Linux.com: Hardware: A Different Approach (Part IV of V): SPARC SystemsNov 27, 2000, 15:06 (0 Talkback[s])
(Other stories by Ross Sanders)
"Some of my readers probably are wondering where I'm rambling to. Others, I'm sure, know where I'm heading. Scalability is the keyword when you are short on cash but need or want an enterprise class system. Last week, I mentioned the alternative practice to being on the bleeding edge of computing. Buy older, create software for newer. Many SPARC systems of the same class using the same OS are backwards and forwards compatible with each other like the UltraSPARC, unless you cross architectural boundaries such as PCI to SBUS."
"The main draw for a SPARC system, be it an Ultra, Super, or Micro, is that you can write your software on a cheaper used computer system similar to the systems you would actually like to run your more resource intensive calculations. Buy a dual-processor UltraSPARC used for about $3000 and this puts you into the position of being able to just recompile the software for a $150,000 64 CPU computer you can gain access to through contacts with minimal problems if any."
"This takes care of computations naturally. The two major drawbacks to this technique is that you don't have access to the $150,000 computer's graphics capability if it has them and not everyone has contacts at corporate R&D departments or a National Lab. However, a two year old SPARC system, as long as it is a PCI based motherboard, can overcome the first problem. Just upgrade the video card. Naturally, it's not the fastest thing on a desk anywhere, but it can match my needs of rendering CCD images and pattern matches quite well. This type of scalability will work with almost any computer system of the same class and basic architecture."