data9791 wrote on 2025-01-26, 21:56:

Can anyone confirm the most powerful CPU these boards can use? Mine is a REV: CA
Apparently my revision supports MMX, https://theretroweb.com/motherboards/s/dfi-g5 … pv-rev.-c#chips shows the MMX OverDrive 200 (PODPMT66X200) being the best officially supported. Are there any faster CPUs these boards support that aren't officially documented?

The ultimate CPU upgrade for boards like this is an Evergreen (or similar) voltage adapter that allows you to use a K6-3 or similar CPU. I have a PCChips M520 that's similar to this board and with the 75 MHz bus speed it runs at 450 MHz. You would also need a patched BIOS to properly support the CPU. I wouldn't recommend this as those adapters are hard to find and expensive these days, but it is something that could be done.

Manuals dating to the release of a motherboard are rarely a good indication of the maximum capabilities as they can - by definition - only tell you about CPUs that were available at time of writing, whereas newer CPUs may work out of the box, may work after a BIOS update or it might be possible to get to work with various hacks.

So let's see what we have here...

This is a pretty generic i430VX board. In retrospect the i430VX platform is generally seen as pretty underwhelming, only adding (primitive) SDRAM onto the featureset of the older (and much more revolutionary) i430FX without addressing any of its shortcuts (mainly around caching). In its day though, it was the fastest money could buy and it wasn't even particularly expensive (unlike its i430HX sibling). So potentially good for nostalgic reasons. Not great for pushing the envelope though.

In terms of CPU support, i430VX gives you a split voltage plane and official support for 66MHZ FSB, but the silicon will usually handle 75MHz if there is circuitry present that can supply it.

This board has documented voltage settings down to 2.5V and documented FSB settings of 50, 60 an 66MHz.

So what could run with those? 2.5V is within tolerance for the K6-3 CPU. It re-maps 2.0x multiplier to 6.0x, so you could run at 6x0x66MHz=400MHz with a K6-3 without modifications, at least, if the BIOS accepts it.

Now, could you push it further? For that, we'd need some detailed photographs of your specific board, in particular the area around the jumpers for bus speed and voltage.

The two jumpers for bus speed give four logical options even though only three are documented. Depending on the PLL chip they connect to, the fourth option would either be 55MHz (not very useful) or 75MHz (which gives you 12.5% more speed vs 66MHz, although it will also overclock PCI and possibly ISA buses at the same time)..
As for voltage, that looks more complex. Sometimes using a resistor instead of a jumper to connect pins will give you a lower voltage. If you can get it down to 2.2V, you could use ubiquitous K6-2 CPUs. Not as fast as K6-3, but much easier/cheaper to find. At 2.0V you could safely try K6-2+/K6-3+, although it's unlikely they would work without BIOS modification. But even if you could get the right voltage, it's an open question whether the VRM/MOSFETs could handle the added current of these faster CPUs on lower voltage. So pics of the components needed before definite advice here.

However maybe a step back: this is a pretty unremarkable board and by no means well-suited to pushing the envelope when it comes to CPUs. Why do you want to try that with this particular board? The answer to that might also let us give you more relevant tips.

Wow, thanks you so much! I'll try to get a picture of my board soon. Its from my first video editing system from the late 90s. I want to rebuild it to be a bit quicker and finish editing an little indie film I started on it decades ago but never finished. Right now it's equipped with a Cyrix 150mhz CPU.I figured I max the RAM, get the best CPU within reason. The video capture card is a Pinnacle DC10+.

data9791 wrote on 2025-01-27, 01:59:

Wow, thanks you so much! I'll try to get a picture of my board soon. Its from my first video editing system from the late 90s. I want to rebuild it to be a bit quicker and finish editing an little indie film I started on it decades ago but never finished. Right now it's equipped with a Cyrix 150mhz CPU.I figured I max the RAM, get the best CPU within reason. The video capture card is a Pinnacle DC10+.

I saw that you would like to update the CPU, and that you currently have a Cyrix 150MHz, but those CPUs have a PR, that is, the actual actual frequency is for the PR150 120MHz, this means that with only 120MHz in some applications it succeeds (almost?) To match a Pentium 150.
However, for a motherboard like this, personally I would put a Pentium MMX 166 MHz, if you need more speed, you can if the CPU allows it change the multi, and set it to 200 MHz, in my opinion for the work you have to do, the Pentium MMX is better, the FPU is 🫡 compared to that of the Cyrix, it is better, I think it is more than double, but this depends on the program, whether or not it rests on the FPU.
The maximum could be a 233 MMX, but we are only talking about Intel, but it depends on the voltage regulators, if they are linear, already at 200 MHz they start to heat up, if they are switching, the speech changes, you can go even further without any problem.

Looking at the manual and the retro web site, the regulators are linear, you can see in the pdf page, that the air flow of the CPU fan, goes directly to the voltage regulators.
So for me i166 MMX, it's the safest and most reliable choice, you can go further, but not much, an i233 MMX in that card works, but I don't know 🤔 what temperatures you would have on the two heatsinks of the linear regulators, I think they will be very hot.

I've taken a closer look at the pics on TheRetroWeb and with a bit of CSI-fu have been able to identify the PLL, it's an ICS9159M, whose datasheet can be found here

Bad news regarding 75MHz on page 1: that fourth jumper setting is 55MHz, not 75MHz. So without major surgery this board isn't going to go faster than 66MHz FSB - at least, the board whose pic is on TheRetroWeb won't. You can confirm if this also applies to yours by looking at the small IC between the SB82437VX northbridge and the 14.318MHz crystal (and top PCI slot). Replacing the crystal with a faster one might allow overclocking (eg. a 16.2MHz crystal would give you 75MHz with the '66MHz' setting) - but that's the sort of stuff I'd do with a board with zero sentimental value and it sounds like this one does have such value for you.

Edit:
Regarding the regulator, agreed that linear regulators aren't great. For info on how much headroom you might have, we need the model of the regulator FETs, i.e. the two three-pinned components bolted to the heatsinks on the bottom of the board.

Assuming P166MMX to be the max, it uses 13.1W at 2.8V, so draws 4.7A, which is (maybe not coincidentally) the same as the P200 non-MMX draws (15.5W at 3.3V). Extrapolating that, it means that at 2.5V you can only draw 11.75W. A K6-3 draws 26.8W, so you can complete forget that without an interposer with its own VRM powered directly from the power supply.

There is one other higher-clocked option that might work without burning out that regulator and that's the IDT Winchip, or rather one of the later Winchip2 or 3 CPUs. Up to the 2A they are single-voltage running at 3.3V, the 2B and 3 are dual-voltage running at 2.8V. The Socket 7 spec defines three BF (Bus Frequency) pins, which - confusingly - regulate the multiplier This board implements the first two, BF0 and BF1, which leaves BF2 permanently high. In that state the Winchip1 CPUs can be run at 4x66=266MHz (1.5x is remapped to 4x), which might be faster than a P166MMX, although I'd doubt it. It would only draw about 13W at 3.3V, so in terms of power it would be safe.
Winchip 2 and 3 CPUs have a different multiplier scheme, with ratios up to 4.5x, but to get that you would have to connect the BF2 pin to ground to pull it low and then select 2.0x for 4.5x. Using that you could run a suitable Winchip 2A or Winchip 3 at 300MHz (respectively 16W at 3.3V, which is only marginallyt above the P200, or 8.4W at 2.8W, which is far lower than the P166MMX). At those speeds they should be comfortably faster than a P166MMX or P200 non-MMX.