Recent searches
Search options
An insigthful analysis on why there are no minicomputers anymore. More specifically it discusses why the minicomputer form factor, architecture, and vendors disappeared.
https://thechipletter.substack.com/p/why-are-there-no-minicomputers-any
@amoroso I would argue that in terms of power and form factor, they are everywhere: a 4U NAS plus a couple of 2U VM servers, a UPS, take up about a half-height 19” rack. They need dedicated floor space and a small group of people using them to justify that usage.
Not all that uncommon.
@ThreeSigma That's a possible way of looking at it.
@amoroso@fosstodon.org In John Mashey posts, he also pointed out the curious fact that while x86 and VAX are both technically CISC, but x86's crude and restrictive ISA ironically made it easier to speed up in a modernized OoO + pipelined CPU. Not as easy as RISC, but easier than VAX. Meanwhile, VAX's powerful orthogonal ISA meant the instructions and addressing modes can be combined in far too many ways, which gave DEC engineers serious headaches. Combined with the smaller R&D budget due to the low minicomputers volume, it basically ruled out any possibility of a fast and modernized VAX CPU.
@niconiconi Right, it seemed an impractical and doomed approach.
@niconiconi @amoroso VAX was replaced with DEC Alpha and VMS ported + emulations. Alpha could run much vax code faster than VAX. That was never the problem. The issue was solely volume. Intel not only had the budget and volume to make a pig fly, they had it to make a pig fly faster than an Alpha. HP figured out where it was going - hence their Intel partnership but they ended up laying a turd (ia64) not a golden egg and let AMD win with x86-64.
@niconiconi @amoroso The other half then was the classic innovators dilemma. Having a mini that's three times as fast as a PC at something is a great market until the PC gets fast enough for most users, and whilst everyone looks at your extra-fast system they all conclude it's not worth the extra. Great example being SGI. The moment PC 3D graphics were good enough for most things SGI did SGI was over.
@amoroso Would argue a bit differently
Big driver was scale. The smaller ones went first, especially those with custom hardware. Others moved to commodity hardware, the big ones lasted longer because they could sustain a market advantage for a while with higher margins driving better than commodity processors but eventually they couldn't.
- Cost of your software/hardware development is divided by customer base size
- Value of your product is armwavingly O(N^2) users/software
All about scale