Dr. Boris Babaian and his Elbrus design team are renowned supercomputer architects whose work seems to have inspired, or at least predated, much of the recent development of VLIW-like architectures in the West. Their work on the Elbrus-1 through Elbrus-3 supercomputers has impressed many for the performance achieved, given the primitive implementation technology. But they have virtually no experience in the science and art of designing a high performance ISA and processor microarchitecture to permit high quality implementations in modern CMOS process technologies. The engineering methodologies involved in building a processor out of thousands of small scale logic chips distributed over many circuit boards is very different from designing a processor out of millions of 0.18 um long MOSFETs distributed over a few hundred square millimeters of silicon.
Imagine if a famous and well-respected leader at theRussian Mikoyan-Gurevich aircraft design bureau announced that they were looking for funding to implement a promising paper design for a commercial airliner, the MIG2k. What kind of scepticism would a leading trade publication like Aviation Week and Space Technology offer to claims that the MIG2k would use standard manufacturing techniques and engine technology, yet cost half of a Boeing or Airbus aircraft of equivalent seating capacity, consume half the fuel per passenger mile, and by the way, do this while flying at Mach 3? Would they say the ‘claims for the [Mig2k] would seem unbelievable, if not for the credibility of the [Mikoyan-Gurevich] team’? The author of a technical report about a ‘paper’ design concept should be influenced less by celebrity and reputation than a careful look at the obvious physical impact of design trade-offs relative to leading edge designs that actually exist and function as intended. This is as true for microprocessors as it is for aircraft.
Be the first to discuss this article!