Intel's Pentium III and Pentium 4 line, there is no doubt of AMD's success.

Among us, the hardcore enthusiasts, AMD is the preferred choice at the moment. A recent poll on FiringSquad showed that 66% of all our readers who voted in the poll were running an Athlon processor, compared to only 34% who had an Intel chip. This is a huge swing in favor of AMD. Four years ago, in a similar poll the people who voted for AMD were slightly over 40% of our audience. Though this isn't a scientific poll, we can also look at the major computer OEMs and see that only Dell offers exclusively Intel hardware.

However, this article isn't about AMD's advances in the hardware arena. It's about the consequences of those advances.

AMD was the first of the two companies to announce, develop, and release a processor with 64-bit extensions to the x86 family. The company has also beaten Intel to the punch with an on-chip memory controller and rumors first started of Athlon 64 dual-core before any dual-core Intel chips. The Athlon's success has spawned a huge support infrastructure for the processor. Big questions upon its launch were "who will make chipsets and motherboards to support it?" We can all remember the first poor motherboards, but eventually the proper hardware materialized.

Intel's 64-bit processor shares most, but not all of AMD's 64-bit extensions. Intel's motherboards are now completely incompatible with AMD's, unlike in days past when a K6 and Pentium could share the same system. A few years ago, Intel experimented with Rambus RDRAM - ostensibly for greater performance but also because it would give Intel something AMD didn't have.

Forces in Action

Page:: ( 2 / 3 )

Remember when Itanium was first announced? Articles about the future Intel processor and its architecture - IA-64, also known as EPIC, mushroomed overnight. Most of these explained that the desktop wasn't going to see IA-64 any time soon, that it was reserved for the far future and that we'd be using IA-32 for a long time. Yet, IA-64 and Itanium were clearly destined for the PC sooner or later. If things had gone according to plan, there would be no 64-bit desktops right now and Itanium would be a successful server processor. Years down the road, the completely proprietary technology would eventually move down to the PC, shutting AMD out. Currently AMD has a license agreement with Intel stretching back to the late 70s which permits AMD to make x86 processors. With EPIC, Intel would be free of the competition.

That's not to say that if AMD was in Intel's position, it wouldn't have tried to do something similar. No, the above example illustrates one of the two major forces driving both companies right now.

The incredibly fierce competition between the two has seen AMD claw its way up into the high-end segment and the low-end server market, areas formerly owned completely by Intel. Athlon FX processors are just as expensive as Intel's Pentium 4 Extreme Editions. Opterons compete head-to-head with Xeons. Though Intel dominates market share and sales, it no longer has the exclusive ability to rely on its high-end hardware to beef up its margins and give it a marketing edge.

AMD is now approaching the point where it can directly challenge Intel and wrest the future of x86 development from the bigger company. Arguably, it has already forced Intel onto the 64-bit x86 path.

This illustrates the second force acting on both companies.

You see, the first force is divergence. Both companies want completely proprietary platforms that they can control and own in full. This, they hope, will permit a virtual monopoly on the desktop similar to what Intel enjoyed in the late 80s and early 90s. Remember the days when a top-of-the-line Intel desktop chip cost $1000 simply because it was top-of-the-line? It didn't have special hardware like an Athlon FX or P4EE, it wasn't limited in availability, it simply cost $1000 because Intel said it cost that much.

The cut-throat competition between AMD and Intel is leading both to try to grab for that kind of market security.

On the other hand, neither company can afford to be left behind on the features front. Intel has SSE3? AMD must have it. AMD has 64-bit extensions? Intel must have it. This converging force is, at the moment, stronger than divergence. Neither company is in position to drive the market in a completely different, proprietary direction. Perhaps both are working on such a solution, perhaps not. For all we know, each has something ready and is just waiting for the right time.

What I can guarantee is that sooner or later, one of the two (or even a third platform), will win.

The Future

Page:: ( 3 / 3 )

At the very least, in the long run x86 will finally have to be abandoned. x86 has been hacked, mangled, sliced, diced, re-packaged, and ultimately trivialized as a universal CISC access layer for two wildly different RISC solutions underneath. Everything since - and including - Intel's original 32-bit extensions to the x86 instruction set has been a hack job. High quality hack jobs, but these modifications are add-ons to an ancient and inferior computing system. Sooner or later, no matter how trivialized x86 becomes, its limitations will grow to be such a burden that both companies will be forced to develop non-x86, and thus very likely proprietary, solutions. Once that happens, the market will decide which it likes more and the competitor will choke and die like Beta vs. VHS. Of course, for all we know, IBM or Transmeta or any other company might come up with a completely different solution that catches on while both AMD and Intel are fixated on x86.

More likely, AMD or Intel will develop their own proprietary instruction set, and deliver acceptable x86 performance at the same time. Itanium tried to do this, but its x86 performance is, by all accounts, mediocre. AMD's 64-bit extensions meanwhile, are still simply an addition to x86, rather than a new instruction set that AMD owns the rights to.

When will this happen? I'm not dumb enough to hypothesize and be immortalized with a quote like "640k should be enough for everybody", but I will say that multi-core processors are certainly a major step in that direction. How long before either company figures out how to pack its x86 core - be it Athlon or Pentium - along with another chip with the unique instructions? From there AMD or Intel would have to ensure that the infrastructure is in place to support it.

Though the market is certainly aware of the consequences of switching to a proprietary system, if the performance boost on that system is significant enough and the package remains affordable, it will happen. From then on, it will be a processor generation or two before the leading company simply kills off x86 support, citing unnecessary costs and low usage statistics.

That's my hot air for the week. Drop by our forums and comment on it.