dirkmirk wrote:

what scores do you get for the doom timedemo?

I will find out! Currently busy playing around with the SX-33 board. I replaced the battery, quite proud of myself considering I'm not good when it comes to repairing stuff. Video in the process 😀

I don't have any 30pin 60ns SIMMs, just 72pin. But I don't think it makes a huge difference between BIOS defaults and lower Wait States.

Well, if you run Doom timedemo you can compare with default versus tuned. I think it may be in the range of 5%.

I made two videos about this board with BIOS default settings:

Doom: (runs too slow) https://www.youtube.com/watch?v=gIQVUK9hhBE

Wing Commander (runs too fast): https://www.youtube.com/watch?v=fJ5BlmbCMBM

The score for Doom is 8539 realticks or 8.75 fps.

Without cache:

3dbench: 8.1

Doom 20545 realticks or 3.64 fps.

Would you kindly re-run the doom timedemo without the screen border but leave the status bar? ie one size less than maximum.

Still 8.8 fps is very, very good for a 386 system, I think mine only achieve about 6.8 or thereabouts, interesting as my systems have 60ns ram must be the chipset that gives that extra oomph, I noticed no performance difference between the et4000 and a 2meg cirrus logic 5434.

edit: Were you aware that you press F5 for low detail for Doom? Doom is perfectly playable on the 386 in Low Detail not that I suggest you do but back in the day it was a great compromise if that's all you had.

I'm currently playing with another board. Also I'm kinda in the wrong thread 😀

I will update it in the FX-3000 thread when I get to it 😁

One thing that might be of interest:
I had a look at the AMI Bios in one of my 386 systems dated 08/08/93 and the CPU detection routine already supports Pentium CPUs.

Motherboard: QDI U386WB
Chipset: UMC 481 and 482F
Cache: 128k 10ns
Memory: 4MB
CPU: AM386DX-40
FPU: No
Oscillator: Eclipse 100MHz

System is not stable.

Are there any 486SX benchmarked with Speedsys? I'm wondering about the CPU/FPU issue that you can see when testing the 386 with a fpu and the score dramatically drops. I only have 486DX variants.

Carlostex, can you please run the test again with himem.sys disabled so you can do the memory test?

I recently bought a 100MHz oscillator as well. I am going to attempt 50MHz on my OPTi 495SX system. This chipset was used in 486 boards, so perhaps it will be able to handle the speed a little better.

Isn't more then 4MB required to do the memory tests?

He's using an older version.

Okay, I tried my system at 50MHz. I am unable to operate my cache at this speed. I will try some faster modules (if I have any left). I think with no cache 50MHz isn't worth it. However, if you used a cpu with L1 cache like the 486SXL or an IBM DLC it might be worth it. I am tempted to try my 486SXL at 50MHz, but I think the last time I overclocked one I fried the cache. So in that case, I'd have to use an SXL2 rated at 50MHz (can be ran in single clock mode)

However, I also have a 90MHz oscillator, and this one seems more promising. I will post results when I am finished testing.

*update* OPTI 495SX best possible timings
Both tests are with L1 cache turned off. I don't have the control software installed on my system to make it work properly with SCSI at the moment.

TI486SXL @ 40MHz

TI486SXL @ 45MHz

Anonymous Coward wrote:

Carlostex, can you please run the test again with himem.sys disabled so you can do the memory test?

I recently bought a 100MHz oscillator as well. I am going to attempt 50MHz on my OPTi 495SX system. This chipset was used in 486 boards, so perhaps it will be able to handle the speed a little better.

Sorry i'm living in England and all my stuff is locked and stored at my folks place in Portugal. I'll be there early November but i'll hardly have any time to test this. I have a Turbo XT computer to build.

Anonymous Coward wrote:

However, I also have a 90MHz oscillator, and this one seems more promising. I will post results when I am finished testing.

I've looked for one too on eBay but never found a single one. So i kinda gave up. Plus i have a suspicion that my branded 10ns cache might not be actually 10ns, but way slower. Again i don't think i'll have enough time to redo the tests with 15ns ones.

My 90MHz oscillator is actually not DIP18, I think it's DIP12 or something. It's shaped like a square. The Chinese seller rigged it up on an adapter that lets me plug into the socket on my 386/486 motherboards. You might be able to get by with a 45MHz oscillator if the motherboard has jumpers that let you use a 1:1 ratio.

I don't think i have any motherboard that allows that. I think the cache is the main problem as the system was quite stable if i turned the cache off. Performance of the system at 50MHz without cache was about of a 386DX-33.

I am not sure if it's a problem with the cache chips themselves. With enough waitstates it should be perfectly stable. I set my wait states all to the maximum which should have done the trick. However, the system was still not really stable. I think it's either a cache controller, chipset or motherboard design issue. Remember even many 486 motherboards can't run 50MHz properly.

I was looking on the internets for an award BIOS for my OPTi 495SX chipset board. I did not find it. However, I did find a nice MR-BIOS. It's much more flexible than my AMI BIOS. It gives me a super fast bootup time and improved DRAM performance. Here are my results for TI 486SXL at 45MHz with L1 cache OFF.

I was also able to get 50MHz going with L2 cache on...but the lower DRAM and SRAM timings hurt memory performance.