A little off topic, but you mentioned using a Cirrus GD5422 card... have you tried any games with that thing? I had one of those, and it simply wouldn't work right with certain graphics modes, gave all sorts of artifacts all over the screen. I'd found some old usenet postings that indicated I wasn't alone with those problems, so I'm curious if yours shows similar issues...

I wouldn't think the ISA bus speed would increase the CPU score that much. When I went from 7.2mhz async, to /10 and /8 I got very tiny changes in the CPU score. What ISA bus setting have you got in your first picture where you get 9.2? Can you try with the /8 divider at 40MHz and see what you get? I am getting below 6.

I do have a maths copro installed though, so I wonder if that has something to do with it. Whether it enables extra floating point tests with the copro installed and the 387 doesn't score very highly so it brings the score down. I doubt it but you never know.

As for the 386DXL I think they are just low power versions for laptops? Maybe there is some other change. I think I have one of those lying around somewhere so I guess I can try it too.

I switch on the retro 386 PC this morning for some more testing, and it doesn't work. Nothing unusual about that! 🤣 First of all, the barrel battery has croaked. I fixed that with 3x AA batteries. I just wrapped the wires around the external battery pins, on the mobo. Red goes on positive. I had to look that up on the net. 😊

I am happy to test the Cirrus Logic card with any "difficult" games. I tried Doom, and it worked fine. I even made this poor quality video Begin Edit: video now removed End Edit., showing the timedemo demo3. I get ~7.5 FPS.

Cirrus Logic:

For the 9.2 Speedsys score, I was using the /6=13.333mhz bus setting. I tried slowing down the bus jumper to /8=10mhz, and I got 9.05! I also ran 3DBench, and got 15.6. Hmmm. There's something funny going on. Either the CPU I'm using is somehow "super", or there's some kind of bottleneck in your CPU performance. Your 3DBench score seems perfectly OK, however! 😀

I hope it isn't the CPU, because I've ordered some of those "CPUs on a pin package thing". It wouldn't be your controller/HDD, would it? I am using a SuperIDE controller, with a compact flash disk only. Absolutely nothing else is in the system. Also - could it be your 100ns memory, I wonder? What else could it be, um...cache chips, perhaps? Are they the right speed?

Memory speed should make quite a difference I think. Seeing that these machines had so little cache to begin with.

What is your memory graph with speedsys compared to DK's? Maybe post a screenshot then we can compare it with DK's charts.

Did you have any luck with the oscillator chips?

Your SuperIDE controller might give you better HDD performance, but it wouldn't affect anything else...

So my guess is the memory. What speed was standard back in the days? I recall 70ns?

Here's Speedsys, running @ /6=10mhz. Regarding oscis, I had success with them - please read my earlier post.

Let's double-check a few things. Jumpers for /6=10mhz settings:

JP3 = On
JP4 = Off
JP5 = 23
JP6 = 12
Pins 15 and 16 only = covered. This is for Turbo. In the manual, it shows 15 16 17. I'm not sure why.

Dang! I somehow missed a whole page 😒

Your memory throughput is almost identical to DK's...

I ran some 3DBench tests on the 386 mobo. For every test run, I used the same CPU, which is an AMD 386DX-40. I adjusted settings on the mobo, in order to produce different 3DBench scores. These adjustments can be seen in the first 4 columns in the table. The 5th column contains the results.

I omitted two types of test. 1) I did not set the Turbo jumper to Off. All tests run had the Turbo jumper set to On. 2) When I could see clearly that 3DBench was going to produce a slow result, eg about 5 or less, I did not include the results in the table. By omitting these two types of test, I removed lots of scores in the "slightly worthless" 1-5 range.

The first column is the Oscillator. This is a silver "tin can" that sits on the mobo. You can change it, to make the CPU run faster or slower. The 80MHz osci makes the 386DX-40 run at 40MHz. The 66 osci makes this CPU run at 33 MHz, and the 50 osci makes the CPU run at 25 MHz. All bus speeds run in "synchronous mode", except for the 7.2 speed, which runs asynchronously.

The results produced seem to be flexible. The only number missing is ~11. Pity about that! I reckon that if you run one of those pesky speed sensitive 386-era games, you can adjust this mobo to get it to run at the desired speed. Of course, there are other hardware options to achieve the same thing. I just did these tests on a 386, because I am interested in old hardware, and it was fun to see what results could be obtained, if you adjust various settings.

You sure were busy there RG100!
As I lay in bed yesterday night, about to go to sleep, I decided to do some reading and gave this mobo manual a go, just to see if I'd missed anything particular (yup, I sometimes read old manuals in bed, 🤣).

And I saw that on the same page as "Contaq-386 UPGHRADABLE (etc)", on the underside in smaller printing, was this:"MS-3124 ver 2.0"

This is a usual model numbering for MSI motherboards in those days.

Is this Contaq really made by MSI?

Edit:I kinda ventured into my attic for the first time in a couple months, it was kinda neglected and could barely walk there.
I had basically stacked everything I've gotten all on a couple piles, unsorted, so yesterday I finally started sorting all of the stuff and once again, I have a pile of about 1 meter high of untested hardware, 🤣!

I was tired so stopped before I had finished sorting through all the 'new' stuff and had no time today.
I still haven't had the time to remove the old barrel battery 🤣.
But at any rate, I want to start making room so I can eventually start toying with my hardware again 😀

Tetrium wrote:

I still haven't had the time to remove the old barrel battery 🤣.

You don't need to! 😀 At least, not immediately. You can attach an external battery to the 4-pin battery header, on the mobo. Red wire goes on pin 1, black on pin 4. If the wire isn't long enough, and plastic covers too much of it, you could strip the wire with some pliers, or something similar. That's it, you're ready to start testing!

Is this Contaq really made by MSI?

Contaq is just the chipset, innit? The sellers probably just couldn't figure out who made it, so they just listed it by the chipset maker... Sorta like how a lot of boards are listed erroneously as American Megatrends because the AMI BIOS sticker is the most prominent marking on 'em.

Awesome to actually see the difference between 386DX 25, 33 and 40. Back in the days it must have been quite noticeable if you had a 40 MHz BEAST.

At 40MHz it's 22.5% faster than 33 MHz. And at 33 MHz it's 33% faster than 25 MHz.

I always thought the 40 MHz machine is just a tiny bit faster, but 22.5% is definitely something you will notice.

RG100:

Just had a thought!

Can you buy a "variable oscillator"?

Like a device where turning the wheel changes the frequency? Or you can do what I did with my 486 machine back in the day. I drilled a hole in the case and a switch that allowed me to change the bus frequency. It even worked on the fly!

Memory speed should make quite a difference I think. Seeing that these machines had so little cache to begin with.

I wouldn't be so sure. The CPU score in Speedsys seems to be affected only minimally by memory speed.
It could be the fast page mode thats doing it, but there's no easy way to tell if that is enabled or not- as far as I know.

The thing is that mine isn't low, going by the reference scores in SpeedSys I'm actually slightly above what is expected for a 386DX40.

I'm just curious as to why RG100's is more than 50% higher than both my score and the reference score.

Well I really can't tell why is that difference, but I can also see it in my 386/40 motherboards. I have one that scores 80mb/s in memory speed and 9,50 in cpu speed, but in 3dbench it has different results. My best board, (which I thought was broken btw, but I'm very glad to see it work again) scores 18,5mb/s and 6,50, but in 3dbench it has 17.6fps, while the first that is faster according to speedsys, can't achieve a score better than 16,5fps. And that is using the same Cirrus Logic VGA.

I'm not sure if you can buy a variable osci. However, I've just ordered a 100 MHz osci. If it works, I might have the fastest 386 on the planet. Hehe!

BTW, I read Vogons user Markk's post here. He mentions using a diode, when using an external battery. Markk, should I use one for my 386 mobo? If you look at my post above, I have just attached the external battery to the mobo, without much thought! 😦

retro games 100 wrote:

If it works, I might have the fastest 386 on the planet. Hehe!

Well, then you might want to get 8x16mb 30pin simms and this chip:
http://upload.wikimedia.org/wikipedia/commons … rix_486DRx2.jpg

More info here: http://reference.findtarget.com/search/Cyrix%20Cx486DLC/

retro games 100 wrote:

BTW, I read Vogons user Markk's post here. He mentions using a diode, when using an external battery. Markk, should I use one for my 386 mobo? If you look at my post above, I have just attached the external battery to the mobo, without much thought! 😦

No you don't have to do that. I would only fit a diode if the motherboard doesn't have an external battery connector, and I have to connect it where the barrel type rechargeable battery was. I've only encountered this just once on a more "recent" motherboard. A soyo socket 7 one, that had barrel type battery soldered(strange for a board of that era) with no external battery connector at all.

I'm not sure if you can buy a variable osci.

Dunno if you can buy exactly the thing you'd need for a 386 board, but it's certainly possible (and not particularly difficult) to build a PLL circuit that would do it. It was a popular method amongst extreme overclockers in the 440BX-era... the "Turbo PLL" as it was called. Of course, one for a 386 would be much simpler, since you don't need the separate outputs for the USB, PCI, RTC and all that.

I have done some more testing with this mobo. The 100 MHz osci I ordered has arrived, and it works. The 386DX-40 chip now runs at 50MHz. I spent about 30 minutes messing about with everything, and I can get the mobo to behave sensibly at a bus speed of 13.3 MHz, but not at the very highest setting of 20 MHz. There's two things which may be vital to ensure that this mobo runs at this high speed (50 MHz CPU clock speed, and 13.3 MHz bus speed).

(1) Inside the BIOS set up area, and inside the "Advanced chipset settings" area, I think that the field "Auto Configuration function" must be set to Enabled. When it was set to Disabled, I couldn't get the mobo to POST successfully.

(2) I believe you must use a compact flash device, if you want this mobo to run at this high speed. But not just any compact flash device. Please look at the photo. Notice that the CF device is mounted directly on to the multi I/O controller card. Now take a look at that cable next to the mobo. That can also use a CF device, but it's built in to an IDE cable. That's no good. There's something "bad" about it, when running at this very high speed. This cable works 100% at a CPU clock speed of 40 MHz, but if you go higher, you need to ditch it in favour of the one mounted directly on to the I/O controller card.

Underneath this photo is a Speedsys screenshot. It shows an improved score of ~10. Previously, with the AMD CPU running at 40 MHz, I got ~9. My 3DBench score has improved from 16.9, to 17.5. I was hoping for a slightly better score, but I'm just using a Cirrus Logic card at the moment. I tried a Diamond Speedstar x24 card, but it displays garbage at the BIOS POST screen. It works otherwise, just not at this high speed.

Hmm I also expected a heaps better score. Because going from 33 to 40 we saw quite a difference.

Right now the board I have in my pc, does 15.8 fps using the default settings, and if I change only the at bus to 20MHz(sysclc2/4) it measures 17.5fps. With some further tweaking in some bios options, I can achieve 17.6, which seems to be my absolute maximum score. It would be interesting to try a 100MHz crystal....
I'm using also a cirrus logic 5422 card. I've got two of them. Not identical though. What's strange is that one has 80ns memory chips, while the other has 70ns. If I use the 70ns one, it's about 0.2fps lower than the 80ns.... I guess the other it's a better card.
Btw, I love that board... I noticed that if have the turbo turned off, it becomes REALLY slow.... Landmark speedtest shows performance equal to a 5MHz AT, and 3dbench does 1.2fps. Now I really need to dig up "zaxxon" 😀 I used to play that game on an Amstrad 1512 we had at school, but any other PC I owned after, was way too fast for it.... 😜