I have done something similar on a Biostar P1 board with intel FX chipset. I discovered, that it not only provided faster settings than documented, but also a slower one. That board did allow my P1 166 to run between 25 and 200 MHz. Sweet till dosbox caught up.

A long time ago I did similar tests on my PVI-486SP3 and found that the FSB can be lowered to 20 and 16 MHz. Since this board has no dividers except for ISA, this results in the VLB/PCI buses running out of spec (VLB: 20-50 MHz, PCI: 25-33 MHz).
With PCI cards installed, the 16 MHz setting did not work reliably, ISA/VLB cards were fine.

Do you happen to have those jumper settings for the PVI-486SP3? I happen to have that same board (not sure what revision I got though) and would love being able to get it to 16mhz!
I also just discovered yesterday that that's my only 486 board that supports a PS/2 mouse!

I have a couple socket 4 boards but they are the Intel made ones which only come with a jumper for 60 and 66 and my only socket 5 doesn't even have any cache chips on it, only has the empty spots.

^
>> Do you happen to have those jumper settings for the PVI-486SP3?

*bump* 😀

retro games 100 wrote:

^
>> Do you happen to have those jumper settings for the PVI-486SP3?

*bump* 😀

I couldn't. When I went to test my SP3 it blew up -_-
Perhaps if I replace/remove the cap that blew up it might boot again though.

My "bump" was to ask your question again, to 5u3. 😀

Oh sorry, I seem to have overlooked this thread (sometimes the forum software "forgets" to highlight new posts).
Unfortunately I didn't write down the jumper settings for 16 and 20 MHz, but I will try to find them out in the evening. Stay tuned...

^ Awesome, thanks very much, and I understand what you mean about new topics on the forum. Actually, speaking of how the forum works, it would be great to search for attachments. Over the years, many extremely useful attachments have been included in to people's posts. I would like to be able to browse these, either all of them, or by some kind of filtered search. 😀

OK, here are some results from my PVI-486SP3: This thing can do 8 MHz! 🤣

CPU: Intel 486DX-50
Board revision: 1.8
Clock chip: VT8228 (if you have a board with the AV9155, settings are different)

1 MHz   16   08   40   20   50   25   40   33   48
2-------------------------------------------------
3JP25  2-3  1-2  2-3  1-2  2-3  1-2  1-2  1-2  1-2
4JP26  2-3  2-3  1-2  1-2  2-3  2-3  1-2  2-3  1-2
5JP27  2-3  2-3  2-3  2-3  1-2  1-2  1-2  2-3  2-3
6JP28  2-3  2-3  2-3  2-3  2-3  2-3  2-3  1-2  1-2

The BIOS displays the wrong speed (16 MHz) when running at 8 MHz, and at 48 MHz it shows either 40 or 50 MHz.

Have fun with this!
Btw, there is another mode at 2-3 2-3 1-2 1-2, but it is unstable (and VERY slow).

5u3 wrote:

...
Have fun with this!
...

Oh I will! 😁 Thanks very much for this info!!!

Yay, now my super slow SX-20 CPU boots at the correct 20 MHz speed! 😀 I "raced it" against an underclocked SX-33 running at 25 MHz using 3DBench, and the scores are:

SX-33 @ 25 MHz = 20.0 // Equivalent to a very slow 486
SX-20 @ 20 MHz = 15.8 // Equivalent to a 386 DX 33/40

Therefore, the SX-20 is 21% slower than a chip running at 25 MHz. This test set up consisted of all the mobo's BIOS settings set to their fastest values. BTW 5u3, please can you double-check your most excellent bus speed jumper table? I see two column entries with the same heading value of 40. Thanks a lot.

retro games 100 wrote:

BTW 5u3, please can you double-check your most excellent bus speed jumper table? I see two column entries with the same heading value of 40.

Sorry, I didn't reformat the table, this is directly from my testing notes, so it may be a bit confusing to read.
40 MHz shows up twice because there are two settings to get 40 MHz from the clock chip.

Meanwhile I got the last setting I mentioned above (2-3 2-3 1-2 1-2) running. It required disabling the caches and forcing the ISA bus clock to 7.159 MHz in the BIOS setup.
It's another 8 MHz setting. I couldn't resist running Speedsys: 😀

It took the CPU more than one and a half hours to draw what seems to be the most pitiful result for a 486 ever. 🤣

retro games 100 wrote:

SX-33 @ 25 MHz = 20.0 // Equivalent to a very slow 486

Somehow I find this part quite funny 😜

Anyhow, I'm wondering if there might be any other boards that can run as slow as the SP3. The "wrong cpu speed" at post does indicate I'll need a better method of testing undocumented jumper settings. Back when I was testing some of my boards I already did have the feeling that sometimes 2 different jumper settings which posted the same speed actually ran at different speeds (sometimes would be VEEERY slow)

retro games 100 wrote:

SX-20 @ 20 MHz = 15.8 // Equivalent to a 386 DX 33/40

Very nice!

That's a 386DX 40 right there...

Remember though that your video card also helps. So if you insert a ISA card (Cirrus Logic or ET4000) then you should have an equivalent to 386DX25 or 33!

This is a great thread! 😀 Tetrium, I think it is sometimes prudent to disregard what the BIOS POST display says about the CPU speed. I have seen this occur, on other mobos. Eg, the BIOS POST will say 300 MHz, but it's really running at 315 MHz or 330 MHz. This example is from a SS7 mobo I was testing recently.

Also, do you remember the P233 MMX we were investigating, about how to overclock them - well, it's possible that the multi may have been successfully increased from 3.5 to 4, but because I looked at the BIOS POST display only to see the CPU speed, I thought that experiment had failed.

One way to find out the CPU speed is to run Speedsys.exe. I think that will be more accurate than relying on the BIOS POST display. 😀

FIC mainboards usually also have four jumpers for bus speed selection. I remember they can be clocket very low. I'm not sure, but I think it was 4 MHz or something.

About the CPU speed reports from the BIOS: I reckon the BIOS checks for the CPUID, runs a simple timing loop and then matches the speed to a table entry. It doesn't seem to take bus speed and multipliers into account.

For example, the PVI-486SP3 BIOS easily gets fooled by unusual bus/multiplier settings. When you run a 486DX4 at 40x2=80 MHz instead of 25x3=75 MHz, the BIOS still reports 75 MHz for the CPU speed and sets the memory timings to very low values, which would be fine for a 25 MHz bus, but could get unstable at 40 MHz.

retro games 100 wrote:

This is a great thread! 😀 Tetrium, I think it is sometimes prudent to disregard what the BIOS POST display says about the CPU speed. I have seen this occur, on other mobos. Eg, the BIOS POST will say 300 MHz, but it's really running at 315 MHz or 330 MHz. This example is from a SS7 mobo I was testing recently.

Also, do you remember the P233 MMX we were investigating, about how to overclock them - well, it's possible that the multi may have been successfully increased from 3.5 to 4, but because I looked at the BIOS POST display only to see the CPU speed, I thought that experiment had failed.

One way to find out the CPU speed is to run Speedsys.exe. I think that will be more accurate than relying on the BIOS POST display. 😀

Yes, I've seen this before also, but didn't think much of it and had since forgotten. I once tested a SS7 board and had it jumpered to a nonstandard fsb to overclock the cpu from 400 to around 415Mhz. It still showed 400Mhz and thought it was odd, but I kinda forgot about it until reading your comment 😉

Right now I'm in the middle of a lot of things, including redoing my entire house causing my attic to be near-impossible to walk through, let alone start doing tests. I reckon it can take a couple months before things turn back to normal here.

So anyone know of a small little program that I can boot from that can do a quick cpu-speed test from a batched bootable DOS disk?
Some of the settings I've tried before were I think really dangerous (66Mhz fsb on a 486 board?) and I'd rather not watch all my precious socket 3 boards go up in smoke you know 😜

Tetrium wrote:

So anyone know of a small little program that I can boot from that can do a quick cpu-speed test from a batched bootable DOS disk?

CHKCPU <-- This one is really nice!