Diversity all but gone in computers and their architecture - good or bad?

[aluminumstudios]

I've been thinking a lot lately about how all diversity is gone from desktop computer systems.

In the 80's there were a plethora of DIFFERENT computers you could choose from - Commodore, Amiga, PC, Mac, Texas Instruments, and on and on.  They all bore unique designs with different strengths and weaknesses.  I remember the large paper catalog from the local computer store and it's various sections for different platforms.  I would always turn right to the Texas Instruments section and look for what game I was going to beg my mom to buy for me next.

The came the 90's with some innovative and powerful systems such as Sun and SGI emerging.  Once again, they brought unique features, innovation, and choice.  When I was in university I used to love visiting the "tech lab" and mucking about on SGI workstations and Sun desktop systems because they were different from the PC and software I had back in my dorm room.

With PC's becoming so ubiquitous and accessible, it was natural that tons of profit from it could go into R&D and the x86 architecture, despite the design superiority and elegance of other architectures, soon found itself R&D'd to price/performance levels that started making MIPS based SGI, or Sparc based Sun systems less attractive.  Likewise, naturally, developers flocked to the popular platforms because they could make more sales, which naturally took a toll on the desirability of smaller platforms.

With supercomputers being constructed out of x86 platforms, tight finances killing SGI and Sun and laregly taking them out of the game, and remaining non-x86 architectures only found in the server room, I'm finding myself a little bored with the computer systems I can get my hands on.  I used to be able to daydream about having a stylish SGI on my desktop.  There was something impressive about the Suns my university professors used to have on their desks in the 90's, but now all of that is gone.  Even if I buy a fancy case, it's still just the same-old, same-old. 

While x86 performance is seriously unquestioned, and the systems give me more than any of the systems of yore ever could, the romance is gone.

I guess I could always buy some vintage systems, but in terms of a modern computer that I use, the choice is gone.  I don't consider a Mac and a "PC" different enough from each other to count, nor only 2 players as "diversity." 

What do you think?  It is good or bad to be rid of diversity in the realm of desktop computer systems?

 

[fractower]

As it happens I could not agree with you more. I remember the good old days of the PC when you could program down to the bare metal. If the OS was using memory that you needed you just took it. If the bios was in your way you wrote to the frame buffer directly. We were gods!

I attended a conference recently where I complained to a young friend of mine that the current round of portable devices just weren't as programmable as the PC. He agreed whole heartedly and then went into a rant about the inadequacies of the SQL implementation on the I-Phone. I think one of us missed the point, and it was probably me. I have finally accepted that I will never again run volume render I wrote for the Burris B20 back when I was a freshmen in collage. That was a machine that was ahead of its time for about 6 months.

[hobold]

IMHO lack of diversity is pretty much always bad. I consider the natural evolution of species as strong evidence for my opinion. When left alone, nature always diversifies. Overspecialization of species only happens under extreme environmental pressure, such as the effects caused by mankind, or after very extreme, very rare natural catastrophes.

However, the current situation in computing is not yet quite that bad. The 'x86 architecture may be dominating the desktop computer, but the traditional functions of that device continue to be accroached by smartphones, netbooks, tablets and other small portable gadgets. And those little ones are certainly not dominated by 'x86 in any way, not even the netbooks (although admittedly, Intel does very well there with their Atom processors). Game consoles are all non-x86, too.

Computation intensive tasks are being delegated to generically programmable graphics processing units in earnest (and those do have a lot of exciting evolution still ahead of them). So 'x86 is under pressure from above as well. And the really big machines are still resisting. IBM is doing fine with their mainframes and POWER servers, even Oracle still sells SPARCs ... Itanium is still not really competitive despite Intel's weight behind it.


In a very real sense, Intel themselves are pursuing diversification with their research projects: Larrabee, the "Single Chip Cloud", and probably a few more that we never heard about. The fact that they all decode 'x86 opcodes is turning more and more into an irrelevant implementation detail.

[David Makin]

There's a "simple" way for chip manufacturers to re-introduce diversity from the programmers point of view - produce a new chip design that has a (runtime) programmable instruction set (or at least partially programmable).
Also to me the real loss of diversity is not down to the reduction in the number of chipsets, rather it's down to too many software designers relying on high level OS routines rather than writing their own code - this is also the biggest problem in terms of bugs - especially if working with Microsoft or Symbian OS/API calls !

[aluminumstudios]

Good points from everyone.

Let me clarify though.  I don't think the computing world is not developing or innovating, nor in danger from lack of diversity.  Evolution (GPU computing, etc) and processor innovation is alive and well and bringing us GREAT things.  I'm just a little nostalgic for the 80's and 90's where there were COMPLETELY different desktop systems to choose from with different designs, different architectures, and different OS's.

Being a bit of an old-school computer geek, the era of smartphones and other portable devices doesn't satisfy me the same way as when I was in University in the 90's and would sit down in front of an exotic looking SGI and mess around on a system that was foreign to me.  Or even when I was in junior high and my friends used to come over to see what I had on my TI 994a and I would go to my friends house and drool over his C64 

As it happens I could not agree with you more. I remember the good old days of the PC when you could program down to the bare metal.

I think about this a lot.  I learned to program under DOS and Unix in University and before that learned BASIC and Turbo Pascal by myself on my Texas Instruments computer then later my Tandy.  While I am far from a great programmer, when I work I think in terms of data and what the computer is doing, not in terms of API calls.  I worry that new generations of computer science students won't get a solid grip of what is happening under the hood because of the ever increasing abstraction.  I also worry that this will encourage future generations of systems to bloat up and separate the programmers and users from the "metal" even more.

[David Makin]

While I am far from a great programmer, when I work I think in terms of data and what the computer is doing, not in terms of API calls.  I worry that new generations of computer science students won't get a solid grip of what is happening under the hood because of the ever increasing abstraction.  I also worry that this will encourage future generations of systems to bloat up and separate the programmers and users from the "metal" even more.

I couldn't agree more - I've thought that since around 1994 !!
The first thing we used to do on the Amiga was remove the OS.

My first program was on a borrowed Commodore Pet. I decided to write a "duckshoot" game. I was very new to programming and wrote the routines to move the ducks, gun and bullets separately - so if one moved the others didn't !! When I sussed that I had to update everything all the time I also realised that doing so would bring the program to a virtual standstill, but did it anyway and got a duckshoot game that ran at around 3 fps, it was more like a strategy game
That was when I realised that I'd have to learn machine code !!
In the end I really learned programming on the Dragon 32 and 64 starting with BASIC but progressing to 6809 assembler (using the fabulous Encoder09 which I think may also have worked on the Tandy - it had "compile to tape" so you could write software that would use the whole available RAM i.e. a massive 32 or 64k).

[Nahee_Enterprises]

You guys should have seen what it was like working with the big computers in the old days.  Having to punch specific holes in tapes to make machines function in the way you wanted.  Or use plug-boards where you set the jumpers so that they would read cards and interpret them correctly.  Or flipping switches at the console of a main frame to have it process a certain way.

Programming back then was a lot more than just writing code, one had to know how to physically work with the computer and all of the peripherals as well.
 

[aluminumstudios]

You guys should have seen what it was like working with the big computers in the old days.  Having to punch specific holes in tapes to make machines function in the way you wanted.  Or use plug-boards where you set the jumpers so that they would read cards and interpret them correctly.  Or flipping switches at the console of a main frame to have it process a certain way.

Programming back then was a lot more than just writing code, one had to know how to physically work with the computer and all of the peripherals as well.
 

Just like photography (my other hobby)!  I started with digital, but have gone back to old fully manual film cameras and developing film myself because I enjoy having "physically work" it.  I doubt the same pleasure could be found in working really old computers though

[hobold]

I doubt the same pleasure could be found in working really old computers though

The hardest part of it is to willingly forget that a software emulation is now faster than the old hardware. In other words, working with antique computers is nicer in simulation than in reality.