retro games 100 wrote:

5u3, please can you tell me what pin I need to cut, in order for the P75 CPU to automatically operate at 5V? Luckily, I have a few spare ADW chips for mad overclocking experiments, and I'm happy to give this idea a go! 😀

Look here for a description of the 486 VOLDET pin and how it works.
I reckon it should be possible to override the voltage detection by cutting pin S4 on the CPU.

5u3 wrote:

retro games 100 wrote:

5u3, please can you tell me what pin I need to cut, in order for the P75 CPU to automatically operate at 5V? Luckily, I have a few spare ADW chips for mad overclocking experiments, and I'm happy to give this idea a go! 😀

Look here for a description of the 486 VOLDET pin and how it works.
I reckon it should be possible to override the voltage detection by cutting pin S4 on the CPU.

Thanks a lot, that worked! 😀 I removed pin S4 from the P75 CPU, and configured this Asus SiS chipset mobo to 4x multi, and 50 MHz bus speed, and the mobo boots @ 200 MHz. Speedsys works, but...

hehe, it was a bit too good to be true, I'm afraid. Quake and Doom aren't impressed by this new arrangement. 😉 Unfortunately, the system doesn't appear to be stable enough. What a shame! That could have been the "retro test success" of the day! Oh well, it was fun trying out this experiment. Nothing has been damaged, because the CPU works OK. I'll label it "5 volts only", and put it in my spares box. 😀

Nice work! Too bad the board didn't live up to expectations, but the lack of a PCI divider makes running a FSB clock above 40 MHz very difficult.

5u3 wrote:

Nice work! Too bad the board didn't live up to expectations, but the lack of a PCI divider makes running a FSB clock above 40 MHz very difficult.

I think the board can cope with the bus @ 50 MHz, in conjunction with using a PCI VGA card. I've run lots of tests with the bus @ 50 MHz, using various PCI VGA cards. No problems there. 😀 But when I increase the P75's multi from 3x to 4x, which increases the CPU speed from 150 MHz to 200 MHz, that's when the system becomes unstable.

udam_u wrote:

If I had free time I would compare SIS, ALI and UMC 486 PCI chipsets. I wonder which is best. Everything indicates UMC... (best overclocking, flexible FSB settings, available FSB/PCI dividers) - what are your feelings on this subject?

I've done some more testing with SiS and VIA chipset-based 486s, and although I can get them to work @ 200 MHz, they fail the DOS Quake 1 timedemo test, which seems to be a good test to check for stability. It does seem that UMC chipset-based 486 mobos are the best, when it comes to overclocking. On one of them, a late revision Biostar, I set the BIOS PCI divider option to 1:1, and ran it @ 50 MHz bus speed, with the P75 @ 200 MHz, and got Windows 98 based benchmarking stable.

However, I think it's worth pointing out something. IMHO, if you want to run later era DOS games and/or Windows 95, the best choice would be to use a Pentium. I think having a fast 486 is important, but stressing this hardware to 200 MHz is perhaps an unwise long-term strategy.

The ironic thing is that my "486 retro strategy" is to use a 486 to run speed sensitive games from the 386 era. I've got 386 hardware, but my testing suggests that using a 486 is more flexible. Typical 486 era games can be run on a Pentium, because during that era games developers must have realised that future hardware was going to get faster all the time, and their games needed to be able to cope with that.

retro games 100 wrote:

However, I think it's worth pointing out something. IMHO, if you want to run later era DOS games and/or Windows 95, the best choice would be to use a Pentium. I think having a fast 486 is important, but stressing this hardware to 200 MHz is perhaps an unwise long-term strategy.

The ironic thing is that my "486 retro strategy" is to use a 486 to run speed sensitive games from the 386 era. I've got 386 hardware, but my testing suggests that using a 486 is more flexible. Typical 486 era games can be run on a Pentium, because during that era games developers must have realised that future hardware was going to get faster all the time, and their games needed to be able to cope with that.

Very good conclusion!

A 486 is definitely more flexible. Especially if you have a Turbo button and BIOS options for L1 and L2 Cache.

But IMO even more flexible is a decent Super Socket 7 system. I had excellent results, the only thing that I had issues with was hardware related and specific to some sort of compatibility issue with Sound Blaster 2.0 and Pro Audio Spectrum 16 cards.

But apart from that I would still recommend a SS7 system as my first choice for building a flexible retro PC. There are a ton of benefits: PS/2 ports, large HDD support, ATX, lower price and more...

PS: My Acer 486 has support for 50MHz, but I don't have a CPU to test this feature...

Mau1wurf1977 wrote:

Very good conclusion! […]
Show full quote

retro games 100 wrote:

However, I think it's worth pointing out something. IMHO, if you want to run later era DOS games and/or Windows 95, the best choice would be to use a Pentium. I think having a fast 486 is important, but stressing this hardware to 200 MHz is perhaps an unwise long-term strategy.

The ironic thing is that my "486 retro strategy" is to use a 486 to run speed sensitive games from the 386 era. I've got 386 hardware, but my testing suggests that using a 486 is more flexible. Typical 486 era games can be run on a Pentium, because during that era games developers must have realised that future hardware was going to get faster all the time, and their games needed to be able to cope with that.

Very good conclusion!

A 486 is definitely more flexible. Especially if you have a Turbo button and BIOS options for L1 and L2 Cache.

But IMO even more flexible is a decent Super Socket 7 system. I had excellent results, the only thing that I had issues with was hardware related and specific to some sort of compatibility issue with Sound Blaster 2.0 and Pro Audio Spectrum 16 cards.

But apart from that I would still recommend a SS7 system as my first choice for building a flexible retro PC. There are a ton of benefits: PS/2 ports, large HDD support, ATX, lower price and more...

PS: My Acer 486 has support for 50MHz, but I don't have a CPU to test this feature...

You could try one of your spare 40Mhz CPU's, take the ugliest one, heatsink it and go ahead and test.
There are very few 50Mhz FSB CPU's available for Socket 3 and the 40Mhz ones seem to be much more common.
I used a spare AMD DX-40 for my FSB tests and it was able to boot even at 66Mhz for a few seconds at least (long enough to view the post screen 😉 ).

@rg100

Hi,
I agree with you. Currently I use PPro rig for late DOS/ early Win95 games. Everything work perfectly without need to overclock parts. But there is something magical in 486 what causes endless pleasure from tunning it. I think that people like me or you, will still try to squeeze last juices from it to overcome early Pentiums - even when others will talk "it doesn't worth"! (;

I'm sure your contribution in 486 overclocking will be remembered on this forum.

Regards! [:

udam_u wrote:

I'm sure your contribution in 486 overclocking will be remembered on this forum.

😀

I hope my accidents in other areas will be forgotten! 😉

I tried another similar experiment with this Asus PVI-486SP3 mobo. With the P75 OC'd to 200 MHz, I used a VLB VGA instead, instead of a PCI VGA. But I get the same results as before: the system works, but Quake timedemo will not run.

And as before, Quake itself runs OK, but unexpectedly exits as soon as you press the ENTER key after entering the timedemo console command. I guess there must be some "sensitive" bit of code run at that precise point, which this OC'd SiS chipset-based mobo can't cope with. BTW, I had the BIOS timings set to their slowest values, and also the mobo's Wait State jumpers were set to > 33.

If you want to squeeze more juice out of 486, I have an idea that would be totally awesome. Somebody could fabricate some special SIMMs using SRAM instead of DRAM. The increase in memory throughput would be incredible. Let's say you got some nice 8ns SRAMs. Then you could disable your L2 cache and run your memory at 0WS.

Over on vintage-computer.com they are currently building SRAM memory boards for XTs, so it's certainly doable

Anonymous Coward wrote:

Over on vintage-computer.com they are currently building SRAM memory boards for XTs, so it's certainly doable

Seriously?? I must've missed it somehow.
But will they be 72p? And what about the memory size?

It could be done for either 30-pin or 72-pin SIMMs.

I imagine these days you can buy some pretty large SRAMs. 32mbit (4MB) modules should be pretty common these days.

I'm testing this Asus mobo again. I used an Intel DX-50 @ 50 FSB, and WinTune running inside Windows 98 worked normally. I then removed this Intel chip, and replaced it with an AMD 5x86 P75, and ran it at "3x & 50 = 150". Again, Windows 98 appeared normal. The BIOS timings were almost maxed out. I had to set the BIOS DRAM speed to "faster" and not "fastest". The cache settings were 2-1-1-1. I am using 2 sticks of 16MB (32MB total) FPM with a 60ns speed rating. VGA is a PCI Virge 325.

I ran Quake for DOS inside Windows 98, and its fullscreen timedemo worked. I get 14.0 FPS, which is quite good. I then ran SuperPi and also Prime95 at the same time, and while these 2 apps were running, I also ran WinTune again. It worked. So, the system appears to be stable @ 50 FSB.

I ran UniFlash version 1.40, and made a back up of the BIOS, for this PCB revision 1.8 mobo. I have attached this file below. I have renamed it to a .txt file. You'll simply need to rename it back to a .bin file. (Vogons does not allow .bin attachments.) The BIOS POST screen shows this information:

#401A0-0306 (That's the BIOS version. I think it's the latest)
06/25/96-SiS-496-497B-PVI-4SP3C-00 (That's the BIOS id which appears at the bottom of the screen)
UniFlash says: Winbond W29X010/011/012/5V series

Wintune, Windows 98, without any other apps running

Wintune, Windows 98, with SuperPi and Prime95 also running at the same time - for a stress test.

I also scanned a sheet from the Asus manual. It shows you extra jumper settings for newer CPUs.

Old Thrashbarg wrote:

But anyway, I'll need to clean up the graphics cards first anyway (clean contacts with rubbing alcohol).

Random tip: pencil eraser. I've found that a couple passes on each side with an eraser does a more thorough job and quicker than alcohol. It's become part of my routine anytime I install a card or stick of RAM that's been sitting for awhile.

I've just tried this, and it works very well. 😀 The eraser I used was one of those "two colour" ones - one colour (blue) supposedly erases ink, while the other colour erases pencil marks. To clean the contacts on an old ISA VGA card, I used the blue coloured side of the eraser, and it did a great job.