in Techspeak |

Great MIPS chips of the past 30 years

Some 30 years ago, a fresh-faced team of Silicon Valley engineers from Stanford University, Motorola, IBM and other semiconductor companies introduced the computing world to MIPS, a new CPU architecture built around an innovative concept: improve system performance by quickly executing simple instructions.

Originally short for Microcomputer without Interlocked Pipeline Stages, MIPS CPUs were initially designed with UNIX workstations in mind and delivered great floating point performance in a super-pipelined architecture that was easy to program and cost-effective.

Gradually, MIPS CPUs found their way into all sorts of devices, from Sony and Nintedo game consoles (anyone remember the Nintendo N64?) to the New Horizons space probe – and everything else in between. To date, MIPS CPUs have been used in billions of devices around the world, revolutionizing the microchip market of the late 20th century.

In the image below I’ve tried to capture some of the most brilliant achievements of the 1985-1998 era, placing an emphasis on a few key metrics such as number of transistors, process node, die area and clock speeds.

Great MIPS chips of the past[Click here for the high-res version]

A look at the R series

MIPS R2000 marked the first commercial MIPS CPU model. Announced in 1985, it was a 32-bit chipset that included the main R2000 microprocessor, a R2010 floating-point accelerator, and four R2020 write buffer chips. The R2000 core featured a short pipeline and a small translation lookaside buffer for mapping virtual memory addresses.

MIPS R3000 was used in high-end UNIX computers by Siemens and DEC and in the Silicon Graphics SGI Personal IRIS 4D/20 graphic workstations. These machines were used to render 3D sequences for movies like The Abyss, Jurrasic Park and Terminator 2. R3000 sold extremely well (over 1 million units were produced). Derivatives of the R3000 for embedded applications included the R3051 used in the Sony PlayStation game console and the radiation-hardened Mongoose-V used for NASA spacecraft onboard computers (e.g. New Horizons).

MIPS R4000 was the world’s first 64-bit CPU used widely in consumer applications. R4000 was a scalar superpipelined microprocessor with an eight-stage integer pipeline and support for up to four coprocessors. The R4x00 processors have been available and used in different versions, including:

  • R4200: introduced with the Pentium performance in a notebook tagline, it ran Windows NT at twice the speed of an Intel 486 processor
  • R4400: doubled the capacity of the primary caches, and contained 2.3 million transistors.
  • R4300: used by Nintendo in the N64 game console
  • R4600: used in set-top boxes such as the WebTV

MIPS R5000 was a 2-issue superscalar microprocessor with a five-stage pipeline and two execution units that can operate simultaneously. The R5000 processor delivered workstation-class floating-point performance and doubled the cache size over the R4600 chip. R5000 was the basis for a second generation of designs in both the networking and printing markets.

MIPS R10000 (nicknamed T5) was a single-chip 4-issue superscalar processor. Introduced in 1995, it gave a 70 to 100% performance boost over the R4400 at 250 MHz, previously the most powerful MIPS processor. The R10000 CPU was used in NEC supercomputers and various SGI workstations.

MIPS R12000 was introduced in 1998 and delivered a level of performance that placed it in the highest class of microprocessors in the world. It featured many improvements over R10000 in areas such as instruction decoding, branching, and system bus performance.

Although many things have changed since and the instruction set has evolved to accommodate new use cases, it is quite remarkable that most software binaries from a R2000 CPU released three decades ago would run perfectly well and quite efficiently on current-day MIPS Warrior CPUs.

This possibly represents the best achievement of modern computer architecture and design.

If you want to read more about the evolution of the MIPS architecture, check out my previous articles here, here, and here.

Creatures of the die

After digging through the archives carefully maintained by Michael W. Davidson at the Florida State University, I’ve also uncovered a little known fact: it seems MIPS engineers enjoyed leaving little traces of their work in the form of drawings on the printed circuit.

Some of these became regular trademarks (e.g. the California license plates):

Others included various cartoon characters such as the Road Runner or Godzilla. And yes, the second image features a T-Rex playing a guitar!

A few microchips featured more personal and/or cryptic motifs such as flowers or elephants.

Even though these days MIPS is best defined by the low-power, high-performance characteristics of the fast-moving mobile and embedded market, it continues to have a longstanding impact on every CPU architecture in use today. For example, all desktop and server processors released since the 1990s have borrowed heavily from many of the energy-saving RISC philosophies introduced three decades ago.

Published