~ wrote:
Octocontrabass wrote:
~ wrote:
They talk about every aspect and service of a classical BIOS, the fully standard base, not just for Phoenix BIOSes in any way.
Do they happen to explain the detection method and calling conventions for 32-bit EISA functions?
The EISA bus is the 32-bit version of the ISA bus. The first volume (System BIOS) talks about using it and all 3 volumes assume EISA throughout the text.
It's apparently one step behind Plug and Play, so those services would also depend on the ROMs of the cards installed, like the video BIOS, as we have already proven.
Have fun trying to find one of the half-dozen or so
EISA motherboards that were actually produced between 1989 and 1992. They were never common (mainly seeing use in office LAN servers), and by 1992, were already too slow for use as a bus for the new accelerated video cards that were coming on the market, hence the development of VLB, AGP, and PCI.
EISA was introduced as an alternative to licensing MCA from IBM, but neither of them really caught on, as the need for a faster bus really wasn't there at the time - the important bottlenecks were elsewhere.
When Big Blue announced the PS/2 line in early 1987, they made no bones about it being an attempt to re-rail their control over the PC market, which they'd lost after the flood of clones that followed in the wake of Compaq's legal victories in 1983-85. While the new bus design,
Micro-Channel Architecture, was genuinely superior, it was mostly a solution in search of a problem - the lack of a suitable operating system was a much more serious stumbling block than the lack of a 32-bit bus was, especially since most MCA systems were actually 16-bit (the 386 systems were only for the very top-end systems), and the history of how IBM bobbled the development and release of OS/2 (and the subsequent rise of Windows) meant it would remain a problem for almost six more years - and more importantly, was completely proprietary, with several patented technologies required to implement both the bus and the interface cards. IBM basically was saying, "sure, we'll let you build clones if you license the design from us first", but then charged so much in per-unit royalties that it would have priced their competition out of business.
A few companies (most notably Tandy-Radio Shack) did license MCA for a few models, but most of the clone builders told IBM to get lost. Still, it left them without a standard 32-bit bus, which if nothing else looked bad, even if the need for it really wasn't there yet. A few played around with
NuBus, an existing standard originally meant for LispMs which both Apple and NeXT had already adopted. I think some even tried to back a 32-bit extension of the
S-100 bus, bypassing the PC design completely, but I'm not sure. But most followed Compaq's lead on EISA.
Compaq had already designed a 32-bit extension of the ISA bus for their Deskpro 386 memory boards even before the PS/2 line were announced, and in 1988 offered to use that as the basis for an 'standard' extended bus. A number of other companies joined in, but given that most of the clone makers were still selling XT class systems (and would continue to for the next three years), it really didn't make much of splash in the marketplace. In the end, only a few companies made EISA motherboards, and almost no EISA cards were released.
Interestingly, the first step in developing the "Enhanced Industry Standard Architecture" was to develop an "
Industry Standard Architecture" in the first place. The term "ISA" was back-developed from the name of EISA, as there hadn't been a formal standard for the PC-compatible bus up to that point. But it all turned out to be a fruitless effort, as events unfolded in unexpected ways that put an end to ISA and EISA even before the standard was officially ratified.
By 1992, the rise of
VESA Local Bus and the
Acelerated Graphics Port had put an end to the development of EISA video cards, and the next year, a new local bus standard,
Peripheral Component Interconnect, was released. PCI was a new design from the ground up, with few similarities to any of the earlier bus architectures (while you could have a gateway to allow ISA and EISA slots on the same motherboard, those were only used for the first few years while new PCI cards got developed), and the combination of PCI and AGP quickly pushed everything else out of the market completely until both were in turn replaced by PCI Express around 2005.
On a similar note, AFAIK, the ABIOS was only used on the MCA systems, and even IBM had dropped support of it by the early 1990s.