You may have heard the argument about RISC versus CISC based microprocessors over the years, especially in the heated platform debate of Wintel versus Macintosh. More than likely, you have been misinformed about the true differences in the microprocessors.
The Beginning.
CISC stands for Complex Instruction Set Computing. The first PC market microprocessors were CISC designs. However, the name came after the invention of RISC and in response to its catchy sound.
RISC stands for Reduced Instruction Set Computing. As the name implies, the earliest RISC chips had a smaller instruction set than comparable CISC designs.
In the 1970s when the personal computer was first becoming a reality, hardware was very expensive. Memory was one of the most expensive chips inside the computer. Because of this price limitation CISC processor designs were chosen because the programs to execute on them were smaller. In today's terms, no one would even notice, but at the time, every byte counted.
The philosophy behind the invention of RISC was based on the quirky CISC designs of the time. The CISC processors had many instructions but each instruction took many clock cycles to complete. For example, a MOV instruction may take 2 clock cycles to perform, while the MUL instruction on the earliest CISC processors took over 100 clock cycles. The RISC revolution began with making ALL instructions perform in exactly ONE clock cycle. The limitation was the number of transistors that can be packed into the chip. To make MUL perform in one clock cycle took more transistors in the processor. So, RISC designers reduced the instruction set.
The advantage is obvious. RISC designs may have slightly larger programs, but on some instructions they were over 100 times faster than their CISC counterparts which more than made up for the difference in program size.
Despite RISC's advantages, RISC didn't take off. Probably because of Intel and the computer revolution that began with IBM's PC in 1980 and it's CISC 8088 processor. It is worth noting here, that Apple's move to the RISC based PowerPC chip sparked a mainstream RISC vs. CISC debate. However, Apple Macintosh computers through the early 90s were based on the CISC Motorola 68000 microprocessor.
Today.
Today's microprocessors have had some twenty years of refinement and hardly resemble their early ancestors. Typical designs features like super pipelining, first and second-level cache, pre-fetch buffers and register renaming are now commonplace and performance enhancing on both RISC and CISC based processors.
So where have we gone and who won? The likely answer is CISC with the popularity of Intel's Pentium line of processors but the truth is RISC stole the market.
The Pentium chip is an x86 CISC compatible chip that is based on a 486 pre-fetch buffer, a hardware translation layer and a 2 processor RISC core. The buffer fetches instructions from memory and feeds the translator that turns x86 code into the RISC instructions used in the core. So, in terms of processor popularity, RISC wins.
Although the Pentium is RISC based, it isn't a true RISC processor. To contrast a modern RISC processor against the Pentium, we will look at IBM and Motorola's PowerPC chip.
The PowerPC chip differs from the x86 platform in a number of ways but two really stand out. First, the PowerPC chip has 32 general-purpose registers instead of the Pentium's 8. These registers are open for the programmer to use as he sees fit, and are not limited by the special purpose design of the Pentium. So, instead of having just the AX register and another to multiply with, you can use any of the 32 general-purpose registers. Therefore more work is done inside the processor with less writes to memory increasing the processor's speed.
Secondly, the PowerPC chip executes even length instructions in exactly one clock cycle and processor cores are always working on an instruction. The Pentium design has to translate un-even length instructions to RISC micro code and sometimes a processor core will stall, not being able to do anything until the rest of the processor catches up.
So, RISC wins today, but what's in store for tomorrow?
The Future.
The future of microprocessor design seems to focus on two new key technologies: VLIW - Very Long Instruction Word and Code Morphing.
Intel's and TransMeta's (Linus Torvald's company -- THE Linux guy) designs use VLIW technology. VLIW takes a series of instructions from memory and packs them into a long block. These blocks are fed into the microprocessor in a way that the processors multiple cores and units all get an instruction. No part of the processor is left to wait while there are instructions to be done. VLIW processors are RISC based so all these instructions complete and exit the processor at the same time and a new block of instructions is fed into the processor. No waiting and very efficient.
TransMeta is also poised to reinvent the microprocessor market with its new Crusoe processor that has Code Morphing technology. In the design philosophy of RISC, TransMeta has created their processor to be very fast with a small number of instructions and a small number of transistors in the chip. There is a software translation layer that converts the incoming instructions to the Crusoe's core much like the hardware translation layer in the Pentium. This results in a processor that is code independent. With the proper translation layer, the Crusoe can run ANY existing software and operating system on the face of the earth. Another benefit from the reduced number of transistors is battery life of over a day in a conventional laptop.
Seems like a science fiction novel, but it isn't. The Crusoe is available now in a nice 700mhz flavor, and Intel's Merced VLIW chip is due out next year. So, even with all this break-through technology, at the heart of it is RISC technology and a philosophy that started over 20 years ago.
I am an american artist.
Thank you for spending your time and hard earned money with me, here and in the real world. You can contact me, hire me, read about me or just check out my jeep.
