rasz_pl wrote on 2026-06-09, 18:32:

does the other one also map to 1GB?

Unfortunately I don't know how to check which address range is used by S3 Trio for VESA modes. Any ideas about the application which can check this?

The S3 VESA BIOS for the Trio64 does not support VESA 2.0 and thus has no LFB support. There is S3VBE20, but this is closed source software without any permission to create derivative works, and that software does not handle LFB for most VLB cards correctly. Possibly you can find some hint for educational use only on what's wrong with S3VBE20.

The S3 Trio64 driver for Windows does not attempt to use the LFB by default on VL cards, if I remember correctly. In any case, a sensible S3 Trio BIOS for a VL card will configure the LFB base address register of that card (on which a lot of bits are not acutally used by the hardware) to an address that is decoded to the LFB, so just reading that register might give an indication what address is used for the LFB. In case of the MK-765VL, as far as I know the address decoding is cloned from the STB card, so the LFB base address would be the same.

Hello,

I've written my experiences with the win95 drivers available for the s3 trio64v+ VLB that I could find here:

https://github.com/oerg866/TRW-BiosMods/wiki/ … o64V--VLB-Notes

There's a common theme with newer drivers that I can't really explain, which is that the mouse cursor tends to break.

These were tested with the TRW Prometheus card but like yours it is also a clone of the STB Powergraph.

mouse cursor is usually supported by using special dedicated hardware sprite. Maybe those drivers dont account(someone forgot and didnt test) for VLB registers being mapped in different way/under diff address?

edit: 🤣 here example just from yesterday Re: UniPCemu progress

I've been writing a Windows 98 driver from ground up for Trio64/Virge so lots of options to customise it. If it is something you are interested in playing with let me know. I will probably need to give you a small app first to pull some of the hardware specific info so I can compile a version that works on your card.

EDIT: Here is a link, happy to some work on it https://github.com/michaeldale/velocity9x if you can run the dos exe I can collect info about your card and try and build a supported driver https://github.com/michaeldale/velocity9x/rel … s/tag/survey-v1

mwdmeyer wrote on 2026-08-15, 03:37:

I've been writing a Windows 98 driver from ground up for Trio64/Virge so lots of options to customise it. If it is something you are interested in playing with let me know. I will probably need to give you a small app first to pull some of the hardware specific info so I can compile a version that works on your card.

EDIT: Here is a link, happy to some work on it https://github.com/michaeldale/velocity9x if you can run the dos exe I can collect info about your card and try and build a supported driver https://github.com/michaeldale/velocity9x/rel … s/tag/survey-v1

This is very cool!!! Perhaps with the hardware workarounds, an accelerated virge driver for VLB is possible. One can dream 😀

I don't understand how Velocity98 would helps here on a real VLB card. From what I read the whole page at https://github.com/michaeldale/velocity9x I don't see any mention it would be intended to run on VLB cards as it relies on PCI VID:DID detection or VBE 2.0 videobios. Even it seems to be targeted/tested on VM not a real PC?

RayeR wrote on 2026-08-18, 00:44:

I don't understand how Velocity98 would helps here on a real VLB card. From what I read the whole page at https://github.com/michaeldale/velocity9x I don't see any mention it would be intended to run on VLB cards as it relies on PCI VID:DID detection or VBE 2.0 videobios. Even it seems to be targeted/tested on VM not a real PC?

Version 0.4 adds VBE/VideoBios work. I still need someone to run my survey tool on the VLB card so I can start adding support for that card. VLB support can then be added afterwards, I have ground work ready for it.

I've tested on RagePro on physical hardware and have run-up work in progress for GMA950, plus Matrox Mill 2 and more testing on S3. Lots of work to do but I can get it done 😀

I have a 486 with S3 Trio64 VLB, so I will be adding that myself later, but no Virge VLB, I will probably need physical hardware to fix the faulty Virge VLB MMIO, but not sure yet......

I have 765VL with Trio64V+ on a 486 system so I can physically test it. But seems to me that also your survey relies heavily on PCI BIOS or PCI scanning while on VLB system it couldn't find anythinh as VLB cards and chipsets didn't have any PCI like id.system. I could assembly my 765VL with a Virge chip but I was warned by this thread about many unresolved issues with it so I decided to solder Trio64 instead...

Thanks I will build an update survey tool soon with VLB support.

Latest version 0.4.2 is fully tested on real hardware with S3 Trio64 PCI. I will be breaking out the VLB cards soon.

ok, let me know when available or i can run survey on a few 486 boards to collect more info when will have vlb support...

mwdmeyer wrote on 2026-08-15, 03:37:

I've been writing a Windows 98 driver from ground up for Trio64/Virge so lots of options to customise it. If it is something you are interested in playing with let me know. I will probably need to give you a small app first to pull some of the hardware specific info so I can compile a version that works on your card.

EDIT: Here is a link, happy to some work on it https://github.com/michaeldale/velocity9x if you can run the dos exe I can collect info about your card and try and build a supported driver https://github.com/michaeldale/velocity9x/rel … s/tag/survey-v1

As far as I know 2D acceleration is very different on Trio64 and Virge because the latter one has removed the legacy S3 2D acceleration and uses its 3D engine instead to perform 2D acceleration.
So, will your driver be a plain frame buffer driver or will it implement 2D acceleration?

Disruptor wrote on 2026-08-20, 08:40:

mwdmeyer wrote on 2026-08-15, 03:37:

I've been writing a Windows 98 driver from ground up for Trio64/Virge so lots of options to customise it. If it is something you are interested in playing with let me know. I will probably need to give you a small app first to pull some of the hardware specific info so I can compile a version that works on your card.

EDIT: Here is a link, happy to some work on it https://github.com/michaeldale/velocity9x if you can run the dos exe I can collect info about your card and try and build a supported driver https://github.com/michaeldale/velocity9x/rel … s/tag/survey-v1

As far as I know 2D acceleration is very different on Trio64 and Virge because the latter one has removed the legacy S3 2D acceleration and uses its 3D engine instead to perform 2D acceleration.
So, will your driver be a plain frame buffer driver or will it implement 2D acceleration?

Yes I believe that is all correct and the driver for hw 2d uses s3d on the virge only.

I do expect adding 2d acceleration on trio should be possible.

Hi Guys,

I now have my driver working on my 486DX/4 with Windows 95 + 2MB S3 Trio64 VLB (real hardware).

I have also reviewed the MMIO issue and believe I can work around that. At this stage I feel it is possible to get the driver with S3D acceleration working. As I don't have any hardware testing this will be difficult, I can build a 86box VM with fake VLB card, so I might do that, but I will wait until someone shows interest to test on their hardware.

I would be especially interested to hear from anyone with:

• A working ViRGE 325 VLB setup
• Details of the video BIOS and installed memory
• The delayed-SAUP2/74F74 or 74ACT74 modification, if fitted
• Windows 95 or Windows 98 available for testing
• Willingness to run a survey tool and share its text report

Thanks

I know from my pal mkarcher who has tried to do a workaround for the MMIO issue for his VLB ViRGE 325 and he failed. He even tried to use a CPLD as workaround but he failed too.
I wish you good luck.
He came to the conclusion that it had a good reason for S3 to not bring a VLB ViRGE.
And on the ViRGEs following the 325 there was no VLB interface at all.

Let me clarify: ViRGE 325 works flawless as framebuffer device in VLB mode. But the MMIO issue stopped his attempts on adapting an accelerated Windows driver.
However he could enable Terminal Velocity's S3D acceleration for his VLB 325. But he could avoid the usage of MMIO in that special case.

Disruptor wrote on 2026-08-22, 06:45:

I know from my pal mkarcher who has tried to do a workaround for the MMIO issue for his VLB ViRGE 325 and he failed. He even tri […]
Show full quote

I know from my pal mkarcher who has tried to do a workaround for the MMIO issue for his VLB ViRGE 325 and he failed. He even tried to use a CPLD as workaround but he failed too.
I wish you good luck.
He came to the conclusion that it had a good reason for S3 to not bring a VLB ViRGE.
And on the ViRGEs following the 325 there was no VLB interface at all.

Let me clarify: ViRGE 325 works flawless as framebuffer device in VLB mode. But the MMIO issue stopped his attempts on adapting an accelerated Windows driver.
However he could enable Terminal Velocity's S3D acceleration for his VLB 325. But he could avoid the usage of MMIO in that special case.

Hi Disruptor, mkarcher,

Do you know for the MMIO issue if you wrote a driver (and if so MMIO at A0000 is itself broken over VLB) or attempted to work around with the existing Virge PCI MMIO driver?

Trio style MMIO puts its acceleration registers into one of two memory ranges CR53 bit 5:

A: the registers appear at 0xA0000–0xAFFFF (the 64 KiB segment normally used for VGA graphics memory).
B: the registers appear at 0xB8000–0xBFFFF (the text-mode segment, 32 KiB).

If A0000 is broken then do you know, which access patterns trigger the corruption? Reads vs writes, back-to-back cycles, specific timing?

I can do some extra testing on Terminal Velocity if unsure, because there might be a direction that way.

It especially approaches on switching a Windows 9x' DOS window from text mode to graphics mode and vice versa - but not only! Especially in the latter case it is not so easy to find a pattern where you see the fails.
We currently cannot tell more about it.

Cool based on this info there are options that may work in the driver, when I get time I will do further testing on this, but probably only if there is someone with one of these boards that can do testing for me once I release a test tool.

mwdmeyer wrote on 2026-08-23, 05:40:

Cool based on this info there are options that may work in the driver, when I get time I will do further testing on this, but probably only if there is someone with one of these boards that can do testing for me once I release a test tool.

If this was 2-3 years ago, I'd have been willing. I had a Virge 325 chip on a VLB card, which I use for some testing with mkarcher. I left it on there for about a year, hoping for a breakthrough. I think we used a modified Windows 95 driver which worked in w95, but I don't recall the details. I have since removed the Virge 325 in favour of the Trio 64V+. The neat factor for a Virge chip on a VLB card would be for S3D game play and 4 MB support.