Voodoo Orchid Righteous 3D is shorter than the Diamond Monster 3D, but not sure if it's the shortest.

risers

Every time I look at the AT motherboard layout with the CPU to the right of the slots, I wonder what they were thinking. They KNEW how big ISA cards had been, how big they still were and how big early PCI cards were, and yet they designed boards in such a way that a CPU heatsink basically always prevents long cards from being used in 2 or 3 slots. Derp.

I have 7 voodoo 1 cards of different brands (12 before downsizing a bit) and they all are about the same length, except for the Orchid as that one is a fair bit shorter.

meljor wrote:

I have 7 voodoo 1 cards of different brands (12 before downsizing a bit) and they all are about the same length, except for the Orchid as that one is a fair bit shorter.

I found interesting of having only Voodoo 1 and 2 cards of interesting colors like Voodoo 2 PROvideo that are blue colored, or the Black magic. Or Voodoo 1 skywell that are black as well, otherwise Creative ones for machines are very good to use, or Diamond ones. But for only collection in boxes the different colored Voodoo are every day better then the green/yellow ones...

Ozzuneoj wrote:

Every time I look at the AT motherboard layout with the CPU to the right of the slots, I wonder what they were thinking. They KNEW how big ISA cards had been, how big they still were and how big early PCI cards were, and yet they designed boards in such a way that a CPU heatsink basically always prevents long cards from being used in 2 or 3 slots. Derp.

If you wonder that, consider the IBM AT itself, as that is what the form factor is based on.

Note couple of things:
- it's HUGE. What you are referring to as "AT" is in fact Baby AT, and is more comparable to uATX than full ATX. A full AT board is bigger than EATX.
- the CPU is NOT in front of the slots.
- een if the CPU were in front of the slots, it wouldn't matter as it's a flat chip which didn't need a heatsink.

"Baby AT" was a form factor introduced by the clone manufacturers as a space-saving measure (IBM went its own way with PS/2 and then settled on LPX for the PS/1 and Aptiva range before ending up with (u)ATX just before selling the whole business to Lenovo), and as with any such measure, you get compromises.

The thing about Baby AT is that it's designed for minitowers, where the upper third of the board has very little vertical clearance due to drive cages hanging a few cm above it. So anything that is taller than say 3cm needs to be in the lower 2/3. Which is where the slots are. That means that as soon as CPUs started needing serious cooling, they inevitably had to start conflicting with expansion cards. You're right in saying that this was impractical - it was one of the main reasons Intel introduced the ATX form factor, to have a standard where big CPU coolers could be accomodated without interfering with slots. But it's completely unjustified to blame that on the AT designers at IBM, or even on the clone vendors who came up with Baby AT. When they were doing that, there was no indication whatsoever that CPUs would need big towers above them, forcing them into conflict with cards - and for a full decade it worked as designed - i.e. fine.

Even in the second half of the 1990s when this became an issue most board designs did their best to mitigate it. You almost always see the CPU located where ISA and PCI meet, so there will be at least one full-length ISA and one full-length PCI slot available, frequently two. Plus the CPU is usually far enough away from the back plate to allow 2/3 length PCI cards (such as most Voodoos) to fit. Only on exceptionally compact (or badly designed...) boards did you get stuck with a few slots that really couldn't handle anything over the length of an ISA slot. But that is no different than today in a very compact uATX case, which only has a few slots available and frequently cannot handle anything but the shortest cards. That's the nature of compromise with compact boards and systems, and the only one who really should have been thinking harder is someone trying to fit server-class (or high-end gaming) cards into a system like that.

Case in point: my original P60 bought in 1995 in the smallest, cheapest, crappiest minitower money could buy. The motherboard was an old MSI, I believe an MS-5103 - Baby-AT board with 4x ISA, 4x PCI and a big So4 with P60 that needed active cooling (definitely in that cramped case). It blocked off four slots, so "all" I was left with, was 2x full length ISA, 2x half-length ISA, 2x half-length PCI (one shared with ISA) and 2x full-length PCI. That's four full length slots. My GUS MAX went into the bottom one. Apart from that, my S3 868 card fitted into any of the PCI slots and I later added a PCI modem and finally an ISA NIC, either of which fitted into any of the available slots. So with an oversized sound card, VGA, NIC and modem AND a CPU blocking stuff I *stil* had 1x full length ISA and 1x full length PCI left (and either a half-length ISA or PCI slot as well). It's certainly possible to imagine configurations that would have maxed this out, but they are not realistic for the market segment these boards were aimed at.

It is possible to position the CPU in corner of a baby AT board where it will not block any slots. See for example the A-Trend ATC-5000:

In fact I bought this board for that exact reason...

meljor wrote:

I have 7 voodoo 1 cards of different brands (12 before downsizing a bit) and they all are about the same length, except for the Orchid as that one is a fair bit shorter.

Most likely because they all follow 3dfx' reference design. I own four different cards currently, which are all the same, apart from the silkscreening on the pcb. Only exception is the A-Trend Helios 3D that has a black pcb - but is still the same otherwise.

Orchid, miro, and Canopus are probably the only cards that don't follow the reference design.

Depends on the maker of a given motherboard. CPU sockets was placed anywere from the top to the bottom.
Another thing that they should have considered, was the placement of the headers for turbo and so on.
And not to speak of the placement of the heatsink for the voltage regulator transistor.

Anyway....
It was in the PC's infant, and stuff seems to have been done by the old "learning by doing" and "learning by own mistakes".
What in return is part of what makes pre-atx interesting to read about and tinker with. 😀

Looks like I need an Orchid Righteous 3d then. Too bad they seem to be fairly rare on eBay... Thanks for the input!

Skywell Magic 3D has a similar PCB size as the Orchid Righteous 3D. So that might be an option as well...

Does anyone have the exact length of the Orchid Righteous 3d and/or the Skywell Magic 3d? I want to make sure they will actually fit in my case before I buy one. Thanks!

Around 153,9 mm +- 5 mm depending on where the measurement begins (PCB or slot bracket)...

From the end of the board to the beginning of bracket.

Orchid... 16.4cm
Diamond... 17.6cm

Thanks!

Plasma wrote:

It is possible to position the CPU in corner of a baby AT board where it will not block any slots. See for example the A-Trend A […]
Show full quote

It is possible to position the CPU in corner of a baby AT board where it will not block any slots. See for example the A-Trend ATC-5000:

In fact I bought this board for that exact reason...

Only problem is that you can't use this board in a minitower (or many desktop cases) with any sizeable CPU heatsink. There's a reason A-Trend themselves moved it downwards on the successors to this board.

Why do you think that? I have this board in a minitower. Not many socket 7 heatsinks were larger than the CPU.