Blog

What’s in a processor?

Generally speaking, the difference between a microcontroller and a “processor” is soft code programming, microcontrollers typically run from a ROM-based set of instructions, which may simply be a function of manufacturing and cost convenience, and not anything intrinsically lacking in the processor. Processors have two basic characteristics, their processing capability, and their instruction set. Capability is measured in MIPS, SPEC, ...

Robert Dow

Generally speaking, the difference between a microcontroller and a “processor” is soft code programming, microcontrollers typically run from a ROM-based set of instructions, which may simply be a function of manufacturing and cost convenience, and not anything intrinsically lacking in the processor.

Processors have two basic characteristics, their processing capability, and their instruction set. Capability is measured in MIPS, SPEC, FLOPS, and other hopefully significant but usually esoteric metrics. Sometimes they are measured in GHz, bit width, “cores,” and occasionally architecture.

The instruction set is the primary source of the IP, and subsequently the licensing of processor technology. AMD, ARM, ARC, Intel, MIPS and others have licensing programs for organizations that want instruction (code) set compatibility with their processors.

In addition to the big, well-known names, ther are other companies with their own proprietary processor designs, or a mimic of one of the others. Sometimes the mimic is allowed to operate without license or redress (as in the case of VIA’s x86 C7 Nano processor), and sometimes they are severely smacked by the IP owner (as in the case of Transmeta.) It’s an inconsistent and inequitable policy of the IP holders based on pique, paranoia, and sometimes genuine fear of market share or revenue loss.

For the others, like, Ambric, IBM, Imagination Technology, Renesas, TI, Tensilica, and others they go about their business without much harassment and often a degree of amusement.

Generally speaking, the difference between a microcontroller and a “processor” is soft code programming, microcontrollers typically run from a ROM-based set of instructions, which may simply be a function of manufacturing and cost convenience, and not anything intrinsically lacking in the processor. Almost everyone has a microcontroller or two in their chips.

In terms of volume, ARM is the dominant supplier, with its design in literally billions of systems from automobiles to mobile phones to spaceships and light fixtures. In terms of scalar complexity, Intel dominates with its processors in just about every PC since 1980 and many STBs and embedded systems.

All of which, in perhaps a laborious and tenuous way, leads us to the question of Nvidia’s long-rumored development of a “processor.” The first consideration is to define what is meant by processor, because in the company’s Tegra products it is already producing a “processor.’ Two, in fact, can be found in the Tegra. But when Nvidia is discussed in terms of processors, it is more or less implied that x86 is being spoken about.

Nvidia is expected (or maybe hoped) to have or be in the process of acquiring x86 capability and to subsequently produce an x86 compatible device to compete with AMD and Intel as those two companies encroach on Nvidia’s turf. The latest rumor to fuel these speculations is the alleged acquisition of shares in VIA by Nvidia. The presumption being that in so doing Nvidia would gain access to VIA’s x86 technology.

If Nvidia were to follow that route they would then be the fourth x86 producer, not a very favorable position in a very aggressive market where one company holds 80% market share.

Secondly, as mentioned above, VIA does not have an x86 license and operates at the pleasure or perhaps indifference of Intel. VIA’s x86 processors are built in Fujitsu’s fab and Fujitsu doesn’t have an x86 architectural license like IBM does. So, it’s questionable how much indemnity Fujitsu can provide VIA.

However, if Nvidia were to build an x86 without a direct license from Intel, it could do so presumably under IBM’s umbrella. And even that would still likely provoke Intel into some form legal action.

Given the pugnacious stance of the two companies and the demonstrated litigious nature of Intel in particular, there is little logic in Nvidia entering the x86 arena, maybe it is absurd to even consider it.

Which reminds me of the time I predicted Nvidia wouldn’t enter the IGP market—and then they did. And, ironically, this issue of licenses and processors is fueled by Intel blocking Nvidia from participating in the chipset
business.

“Simple pleasures are the last refuge of the complex.”—Oscar Wilde