It depends on what kind of slot your riser card uses. If it's a proprietary slot that includes the signals for VLB, then you could get (or design) a new riser card to exploit that. If the riser just plugs into a standard ISA slot, then there's no way that it's going to work. Even if your board has integrated VLB graphics, there's no way to add external slots unless you can find a way to break out the signals.

As far as I'm aware, most of the LPX 486 boards like that do use VL-Bus for the graphics and often for the IDE too, maybe something like speedsys could confirm either way?

🤔 possibly, datasheet for the gfx chip would help also.

Anonymous Coward wrote on 2022-09-17, 01:01:

It depends on what kind of slot your riser card uses. If it's a proprietary slot that includes the signals for VLB, then you could get (or design) a new riser card to exploit that. If the riser just plugs into a standard ISA slot, then there's no way that it's going to work. Even if your board has integrated VLB graphics, there's no way to add external slots unless you can find a way to break out the signals.

That’s what I’m talking about doing. Breaking the signals out of the processor socket.

There are ISA riser cards with an integrated ISA-VLB bridge chip. There´s both ISA and VLB slots on the riser card so you can *use* VLB cards - but of course at ISA bus speed (or a little less).

Sersiously, this exists? Can you give an example?

Yes , they do. But I´ve only seen 2 or 3 on Ebay in all those years.

But now that I think about it it might have been riser cards with a PCI-VLB / VLB-PCI bridge chip. (I didn´t save any pictures...)

After seeing Atheatos "Landing the 386 to the moon! New VDD Mod! New Am386DX overclock record at 55Mhz." I commented asking about VLB mod
https://www.youtube.com/watch?v=c6CAnS1GbrQ&l … It0V9O7pc9xLEzg
>Atheatos Have you investigated modding motherboard to add external VLB slot? 😀 VLB does work on 386, and there were a few very rare 386/486 combo board with VLB. VLB wires VGA chip directly to CPU Address and Data buses, gives at the minimum a 2x write speed. You would probably need to solder it on the back of the board, or design CPU interposer. Dont know how >50MHz VLB would compare to 18MHz ISA, or if it could be made stable at that clock at all. Can you run VIDSPEED on your setup?

and got positive response 😀 later mentioned in another video as the ultimate mod https://youtu.be/9DD7V8_SxwU?t=887
so Atheatos is a good person to contact if you are looking for collaboration

majestyk wrote on 2022-09-17, 08:25:

There are ISA riser cards with an integrated ISA-VLB bridge chip. There´s both ISA and VLB slots on the riser card so you can *use* VLB cards - but of course at ISA bus speed (or a little less).

majestyk wrote on 2022-09-17, 08:48:

Yes , they do. But I´ve only seen 2 or 3 on Ebay in all those years.

But now that I think about it it might have been riser cards with a ISA-PCI bridge chip. (I didn´t save any pictures...)

I think those risers already have VLB signals wired, and use VLB-PCI bridge chip to provide PCI

majestyk wrote on 2022-09-17, 08:48:

Yes , they do. But I´ve only seen 2 or 3 on Ebay in all those years.

But now that I think about it it might have been riser cards with a ISA-PCI bridge chip. (I didn´t save any pictures...)

I think those risers already have VLB signals wired, and use VLB-PCI bridge chip to provide PCI
[/quote]

You are probably right - I had just corrected that after I wasn´t so sure anymore. What I remember for sure are ISA + VLB riser cards with a bridge chip onboard.

It seemed kind of retarded to strap VLB onto ISA, but it's technically possible...and dumber things have happened.
Even VLB riser with a PCI bridge would be cool to have. That makes a little more sense, and we know OPTi loved to use those bridge chips on their motherboards to fake PCI.

It would make more sense if you retrofit another 486 LPX board inside that Gateway, which actually has a real VLB slot on the riser, rather than struggling to tap into the CPU socket with adapters and make an unstable mess.

computerguy08 wrote on 2022-09-17, 09:29:

It would make more sense if you retrofit another 486 LPX board inside that Gateway, which actually has a real VLB slot on the riser, rather than struggling to tap into the CPU socket with adapters and make an unstable mess.

The length of the traces / wires would be critical and reduce maximum FSB. Riser cards with ISA + VLB slots in most cases only had one VLB slot - the one next to the socket.

Heck if we are gonna talk about bridges, may as well bridge straight to pci.

Is there a isa to pci bridge?

You cant go from ISA to VLB due to missing address lines, probably same problem with PCI. VLB is exposing raw 486 CPU Address and Data buses, even supports 32bit bus mastering. Imo there is nothing fake about VLB-PCI bridge, the only difference from more integrated solutions is not being located in same chip as memory controller and thus perhaps less optimized/slower? Afaik VLB itself is just a small arbitration circuit added to 486 bus.

VLB has a 16-bit operating mode. In this mode it only uses 24-bit addressing, just like ISA. You can find the odd 486SLC board here and there with VLB.
The bus mastering thing is a non issue, since only a handful of cards use it.
For me a VLB to ISA bridge would only be interesting in the sense that you could try out a few graphics cards that were not released as ISA versions...but probably the performance would be terrible and not worth it.

24bits of address bus would remove second, after speed, advantage of VLB graphics, the linear buffer.

You can still have a linear frame buffer, but like with ISA cards it's placed between 12 and 16MB, which is not ideal if you have 16MB installed.

Sphere478 wrote on 2022-09-17, 09:46:

Is there a isa to pci bridge?

What would be the point? You're still limited to ISA speed that's like 2MB/s at 8MHz.

Before VLB there were lots of local bus implementations. Some had a connector, most were just onboard and hooked the VGA to to local CPU bus. To do just the latter requires you to make an interposer board that can operate at 33MHz. Not an easy task. If you are trying to hook an existing vlb card to that CPU with wires I doubt it will work. Even a custom board that fits the CPU socket AND the correct location for a VLB slot may be too long from a latency standpoint. If you look at a typical VLB card the traces are usually pretty short, like an inch to 1-1/2. The CPU socket or chipset is usually under 2" from the furthest VLB slot. Oh and the PCShits with 8 VLB probably never worked properly.

If you have experience with doing a local bus video implementation you may have the electronics knowledge to attempt/diagnose this. But I still think it will be very difficult to get it to work.

Slow yes, the point: because we “maybe” can! Idea is to maybe put a voodoo or random Indescript 3d pci card in a late model isa system. That may have found its self with eisa or vl bus in another life and would have been able to do this. Performance hit still, sure. Fun, yes!

rasz_pl wrote on 2022-09-17, 10:34:

Afaik VLB itself is just a small arbitration circuit added to 486 bus.

The "small arbitration circuit" is the key to implementing local bus add-on cards (like VL, or earlier local busses). A local bus cards "takes over" the frontside bus if it is addressed and stops the chipset from processing the cycle initiated by the host processor. This makes the local bus card responsible for driving the data lines on a read cycle and generating the "ready" signal to finish the cycle. Without a way to take the local bus away from the chipset, the chipset and the add-on card will be in conflict on handling the cycle.

Modern 486 chipset have a dedicated input, typically called LDEV, to signal that a local bus card takes over the cycle. That pin is not on the 486 host bus, because it's an agreement between the chipset and the VL card, the host is not involved at all. Early 486 chipsets often support the 4167 coprocessor, which actually also is a local bus target. Depending on chipset design, the board decodes the C000'0000-DFFF'FFFF to the Weitek coprocessor internally (in that case, the Weitek support is of no use), while in other cases, an output pin of the Weitek processor indicating "chip selection" is used to tell the chipset that the Weitek takes over the cycle. That's exactly the same mechanism as is used for VL devices. I've got an early Opti Local Bus board that does exactly that: The LDEV pins of the 3 local bus slot and the Weitek select pin are fed into a 4-input AND (that's an OR for active low signals) gate, and the output is fed into the "weitek select" input of the chipset.

So if you want to add a VL slot to a board, and have it be connected to the FSB, you need a way to claim cycles from the chipset. You need to know how to interact with the specific chipset of the mainboard, the 486 data sheet won't help here. If there is online local bus graphics, you might be able to find the LDEV signal at the graphics chip even if you can't find a data sheet for the host chipset.