That is a rad motherboard! I've never seen anything like it.

No clue about RISC stuff, but some early socket 4 boards came as VIP packages (VESA, ISA, PCI) and I have heard they were quite buggy.

I found the briefing of the onboard logicore chipset with a diagram showing the bridging between SYSAD and VLB.
As for consumer x86 systems, VLB stretched to some early socket 5 boards.

I have heard about some 386 boards with VLB. VLB was a stopgap, and one that had a lot of issues even on 486 platforms. One interesting advantage it had was on systems with a 40mhz FSB it outperformed PCI by a considerable amount. (The calculation is 32 bits 40,000,000 times a second on full PC bus attention to the card, of course things like other cards and interrupts will change that speed a lot, so it's not a completely cut and dry speed like PCI-E is.)

Am I the only one who is impressed that a RISC board has plain ISA ?
Both PCI and EISA were semiprofessional and well standardized,
but ISA was historically an x86/PC-AT creation. IRQ lines,
ROMS with 8086 code, hardcoded port addresses, etc.

Many of the "PC RISC" designs worked by exposing a x86 bus from the system bus. With such design a manufacturer would add some off the shelf Intel components and have for example an EISA bus (with the Intel 82350 chipset). VLB is more or less an exposed 486 host bus and the Logicore documentation lazibayer found states exactly that:

HCHIP creates a VESA Local or ¡486 host bus with the signals necessary to interface to Intel compatible peripheral chips.

It was all about costs - workstation manufacturers wanted to enter the high-end PC market with RISC chips, so they made motherboards using popular buses.

Jo22 wrote:

Am I the only one who is impressed that a RISC board has plain ISA ? Both PCI and EISA were semiprofessional and well standardi […]
Show full quote

Am I the only one who is impressed that a RISC board has plain ISA ?
Both PCI and EISA were semiprofessional and well standardized,
but ISA was historically an x86/PC-AT creation. IRQ lines,
ROMS with 8086 code, hardcoded port addresses, etc.

As I said, it was all about costs. You couldn't sell PCs in this market and say: throw out everything you have, use our ultra fast proprietary bus. And don't look at it from the CPU perspective - the MIPS on this board knows nothing about the ISA bus.

Jo22 wrote:

Am I the only one who is impressed that a RISC board has plain ISA ? Both PCI and EISA were semiprofessional and well standardi […]
Show full quote

Am I the only one who is impressed that a RISC board has plain ISA ?
Both PCI and EISA were semiprofessional and well standardized,
but ISA was historically an x86/PC-AT creation. IRQ lines,
ROMS with 8086 code, hardcoded port addresses, etc.

I reckon the OPTi 82C499 chip is responsible for bridging VLB and ISA buses.

Nice bit of bizarrechitecture. I assume it was done deliberately so that the system could use the wide variety of VLB video cards, SCSI devices, etc. (that were already supported by Windows.)

As I said, it was all about costs. You couldn't sell PCs in this market and say: throw out everything you have, use our ultra fast proprietary bus.

Thanks, but I meant something different here (no money on the mind). PCI and EISA were advanced designs that
were at least imagined to be used on different architectures, too. ISA cards were much more low-level,
sometimes they had no bus controller chips at all (just think of the gameport).

That's what I meant. Simple designs, like serial/parallel controllers or IDE host interfaces could be used from
any machine type. But what about network cards w/ boot socket, RAID controllers with their own BIOS,
proprietary CD-ROM drives, FM radio cards, etc. Such devices were of little use in a RISC machine,
because of lack of software

And don't look at it from the CPU perspective - the MIPS on this board knows nothing about the ISA bus.

..or because of lack for direct hardware access from the CPU side. 😉

Or in other words, the interesting kind of ISA cards were always those that were closely tied to the machine.
I.e., special hardware. Like CNC controllers, cards for measurements, special I/O cards, digitizers, etc.
They wouldn't work without x86 fluff, because they worked so much on the bare metal.
Both PCI and EISA are intelligent, in comparison. Their I/O ports can be changed, DMA works different,
32-Bit word size, etc. Heck, they could even be detected and categorized by a foreign architecures' firmware! 😁

That's why I'm impressed or surprised about ISA in a RISC architecture.
The cheap stuff was also available for PCI - with a much higher performance (which RISC was all about).

Jo22 wrote:

The cheap stuff was also available for PCI - with a much higher performance (which RISC was all about).

Keep in mind that the DeskStation was released in 1994 and PCI just became available. There were almost no PCI cards on the market at that time. It would also be hard to test and validate such design without cards being available.

They could of course use EISA, but my guess is they either wanted to make it cheaper or they haven't had time for it. EISA would require getting part of the Intel EISA chipset (several chips). SNI used EISA via the Intel 350DT for their MIPS workstation: https://www.linux-mips.org/wiki/RM_400 . Also the first Alpha PC, the DECpc AXP 150 used the 350DT chipset, but the beta system used the ISA bus with the Intel 82380 chipset. You can read about these here http://www.1000bit.it/ad/bro/digital/djt/dtj_v06-01_1994.pdf - the article has the title "The Evolution of the Alpha AXP PC".

Deskstation uses OPTi 82C499 for its VLB support, just as these few well know Pentium boards with VLB do. Is well know this "emulation hack" doesn't perform as fast as the real "VLB" found in the 486 boards, which exposes real 486 cpu pins to hardware.

torindkflt wrote:

Pentium motherboards with VLB slots do exist, but they require special controllers to translate between the PCI and VLB bus.

Maybe some Pentium chipsets worked that way, but on OPTi chipsets it was the other way around.

I am not entirely sure about how VLB is glued to a Pentium chip, but what I do know is that all of the initial Pentium chipsets from OPTi were VLB, because at that time VLB was very popular and it was not clear that PCI was the future. OPTi bet the farm on VLB...but then at some point in 1994 they realized they made the wrong choice, and desperately tried to get a PCI chipset on the market. They were not entirely successful, and this resulted in the 82C822 VLB to PCI bridge interface. Most of the "PCI" OPTi boards used this bridge chip, but the PCI performance was really bad (VLB cards worked ok). I believe they eventually developed a real PCI chipset, but it came out pathetically late in the game.

hyoenmadan wrote:

Deskstation uses OPTi 82C499 for its VLB support, just as these few well know Pentium boards with VLB do. Is well know this "emulation hack" doesn't perform as fast as the real "VLB" found in the 486 boards, which exposes real 486 cpu pins to hardware.

82C499 is not just a VLB support chip though. It is a self contained 486 chipset.

Ampera wrote:

VLB was a stopgap, and one that had a lot of issues even on 486 platforms.

Maybe in retrospect VLB looks like a stopgap solution, but it wasn't intended to be one. VLB and PCI were developed by competing organisations, both of which intended to dominate future PC bus specifications. There was actually a 2.0 specification of VESA local bus which saw implementation in some chipsets, but unfortunately not in expansion cards (by that time, PCI was the obvious winner).

Ampera wrote:

I have heard about some 386 boards with VLB. VLB was a stopgap, and one that had a lot of issues even on 486 platforms. One interesting advantage it had was on systems with a 40mhz FSB it outperformed PCI by a considerable amount.

I have one. For some reason, it will no longer take any other chip than the one in there now (Am386DX) but it does work quite well with it. I had a 486 DX50 back in the day and I don't recall having any timing issues with VLB, as well, I had a P60 machine with PCI and VLB - I never tested that one with VLB video as it quickly left my floor at a steep discount due to a 486 being more powerful at the time.

Some PowerPC Reference Platform machines had ISA slots too.
As did the BeBox, for the purpose of allowing buyers to use the "PC clone organ bank" as JL Gassée said at the time.

lolo799 wrote:

Some PowerPC Reference Platform machines had ISA slots too.
As did the BeBox, for the purpose of allowing buyers to use the "PC clone organ bank" as JL Gassée said at the time.

Well, if you have a PCI bus it's pretty easy to have ISA as well - you simply add an Intel southbridge. Unfortunately, it's often the old, slow SIO - this is the one on the BeBox schematics, but I have it also on my 164LX Alpha motherboard from 1998.