Heavy Artillery

When the Merced, seventh generation microprocessor from Intel and Hewlett-Packard is launched in 1999, it will not just take a piece of the action. It will be THE action

The average age of a microprocessor design is between eight and 14 years. Within that period, the manufacturers would have squeezed every megahertz out of the silicon. Our current favourite, which is used by over 90 per cent of the world, is the aging X86 architecture to which belongs the Pentium, MMX, PIIs and the 486s and 386s of an age gone by. Who do you think Intel, Andy Groves paranoid child, would turn to?

 

To another giant one that is perhaps even more paranoid that it hasnt made a loss in the last fifty-two years. A company that steadfastly refuses to take any risks ever and whose corporate culture does not include information technology religions and whose in-house joke, which its India chief repeats with relish, goes that If I were asked to market Sushi, Id market it as cold-dead-fish. Thats the truth.

Apparently, Grove liked the joke.

The Hewlett-Packard-Intel alliance recently announced that the next processor from Intel would be based on a new technology called EPIC (Explicitly Parallel Instruction set Computing). It has long been accepted that parallel computing can provide great performance, and the Merced processor promises to make that affordable.

EPIC is the next generation technology, a kind of a generic philosophy or collection of techniques, for example, like RISC or CISC. Its an instruction set technology. IA-64 is an actual instruction set, as for example, IA-32 or PA RISC.

Now, getting more concrete, Merced is a code name for first processor thats going to implement this IA-64 instruction set, as, for example, the Pentium II processor or the PA 8500. It is scheduled for production in 1999. The processor will be produced on Intels 0.18 micron process technology, which is currently under development. The smaller micron process would mean that Intel would be able to wring more processor out of each silicon wafer, thus making for cheaper processors.

The Merced processor will extend the Intel architecture with new levels of performance and features for servers and work- stations.

Intel had its expertise with the volume market and its needs. HP on the other hand brought enterprise system technology. HP had a lot of activity with PA-RISC and did a really excellent job of pushing performance ahead in high-end systems. So they brought that expertise to the partnership.

But it still is not goodbye for the 32-bit x86 architecture. The new architecture, dubbed IA-64, will be backward compatible with the existing x86 architecture, just as the 32-bit architecture of the 386 was compatible with the 16-bit 286, 8086, and 8088.

Lets explain that. The Merced microprocessor will run all existing software for the Intel platform without any change, declares Stacy Plemmons, VP Computer Systems, HP India. That is complete backward compatibility, and much better than what Apple could do when it effected the changeover to PowerPC. Or what Sun Microsystems could do when it came out with a newer version of Solaris, its server operating system.

HP, the company most known in India for its overwhelming marketshare in printers, is also a microprocessor manufacturer, and it has a successful line of microprocessors in the PA-RISC range. The company realised that while its RISC processors were evolving and were the market leaders in the RISC/UNIX market, that was simply not enough for the requirements of the enterprise servers of tomorrow.

Meanwhile, Intels x86 range was getting a little long in the tooth. The company had to leapfrog its competitors RISC range by introducing a processor that was miles ahead, and at the same time could maintain backward compatibility with existing software written for the 32-bit processors.

HP, true to its reputation for not taking risks, started looking for a partner who had lots of cash to sink in a new microprocessor fab. A new microprocessor fabrication facility is an expensive proposition, and HP didnt want to risk a few billions in it.

Partnering with Intel would mean that the chip should be a next generation processor for both the companies existing range. Intel, with a stranglehold on over 90 per cent of the world microprocessor market, would be able to manufacture the chip in high volumes that would make it much cheaper. In fact, if HP had gone it alone and developed the processor, the cost would be so prohibitive that nobody would buy those systems.

The arrangement with Intel provides HP with a smooth upgrade path to 64-bit microprocessors. It would also rid HP of potentially unprofitable processor manufacturing in the future,considering Intels growing clout. Intel would sell the processors to HP at the same rates as it would to others, but the low price would enable HP to club several of the microprocessors in a single high-end server or workstation for mission critical operations, while Intel would try and market it into everything from high -end servers to desktops to embedded consumer equipment.

The Intel-Architecture 64-bit Merced processor will offer as much as 15 times the performance of current processors while maintaining compatibility with existing code, says Stacy Plemmons, VP, Computer Systems, HP India. And the code compatibility extends to the HP software that currently runs on HP processors.

Although RISC processors from HP competitors will improve at a 50-percent annual rate, HP expects a 500-percent to 1500-percent improvement in performance from IA-64.

This is a quantum shift in processor technology, says Vikram Mehta, Director Enterprise systems and marketing, Asia-Pacific, HP. And we have achieved the architecture shift much better than anybody has ever done. Merced will contain an Object Code Translator that will translate PA_RISC instructions into something that the IA-64 instruction set can understand.

But such paradigm shifts -- or in the words of industry observer John Berst, inflection points -- are fraught with danger. Customers are concerned less with technology than smooth day-to-day operations, and putting them through a difficult transition phase is a recipe for disaster. It is to both HPs and Intels credits that they have made the transition path almost unnoticeable.

The implications of Merced are many for the IT industry. Not ever before has any technology received such overwhelming industry acceptance before the Merced, says Mehta. Every IT major has announced that it will support Merced, and they include even Intels and HPs competitors such as Sun Microsystems, Digital and SCO.

The wider acceptance of Merced also means that it will further enhance the strength of UNIX as an OS for high-end applications, says Mehta. Contrary to the widespread belief that the Netware and UNIX marketshares are shrinking, he says that all operating systems have grown in the past year even though NT has grown more than others. A 64 bit processor such as the Merced will flex its muscles even more on a 64-bit operating system, and this is where Microsoft is lagging behind. The company has not yet committed itself to a 64-bit OS, while Sun will have a 64-bit version of Solaris when Merced is launched. Digital will do the same, and HP and SCO together will launch 3DA UNIX that will take advantage of Merced power. It could be said safely that while the last decade saw a battle between microprocessor platforms, the coming decade will see a battle of different operating systems on a single platform.

The flocking of competitors to Intel should be quite confusing for many, but the fact is that even competitors realise that they need to offer a choice to their customers. So a reluctant Sun has to offer a version of its OS to run on Merced, while its own hardware division grumbles. The cheaper prices of Merced based systems is likely to hit Suns high-end hardware sales, but the company is left with no choice at all.

Digital, which already has its 64-bit Alpha processor, is going nowhere with it, and is widely expected to shelve Alpha when Merced arrives. The company has already announced its tie-up with Sequent Systems to develop its 64-bit UNIX for Merced. Once that happens, and given the fact that Intel now controls Alpha production, its is unlikely the the Merced will face any major threat.

For both contendors to high-performance crown, in the long term, it is either sink or swim with Merced. The demand for their operating systems will dwindle if they dont offer a choice on Merced too, but that will adversely affect their hardware business. Damned if you do, damned if you dont.

Merced will also considerably firm up Intels position on the desktop. Examine Intels history, and you know that despite their denials, Intel has introduced new processors at the high-end, and as soon as they could be manufactured in volume, moved them to the desktop. They fool no one when when they say that the new processors are for servers amd workstations. Meanwhile, Intels Pentium clone competitors such as AMD and Cyrix do not seem to have any coherent strategy to face the challenge.

But some industry observers have been predicting that growth in the next century for microprocessors will be more in consumer appliances than on the desktop, and Intel would meet its nemesis here. It is difficult to make Intel comment on that, but HP is only too happy to gush that Yes, Merced will be there on the desktop, high-end servers and workstations such as ours, and even consumer appliances. That is the volume market, and those are the volumes that will make our machines cheaper.

And when that happens, it would lead to a further consolidation of its position for Intel, and better prospects for HP as a systems and OS maker. UNIX would see a longer life at the top than previously expected, and your desktop will finally be capable of taking on those dream applications of tommorrow.