Which model PCChips board is it? I might be interested, and I think I can get you what you're looking for in trade.

M919. Elpina brand. Also, I don't know if it'll make much difference but the DX4 chip I'm offering along with everything else is the Overdrive version.

So you're not keen on keeping spare parts?

I figure if someone else could give the stuff a better home, I'm all for it. I'm currently in discussion with Old Thrashbarg right now; I'm actually banking more on trading for stuff to improve my current retro rig with.

SOLD!

I just bought a similar 486 from ebay. I think it's an older version of this mobo. A photo of it can be seen here. If you look carefully at this photo, the cache chips are socketed, and so I'm hoping that will increase the possibility of them not being fake. I wonder if this UMC chipset-based mobo can accept EDO RAM? Perhaps the currently installed 32MB SMIMs are EDO RAM? Also, I wonder if the bus jumpers can be set to undocumented settings, such as values greater than 40 or 50, such as 60 or 66?

Old Thrashbarg, when you get your mobo, we'll have to race them with some benchmarking utils, to see what they can achieve.

If you look carefully at this photo, the cache chips are socketed, and so I'm hoping that will increase the possibility of them not being fake.

I had some of those G-Link chips (which I actually just sent to AlphaKilo as part of a trade), and they were definitely real and functional. I guess there's a possibility that the markings on yours are fake, but usually the PCChips fake cache either didn't have any particular brand markings on 'em, or faked the markings of better-known brands like Winbond and Alliance. So I think it's probably a safe bet that you're actually getting real cache on that one.

EDO support depends on the revision of the chipset... unfortunately, that information is hidden underneath the round green sticker. Given that it's an older revision of the board, it's gonna be a coin-toss whether you get the feature or not. I do know you're probably going to want to pull out two, maybe even three of those SIMMs... the thing can officially cache 64MB RAM, but from some of the reading I've done, reportedly there's some weirdness about it and it may only be able to cache 56MB in actuality. (I'm a bit dubious about that claim, as I can't think of a technical reason why there'd be a 56MB limit, but that's the story as I've heard it, anyway.)

The thing should support 60mhz bus, possibly even 66, in a similar manner as other UM8881 boards... but whether the board design and component quality is good enough to handle that speed stably, remains to be seen.

One thing I find odd about the board you got, is that it only has one power transistor on it... the thing with the heatsink right above the VLB slot. I'm not sure what effects that may have in practical terms, but considering that PCChips is known for cutting corners, and later revisions of the board have two transistors installed there, that may be cause for concern.

Anyhow, a race it will be. I plan on tweaking the hell out of mine, so sharing any tips and tricks we find should be fun.

Old Thrashbarg wrote:

One thing I find odd about the board you got, is that it only has one power transistor on it... the thing with the heatsink right above the VLB slot. I'm not sure what effects that may have in practical terms, but considering that PCChips is known for cutting corners, and later revisions of the board have two transistors installed there, that may be cause for concern.

There's always cause for concern whenever PC Chips are involved. 😉 BTW, I had lots of fun with the revision you're about to receive. At least I think it's the same revision. I talked about it here.

Old Thrashbarg wrote:

Anyhow, a race it will be. I plan on tweaking the hell out of mine, so sharing any tips and tricks we find should be fun.

😁

My experience with PCChips products is that making them run fast isn't as hard as making them run stable.

Yes that's my experience too. I had an M919 with fake cache and with a bit of tweaking I could actually get respectable performance out of it. It was quite flaky with hard lockups and EMM386 errors which seemed to get worse the more I used it. I've since replaced it with a different board with real cache and it's much more reliable 😀

I've just received the 486 mobo, and I'll test it today. The RAM has EDO written on it, so that's good. The 4 SIMMs are 32MB each. The socketed cache chips have -15 written on them, but I'll need to run cachechk(7) on it, just to make sure they're not fake. The mobo's build quality looks reasonable, but it's unmistakeably an M919 board. It's version 1.4, and info about these boards in general can be found here -

http://motherboards.mbarron.net/models/486pci/m919v1.htm

However, looking at this info, the jumper info is not 100% correct for my board. There's a jumper on it that says jumper JP23 is on for 3.3V, or off for 4V. Oh dear, no 5V option then? However, there's another bank of 4 jumpers labelled 3.3V or 5V, so I'm not sure what that's all about. Perhaps JP23 is for I/O voltage?

The bus speed jumper layout is unmistakeably UMC based, and the PCB has a 50 MHz configuration written on it. I'm confident that I can select the undocumented 20, 60 and 66 MHz speeds, but the real test will be to see if those two higher 60 speeds are stable.

retro games 100 wrote:

However, looking at this info, the jumper info is not 100% correct for my board. There's a jumper on it that says jumper JP23 is on for 3.3V, or off for 4V. Oh dear, no 5V option then? However, there's another bank of 4 jumpers labelled 3.3V or 5V, so I'm not sure what that's all about. Perhaps JP23 is for I/O voltage?

That seems to be a correct assumption 😉.
Those very late boards weren't expected to run any 5v chip anymore and we all know PC Chips loved to make their boards as cheap as they could and still get away with it 😜

Regarding the 4V / 5V volts issue. I think I have worked out what the jumper specifications mean. If you want 4V, you adjust jumper JP23. However, if you want 5V, you leave JP23 alone, and simply alter the "3.3V or 5V" set of jumpers elsewhere on the board.

The BIOS appears to be the 2nd release for this board: 2/16/96. I have been testing this board for a couple of hours now, and it seems OK. I have managed to do these tests:

AMD 5x86, 4x40=160, BIOS completely maxed out.
AMD 5x86, 3x50=150, BIOS completely maxed out, except Read=1 (not 0)

With these 2 test configurations listed above, I've run Quake, Cachechk(7), Speedsys, and 3DBench so far. I also ran Windows 98. I got a couple of weird problems. Sandra 2002 Pro pops up an error message when you run its benchies, but continues to run them anyway. I may even do a fresh Win98 install, using this board.

Cachechk gives me values indicating that the 256Kb of onboard cache is not fake. I'll post screenshots etc a bit later. What I now need to try is the 2 big tests. The 3x60=180, and 4x50=200 MHz speed tests. For that, I'll probably need to alter the voltage. So far, I've been putting the normal amount of voltage through this ADW P75 5x86 chip.

Edit: 3x60=180MHz works! All I had to do was set the voltage to 5V, and the bus speed jumper to "off on off", which is the UMC chipset-based undocumented setting for 60MHz. Screenshots etc to follow...

With the board set to 3x 60 = 180MHz, here are some benchmarks. I ran ChkCPU.exe and CacheChk(7), and sent their output to text files. They can be read below, as attachments.

I ran Quake, but it aborted with a "bad surface extents" error message. I didn't spend any time messing about the BIOS settings, in an attempt to get it to run. I'll try that later. I ran doom -timedemo demo3, and got 52.340 FPS. That was in full screen.

Here is the thing in action. The next thing to try is:

Get Quake working.
Try 4x 50 = 200 MHz.
Get Windows 98 working. I'll go for a reinstall, just to be sure.

I got Quake working. This game seems to push the cache quite hard. On all previous tests, I was able to max out the BIOS cache timings, but not for Quake. I had to loosen up the cache timings to 3-2-3, which is their slowest setting. However, I have the RAM timings set to 1-0 for read-write.

Please note that this 3-2-3 cache timing setting is only necessary when the board is set to 3x 60 = 180 MHz operation. Quake timedemo demo1 gave me 16.7 FPS, in full screen mode. That's not bad. And that's just using a basic Virge 325 card. On other UMC 486 boards I get about 17.1 FPS, I seem to remember. (That was probably with a Diamond Viper V330 card.)

Also, I tried 4x 50 = 200 MHz, but DOS would not boot up. It got stuck as the auto file was executing its instructions.

I did a fresh install of Windows 98 SE, using this mobo. It worked OK. However, when I run Sandra 2002 Pro, it gives me an odd error and then after that, all of its benchmarks appear to run normally. I set the mobo's BIOS timings to maximum, except for the cache timings, which I set to 3-1-3. That's "two notches" slower than the fastest setting possible. I tried "2-2-2", but Win98 wasn't stable enough. The other setting is 2-1-2, which is only feasible for DOS - but not for Quake! I'm only using a 2MB Virge 325 video card at the moment. I ran some benchies:

WinTune 98

Sandra 2002 Pro

I removed the AMD P75 5x86 133MHz CPU, and replaced it with an AMD DX4-100 WB CPU. I have discovered that you can set a jumper on the mobo to force this CPU to run in 2x multi mode, rather than its default 3x multi. (Is it possible to go beyond a 3x multi setting?) With this chip set to run @ 2x multi, I increased the mobo's bus speed to 60 MHz, giving me a CPU clock speed of 120 MHz.

I ran Quake, in full screen, and I get 11.7 FPS. That's a bit rubbish, compared to my other experiments, but it was interesting. The voltage setting was set to 4V. The BIOS cache timings had to be set to their slowest setting of 3-2-3.

Whoa! RG100! SO many classic informations! I need sometime to digest 'em all! Great going, I'll have a look at these datas later!