As told this is a prototype based on datasheet.

jmarsh wrote on 2026-08-27, 13:18:

If you set bits 4-7 of SR7 to a value other than 0, does the LFB appear at that MB boundary?

Edit: from the pics in the first post, MEMR# and MEMW# are definitely not connected. So no memory access beyond 1MB is going to work.

the design is already updated to using the lfb mem r/w pins so this is already patched in - still did not made any tests to see if it really matters but better to have it than not
thanks rasz_pl for the heads up 😀

After seeing datasheet with a bit more attention, I don't see much advantage in ISA bus for Linear Addressing, because it's only working if you have 14mb or less. Either way we will use jumpers for this situation, so you can choose if you want to use or not.

Marmes wrote on 2026-08-31, 09:30:

it's only working if you have 14mb or less.

thats where BIOS "Memory hole at 15/16MB" comes in

Exactly.

Interesting that the PC98(21) series of computers does it almost exactly the same.

They have a (non-VGA) framebuffer that is accessed much the same way (15-16mb), but, if that region is not available, the bios instead relocates it to 4095mb.

Either way we made tests with a Diamond Speedstar 64 ISA card on a p133 with 8mb of ram with univbe 6.7 and LFB activated at a resolution of 640x480 16bpp and the interesting thing is in PCPBENCH with Nolinear was faster than LFB. Don't ask me why, in SVPRO we just followed the datasheet so no surprises are found.
Can anyone point me some software/game that uses LFB so I can compare both? This is just out of curiosity, because board will have both options.

Tests made with modded SV on pentium 166 motherboard .
With 14-15mb memory hole in MB bios and LFB activated by univbe and 16 MB of ram, LFB not available.
With 15-16mb memory hole in MB bios and LFB activated by univbe and 16 MB of ram, LFB not available.
With 14-15mb memory hole in MB bios and LFB activated by univbe and 8 MB of ram, LFB available and graphics card configured for max performance.
With 15-16mb memory hole in MB bios and LFB activated by univbe and 8 MB of ram, LFB available and graphics card configured for max performance.
In PCPBENCH at 640x480 8Bpp with LFB I get around 7FPS , with no linear buffer I get 10.2FPS
Isn't LFB supposed to be faster?
What am I doing wrong?

No idea. Where is the LFB being mapped to and how big is it? I suspect 2MB vram makes it not fit in standard 1MB hole BIOS is making. Maybe there is some merit to adding a way of limiting video memory to 1MB.
I would try an actual game, like Duke3D in 640x480 with https://www.moddb.com/games/duke-nukem-3d/dow … mark-for-ms-dos to get detailed speed stats.
Experiments with fastdoom show some chips are just faster when operating in VBE2 modes Re: FastDoom. A new Doom port for DOS, optimized to be as fast as possible for 386/486 personal computers!

it can access in 64kb banks and this is supposed to make it slower. But it's not.

I assume that copying data directly to an LFB in DRAM is throttled by the ISA bus, and therefore the advantage of a LFB is negated.

Using a Pentium also means you have to deal PCI to ISA bridge shenanigans (I think).

In ISA bus this may be not so important as there is no big or any difference. Maybe on PCI or VLB is another story.

This might be out of scope for the project, but was there any consideration to use a DVI-I port for optional LCD support instead of just VGA?

I can't even think of a chipset that would make sense which would support that. Feature connector is there to go nuts with. Or make a pin compatible FPGA implementation of a DOS era 2D SVGA and put in DVI, hardware Mpeg etc.

Isn't that what a DVI transmitter does?

https://theretroweb.com/chips/5750

we have not considered any kind of DVI connector (with digital capabilities) for this since there is no direct support from CL5434
we only really added the feature connector because the CRT Terminator exists tbh

One question.
It is said the SonicVision PRO is for older 386/486 builds.

What is the reason it is not recommended for let's say a Pentium MMX 233MHz build?

Let me preface my personal opinion with an important point: any system (let's say a 386 or higher, just to ensure the card's resources are fully utilized) with a fully functional ISA slot—meaning one assigned the proper IRQ and DMA resources—can use the card correctly.

That said, a system with a Pentium 233 MMX will certainly have a sufficient number of PCI and ISA slots; this allows you to use a higher-performance PCI video card that remains highly compatible with DOS titles, install a faster PCI network card (perhaps a Wi-Fi card with Windows 9x drivers), and then use the remaining PCI and ISA slots for other peripherals (including any sound cards).

Simply out, on a Pentium 233 MMX you can use higher-performance hardware that offers greater capabilities and convenience (such as a Wi-Fi network card).

I’d also like to add that, for old 386 and 486 systems, it is difficult (and extremely expensive) nowadays to find a high-performance video card (like the CL5434 with 2MB of RAM on ISA) or a network card with an RJ45 port (on ISA), so the SonicVision Pro is a welcome arrival!

As for the card's audio component, I consider it a nice bonus (our friends have certainly spoiled us with the likes of the Orpheus, MK, and GoldLib) and it’s a feature I’m quite happy with!

crusher wrote on 2026-09-15, 05:53:

One question.
It is said the SonicVision PRO is for older 386/486 builds.

What is the reason it is not recommended for let's say a Pentium MMX 233MHz build?

It would bottleneck performance on the Pentium due to the slower ISA bus.

Even on 486 machines, 32 bit VLB and PCI slots were introduced to provide greater bandwidth and performance for expansion cards.

In one sentence, the GD5434 is among the fastest video chipsets you can put in an ISA slot, but remains much slower than the slowest video card you can put in a PCI slot.

This is true for even slower CPU and older gen PCI chipsets. I had an Orchid Kelvin ISA with GD5434 in a Pentium 60 for a few years, due to all my PCI video cards being otherwise occupied. Then as I got more to play with, I had a GD5430 PCI, now these are only 32bit vs 64bit, and some 30% slower than the rest of the GD543x series, which hang in the average end of 2D cards with others like TGUI9680 etc. The only one comparably as bad which passed through my hands was a Trident, not sure if it was late TVGA or early TGUI 9200 maybe. Well the point is, the 5430 is no fireball. I knew I had my best ISA card in that system, did I want to curse it with my worst PCI?? Well I didn't for a while until I needed the ISA elsewhere, then it was kind of a "sorry brah, gotta shuffle my cards, you get the deuce..." BUT, it turned out noticeably faster than the 5434 ISA. So yeah, 32 bit on 32 bit beats 64bit on 16bit.

Would be cool sometime to get a VLB/ISA/PCI system and cross benchmark the VLB/ISA/PCI versions of 5430 and 5434 and also others that came in at least two flavours like the trident 9440 and also earlier chips 5428 on ISA and VLB and get a thorough picture of how each bus performs for graphics and quantify how faster chipsets react to the slower buses.

have been testing the latest prototype and so far found no issues... for your viewing pleasure I attach some shots of the drivers for win3.x
I have uploaded some more dos shots on the card's page , just because they look cool and there is no other real info yet 🤣

edit:
I do not understand why the AUDIOSESSION program complains about a missing SBPRO.DLL when it launches, the driver files were downloaded from opti's archived page...