Over the past week or so I've finally gotten around to working on my own 386DX40 a bit. Pics and full specs of it can be found here, so I won't bother restating all that... though there is one minor update: I discovered that the original cache wasn't working, so I borrowed 128K worth off a 486 board until I can get around to testing the old chips and weeding out the bad one(s).
Anyhow, now that I've got all that squared away, I ran a couple tests on it. Here's the first Speedsys result I got:
Not too shabby... my memory bandwidth is a little higher than most others I've seen in this thread, but I'm also using a completely different board with a different chipset.
I noticed that my CPU score is much better than what DonutKing was seeing, but still nowhere near the CPU speed numbers RG100 was getting. So I got to investigating the discrepancies a bit. I had a bit of a suspicion about what was going on, and on that hunch, without changing anything else, I pulled out the coprocessor and re-ran the test:
Fuckin' aye, that did it. 😳
Looks like RG100 has a bit of competition now. 😁
As for the difference between my and DonutKing's CPU scores with the coprocessor installed, his being around 5.80-5.90, and mine being 6.83... that could be accounted by the fact that my copro is quite a bit faster... I have a Cyrix 87DLC, whereas I believe he's running an Intel 387.
So, I think we can safely call that mystery solved. Speedsys just runs a different test on FPU-enabled systems, and doesn't bother to make any mention of that fact. 😒
I also ran 3DBench both with and without the copro installed, and despite the significant differences in Speedsys CPU scores, I got the exact same 3DBench result both times: 15.6. That number is perfectly in line with what others have been getting at 40mhz CPU / 10mhz ISA.