My colleagues and I at BDTI frequently evaluate processors – sometimes on behalf of processor suppliers, and sometimes on behalf of processors users. Comparing processors can be complicated, given their many attributes, including some (like energy efficiency) that are difficult to compare accurately, and others (like ease-of-use) that are inherently subjective.
Several recent conversations with processor suppliers have reminded me of another serious hazard in processor comparisons, one that I call the "Uncertainty Principle of Processor Comparison."
Here's how it works: Supplier Z is developing a new processor. Naturally, the engineering and marketing managers at Supplier Z are aiming to make their new processor superior to competing offerings. The problem is that Supplier Z doesn't know what new products its competitors are working on; it can only compare its product under development to its competitors' currently shipping products.
This is where the hazard arises. Processor performance and related metrics – like performance-per-cost and energy efficiency – don't always improve in a slow, steady fashion, governed by silicon fabrication process improvements. Rather, processors can improve quickly – and this is particularly true for processors that incorporate specialized features targeting specific applications. So, comparing a processor under development (which may not begin production until a year or two in the future) with a processor that's already in production typically yields exciting results. "Wow!" say the product managers at Supplier Z, "Our new Z6 processor has 3X the energy efficiency of Supplier Y's Y5 processor!" But by the time Supplier Z's new Z6 processor reaches the market, it probably won't be competing against Supplier Y's Y5 chip, but instead against its successor, Y6. And guess what? Y6 may also be 3X more energy efficient than Y5.
When I point out this hazard to processor suppliers like Supplier Z, the usual responses are, "We can only get data on our competitors' shipping products, not on products they're current designing," and "customers want to compare our new processor against what they're using now." These are valid and practical reasons to compare Z6 vs. Y5. But often, Supplier Z winds up deluding itself by thinking that this kind of comparison means that when its new processor emerges, it will have a compelling competitive advantage.
This can be a dangerous delusion, especially for start-up processor suppliers hoping to displace established suppliers. Engineers and engineering managers know that there's often significant cost and time associated with switching between processor families. And they know that there's risk in adopting a processor from an unproven supplier. So, for a start-up processor supplier to win business away from established competitors, its products must deliver more than a slight advantage.
In other words, faulty logic that ignores the Uncertainty Principle of Processor Comparison can spell doom for processor suppliers. Yes, it's true that you only have access to data on competitors' current products. And it's true that customers are interested in comparing your future product to what they're currently using. But, ultimately, the comparison that matters is between your future product and the competing products that will be available at the same time. Yes, that's a much harder comparison to make. But processor suppliers that fail to make it risk deluding themselves into oblivion.