Now’s a great time to be thinking about the future of computing processor architectures. We have five mainstream survivors of the first 50 years of computing – S/360 (1964); x86 (1979); SPARC (1987); Power (1990); and Itanium (2001). I have five retired architectures in mind: VAX (1977); 68000 (1980); MIPS (1985); PowerPC (1992) and Alpha (1992). Yes, I know that PPC is still appearing in supercomputers and game consoles, but without Apple it is out of the mainstream.
Note that technical superiority does not define success: execution of the business model is the primary determinant. Both the S/360 and the x86 architectures are cranky artifacts of the days when assembler was the only way to program, but they are two likely to stand through the next 20 years.
So, first, did I miss any architectures? The reasons for demise, and the ultimate resting place, of the retired architectures are interesting pointers for the future.
Then, which of the five current architectures are most likely to make it through 2020? I have my own ideas, but am interested in yours.
4 responses so far ↓
1 Gerry Van Zandt // Apr 1, 2009 at 5:32 pm
Martin,
PA-RISC is one, and I think it’s pretty significant as it innovated within the RISC space and had a healthy life-cycle. Perhaps it even has a bit of DNA in the Itanium architecture =)
I don’t know if you’d include the ARM instruction set as another, as it is more low-power/mobile-oriented, but it’s certainly for computing and it’s strong in its core markets.
Best,
Gerry
2 Nick Jones // Apr 7, 2009 at 4:09 am
I’d suggest that the question doesn’t really matter any more. The processor architecture is increasingly irrelevant. The applications that really matter will be written in Java, Python, Ruby and a pile of similar languages that run on almost anything from a mobile phone to a supercomputer. The point at which we need to know about hardware architecture has moved up and away away from the chip architecture. We do still care about issues like how clusters are connected and what APIs we use to program for massively clustered systems. But why should we care about the processors running on the nodes of the cluster? And if we still care in a decade, it will probably be for reasons which are nothing to do with chip architectures or marketing but will be driven by things most of us would consider irrelevant today, like MIPS per kilowatt hour.
3 Stephen Shankland // Aug 6, 2009 at 1:43 pm
I’m not sure I’d call S/360 and its derivatives “mainstream” when PowerPC is merely used in gaming consoles. Also, Power is closely related to PowerPC, should you really divide those two into separate architectures?
4 Martin Reynolds // Aug 6, 2009 at 10:22 pm
Thanks, Stephen. The S/360 line has proven to be remarkably resilient. It may not be front and center, but it remains core to much of modern business. I divided PowerPC and Power, despite their common heritage, because PowerPC had such a prominent space in volume computing. PowerPC’s position in game consoles is dominant, but it is no longer a general-purpose computing processor. The Power processor tops and tails the PowerPC era, so maybe PPC is a red herring.
I’m guessing that SPARC isn’t going to get a whole lot of love from Oracle.
Leave a Comment