The End of Open Architecture?

The Success of Open Architecture

While mid-range and high-end machines have always been proprietary, the IBM PC has been an open architecture since it’s inception. The ability for any company to make a better/faster/smaller/cheaper replacement component has been the driving force behind competition, encouraging innovation and lowering prices. Though several attempts have been made in the past to implement proprietary solutions, most have failed as competitors teamed up to create viable open standards and products. Now, it seems that as the line between low-end and mid-range systems is blurring, another company is attempting to force it’s visions and designs upon the industry. Will they succeed?

How big must a company be in order to be able to force the PC industry to follow it’s direction with proprietary solutions? IBM thought they were big enough in 1987 to introduce the Micro-Channel bus, but just one year later, over 60 companies combined to introduce the EISA bus, and succeeded in ousting this usurper. You can’t get much bigger than IBM, so it seemed that nobody could take complete control of this industry – until now…?

Intel had been one of the leading innovators of open solutions even before the IBM PC. Back in the late 1970s they cross-licensed their processor technology with AMD. Among other innovations, Intel gave up their property rights for Socket 7 to the industry, which has helped to create the huge and successful PC market we see today. Intel had also introduced the PCI bus architecture and had given the intellectual property rights to the industry as an alternative to the VLB (open) standard which had become very popular on the 80486 based systems. PCI has been such a successful I/O interface that it has become the de-facto standard for low- and mid-range systems of all types.

When Intel introduced the Socket 8 for Pentium Pro processors, they retained the rights to it, unlike the Socket 7, preventing any other processor manufacturers to develop CPUs for it. This should have been a ‘red flag’, but at the time, the Pentium Pro was considered a mid-range product, and since the bulk of desktop computers were Socket 7 based, nobody really seemed to care a great deal. As the power of the Pentium processors (and clones) increased, more and more mid-range systems switched from using Pentium Pros to the less expensive (but almost as powerful) Pentiums.

Intel Taking Control

Suddenly, in 1997, the peace was shattered by the announcement by Intel that they would stop producing Socket 7 processors and chipsets, and would instead switch to their new Slot I processor interface – which was proprietary. Initially, there was great concern in the industry that Intel would increase their dominance and all but eliminate their competition, however Cyrix and AMD were able to wrest some of Intel’s market share away by developing powerful Socket 7 processors of their own, while creating a marketing opportunity for alternative chipset makers VIA and SiS. In additon, these chipset makers have used their momentum in the Socket 7 market to attempt to make inroads into the Slot I market, while Cyrix announced that they had the rights to produce Slot I based processors. Once again, the industry appears to have thwarted a power-play.

Now we see that Intel has intentions to create another proprietary solution – this time in the memory market. Direct Rambus DRAM (DRDRAM) will soon be supported on Intel chipsets, eventually replacing all current memory solutions, including SDRAM (their roadmap shows this as happening by the year 2001). The industry has responded with DDR SDRAM and SLDRAM, but so far there seems to be little momentum for these solutions. Only one chipset manufacturer (VIA) has provided support for DDR SDRAM, and only recently has any memory manufacturer even shipped any DDR SDRAM chips. So far, Intel has apparently not made any move to support these memorys, so if the industry wishes to combat the proprietary DRDRAM ‘threat’, support for SLDRAM needs to be implemented quickly. Fortunately, Slot I/Pentium II chipsets from SiS and VIA are now being made available, and SLDRAM support should be appearing ver soon. It is possible that this move by Intel may also be thwarted, if enough users and OEMs decide that open architecture is preferable to proprietary solutions.

Just this past week IBM, Compaq and HP announced that they have proposed an enhanced PCI specification to increase the speed to 133MHz, and the throughput to 1GB/sec (up from 132MB/sec currently). This announcement was claimed to be a way for these manufacturers to gain more control over the NT Server market, which Intel has slowly taking over, however there have also been hints that Intel has also been working on a PCI enhancement – proprietary, of course!

Is Proprietary Better?

The danger of all this is that Intel is still the dominant supplier of processors, chipsets and motherboards. They have recently entered the graphics chipset market and the networking market. The AGP bus was ‘given’ to the industry to solve bandwidth issues, but how long until they develop a proprietary enhancement for that as well? Can Intel actually force a proprietary solution on the industry? With DRDRAM, memory manufacturers would be forced to pay a royalty to Rambus, Inc. and to Intel to build modules, giving Intel a dominant role in the future direction of memory design. Dell, the largest PC manufacturer, does not seem to mind this at all, since they currently purchase all of their technology from Intel anyway. Those who have ‘broken away’, such as IBM, Compaq, HP and others, will likely suffer for their deeds if Intel succeeds, and in the long run, so may consumers

An argument can be made that proprietary solutions are actually good, since it tends to create stable, compatible components. Unfortunately, it also tends to stifle innovation and keeps prices high. While businesses prefer solid, reliable solutions and are willing to pay a bit more for them, consumers generally don’t want to pay the higher price. Now that computers have almost become a necessity, like TVs and Microwaves, pricing becomes a major issue. On the other hand, many consumers would prefer that the PC become an appliance that they only need to push a few buttons on to operate, and that they can easily purchase add-ons for that are guaranteed to work together. Without some form of strict industry standards, like those imposed by a proprietary solution, this is a rather remote possibility. Just look at the DVD marketplace – it hasn’t even really begun and there are already three proposed standards from competing companies.

As indicated at the beginning of this editorial, proprietary solutions have always been a part of the mid-range and high-end systems from IBM, DEC and others. Businesses pay big premiums to ensure their systems are reliable, and replacement parts are fully compatible. Now the question is – where is the line between mid-range systems (for business) and low-end systems (for consumers). Intel’s PII-450, which is by all accounts a server processor, can now be easily afforded by many consumers, and is selling very well in the consumer marketplace. Even the ‘low-end’ Celerons are now Slot I based, and will soon be implemented again as a socketed processor – in a proprietary package.

Which Will Prevail?

While Intel is readying the Xeon processor line for the mid-range market, the consumer market is being dragged into the world of proprietary solutions. What is very interesting is that IBM, Compaq and HP, who have traditionally implemented proprietary solutions and are very big in the server market, have recognized the advantages of open architecture in this marketplace. Now that Intel is starting to encroach on their territory, these players are working loosely together to prevent Intel’s proprietary solutions from being accepted by proposing open solutions of their own.

In order for open architecture to continue to prevail in the home, and even the server market, the industry is going to have to give more power to the trade organizations which maintain the standards. All too often, companies will join these groups in an attempt to get their own solutions approved, and when someone else’s in given the nod, these companies go off and implement their own standard anyway. This kind of activity only provides more impetus to the success of proprietary solutions. Businesses and consumers eventually get fed up with the compatibility problems and actually welcome components that are guaranteed to work. It remains to be seen whether manufacturers can control themselves in order to prevent Intel from controlling them