eton975 wrote on 2022-10-21, 10:21:

Hi, last month I received this motherboard paired with a Pentium MMX 200. I have since updated the bios with Jan's modded J.2 ve […]
Show full quote

Hi, last month I received this motherboard paired with a Pentium MMX 200. I have since updated the bios with Jan's modded J.2 version. I then upgraded it with a K6-2/500AFX (running at only 400MHz due to 66MHz bus limitations and the 6x multiplier limit). I am using 128MB SDRAM (2x64MB high density sticks) and a Corsair CX500 with 20A on the 5V rail as PSU. OS is Windows 98 SE.

Currently I am using a Matrox Millennium II paired with an Orchid 3dfx Voodoo for video. While a razor sharp and fast 2D card, this combo is very lacking when it comes to later 3D games. I would like to keep the Voodoo as a secondary card for games like EF-2000, Tomb Raider etc that use Glide or 3dfx-specific programming, but I would like to change the Matrox out for an ATI or NVIDIA . I would be playing games like Quake II or III, American McGee's Alice.

I tried a HIS Radeon 9250 (expensive!) with the universal PCI connector and it would not POST in this system. It is not a faulty graphics card, it works fine under Windows 10 in my more modern ASRock 980DE3/U3S3. So I am pretty reluctant to try another Radeon 9250 or 9000 series unless this one just doesn't have the right voltage regulation circuitry or something to step down 5V to 3.3V.

Currently I have my eyes set on the Radeon 7000 PCI for this board, but if this is not an option I am open to the NVIDIA TNT2 M64 or other options. Does anyone here know what the fastest and most compatible option is for a K6-2 at 400MHz? Does the card vendor or amount of RAM matter? (besides the ones with a BIOS for Macs of course) Of course, it can't draw too much on the 5V or 3.3V rail, but I am not running a power-hungry Athlon to be fair.

If nobody here knows I have ordered a cheap POST card off Ebay with an LED that indicates 3.3V support, but it's nice to make sure before I spend nearly 100 more AUD on a graphics card that may not be compatible with this motherboard's PCI voltages or PCI spec revision.

As for why I am doing this when I could run the games on a modern system easily, it's just fun getting stuff to work and you also get a unique feel running at a lower framerate at a natively supported lower resolution. Sounds odd, I know, but this is VOGONS where people could probably just use emulators for most things if they didn't care about authenticity and the unique soundfonts of old sound cards.

Thanks for taking the time to read this and perhaps respond.

I'm running a Voodoo3 with a 430TX-based ASUS TX97-XE and AMD K6-III+ and it's a very good pairing.

Since V3's are now priced like dinosaur-eggs, a PCI based RivaTNT/TNT2 might be a better option. It would be really close to the Voodoo 3 performance-wise and have compatible drivers game/time-wise. A TNT2 M64-version is basically as fast as first gen TNT, maybe a little slower - so not as fast as a V3.
A Matrox Millennium G400 is a little faster than V3/TNT2, but is fairly hard to find. The G450 is a clock-reduced version - roughly in the same performance category though.
A faster card like a GeForce 2 MX PCI might still work really well for a system like yours. Some ATi cards pre-Radeon might be comparable in performance, but the drivers back then were so hit-or-miss I hesitate to recommend them. Radeons on the other hand are starting to be too modern (as you're finding) and will likely be bottlenecked severely by the CPU even if you get your system to boot.

So the POST card arrived and it appears to indeed be a 5V-only board, when the card is stuck in the bottom PCI slot.

I got a PCI TNT2M64 and it works well with the NVIDIA Detonator PCI 2.08 drivers.

eton975 wrote on 2022-10-21, 10:21:

[...]

I would be playing games like Quake II or III, American McGee's Alice.

With those games, your VGA / 3D is the last of your worries 😮

Quake 2 might perform acceptably with a good 3D option, but Quake 3... Take a look here:
Sharing my benchmark results on a AMD k6-2 with various AGP cards

That's 500MHz on a 100MHz FSB board, so CPU is clocked 25% faster and cache, bus & memory 33% faster than in your case, and video goes via AGP, not PCI - It's entirely likey your results will be at least 25% slower. In these benchmarks the results are competely CPU-bottlenecked at anything over a TNT2-M64 and even with lowest settings he only hits 39fps. If scaling is linear you can expect max 30fps regardless of 3D option and I fear scaling may be worse than linear.

I fully realize back in the day we accepted lower frame rates than now (I ran UT on a K6-2 350 with ATi Rage Pro Turbo when I couldn't afford anything better. "It ran" is the nicest thing I can say about the experience), but with the luxury of hindsite - and over two decades of increasingly smooth framerates - I really don't think you will be doing yourself any favours with making this a period-correct low-end upgrade build, the sort of thing someone without enough money for a full system replacement in 1999 might have bought. If you want to play those games without ruining your nostalgia, go for a faster build (I find my Tualatin 1400S still a bit lacking in Q3A - that game really likes fast CPUs) and/or dedicate this machine to a really good experience on DOS or Windows 95 stuff.

A PCI MX4000 with the 5V slot type unfortunately does NOT seem to work in the Titanium IB+...

eton975 wrote on 2025-11-13, 04:52:

A PCI MX4000 with the 5V slot type unfortunately does NOT seem to work in the Titanium IB+...

Do you have a picture of the card? Often late model PCI cards might have the 5v notch, but when you look at the card, you will see a spot for a voltage regulator that wasn't unpopulated. That's a strong hint that it could have had 5v support, but to save some money, it doesn't. Other times it seem like really new PCI cards might have a BIOS incompatibility with pre AGP motherboards.

Which is a shame because an MX4000 would likely be a very good performer on that system.

douglar wrote on 2025-11-13, 17:56:

eton975 wrote on 2025-11-13, 04:52:

A PCI MX4000 with the 5V slot type unfortunately does NOT seem to work in the Titanium IB+...

Do you have a picture of the card? Often late model PCI cards might have the 5v notch, but when you look at the card, you will see a spot for a voltage regulator that wasn't unpopulated. That's a strong hint that it could have had 5v support, but to save some money, it doesn't. Other times it seem like really new PCI cards might have a BIOS incompatibility with pre AGP motherboards.

Which is a shame because an MX4000 would likely be a very good performer on that system.

Yes, I have several pictures from the eBay seller. Images attached.

I failed with a Matrox G450 PCI. J.2 BIOS. (Not yet have had time to upgrade to J.3)
But the upgrade from a K6-2+ 533 @ 400 to a K6-3 400 was fine.

Also have some pictures of a Radeon Mobility 7500 PCI 128MB with the 5V keyed slot, that also doesn't work in the Titanium IB+ but does in a P3 Advance 10F system. Any advice on what kind of regulators and capacitors I need to add?

leonardo wrote on 2022-10-21, 11:36:

Since V3's are now priced like dinosaur-eggs, a PCI based RivaTNT/TNT2 might be a better option.

It's funny to see that happen (even with the AGP version). For years everyone shit all over the Voodoo3.

Disruptor wrote on 2025-11-13, 22:06:

I failed with a Matrox G450 PCI. J.2 BIOS. (Not yet have had time to upgrade to J.3)
But the upgrade from a K6-2+ 533 @ 400 to a K6-3 400 was fine.

Ok, great vogons user mkarcher has patched Jan's J.3 BIOS for me.
It now uses a FSB of 66 MHz during the whole initialisation.
Before it tried a safe boot at a FSB of 50 MHz - until the configuration from the CMOS was processed.
That result was a 25 MHz PCI clock which is in PCI specification but definitively too low for the G450 PCI.
Now my Matrox G450 PCI boots fine.

So don't use mkarcher's patch when you use Pentium 75, 90, 120, 150 or other processors that require a FSB below 66 MHz like 60 MHz or 50 MHz.

Disruptor wrote on 2026-02-26, 10:27:

Ok, great vogons user mkarcher has patched Jan's J.3 BIOS for me. If now sets or leaves the PCI divider at 1:2 (=33 MHz) during […]
Show full quote

Disruptor wrote on 2025-11-13, 22:06:

I failed with a Matrox G450 PCI. J.2 BIOS. (Not yet have had time to upgrade to J.3)
But the upgrade from a K6-2+ 533 @ 400 to a K6-3 400 was fine.

Ok, great vogons user mkarcher has patched Jan's J.3 BIOS for me.
If now sets or leaves the PCI divider at 1:2 (=33 MHz) during the whole initialisation.
Before it was likely lower (1:3?), on a safe state even for users that overclock to 75 or 83 MHz - until the configuration from the CMOS was processed.
That result was a 22 MHz clock which is definitively too low for the G450 PCI.
Now my Matrox G450 PCI boots fine.

That's pretty cool. This is one of my favorite boards. I'll have to try out the BIOS.

The VGA signal quality from the G450 card is gorgeous, so I can understand the attraction if your mainly doing 2d stuff and directx <= 6 stuff.

Are you doing any newer 3d stuff with this card? I remember seeing a performance drop in 3d titles when going from the G450 AGP to the G450 PCI. I suspect it is because the G450 PCI had that unique PCI-AGP bridge chip.

Please find attached a version of the Titanium I B+ that initializes the on-board components in a slightly different order.

The original BIOS first performs a device scan and assigns resources to the PCI bus before it initializes the clock chip. This is theoretically a good idea, because the SETUP contains an option to disable clock to unused PCI and SDRAM slots (to mitigate EMI). So it makes a lot of sense to perform the PCI device scan before the clock chip initialization. Specifically for this BIOS, it is pointless, though. This BIOS supports clock disabling only in the case that all RAM is EDO. In that case, the clock is shut down from all DIMM slots. There is no code to disable PCI clocks (although the clock chip has enough dedicated PCI clock output pins that a per-slot PCI clock could be implemented, and due to driving/loading concerns, it likely is). So for this BIOS, initializing the clock chip before the PCI bus does not limit any functionality.

The advantage of initializing the clock chip first is that the clock chip is swiched from FSB50 to FSB66 before the PCI cards are initialized. The G450 PCI made trouble with initialization at 25 MHz (PCI is alway half of the FSB), but works perfectly at 33MHz. If there is nothing more to the story than a PCI clock of 25 MHz, the G450 is non-conforming to at least some PCI specification versions. I'm unsure whether this is true for all versions, but originally, PCI was specified as 25 to 33 MHz.

The appended "M" to the filename is due to my username being "mkarcher". It is based on Jan's J3 BIOS.