I was finally able to locate a contemporary review of the Micronics board in addition to the AMI Enterprise III (which uses the SiS chipset like the NiCE board). Looks like despite having only write-through L2 cache, the Micronics is indeed the faster of the two when equipped the same. My speedsys results clearly show this. Of course the Super EISA board has an edge in that it can take more cache and run at faster bus speeds. The review also shows how lousy VLBus SCSI implementations are. I wonder if Adaptec's cards would show similar performance differences.

OK, got another EISA board (it popped up on ebay and I couldn't resist). Its a Tyan 486 VL EISA marked "S1437". Same SiS chipset as the MCCI NiCE board and the BIOS appears to be identical too. I'm seeking info on this one HERE.

It came with 8x1MB parity SIMMs. I'll do some testing this weekend to determine if those 4MB SIMMs are bad or its the NiCE board acting up.

That was a pretty interesting review. I just wanted to point out one thing that I thought was a little unfair though. The author states that VL-bus is faster than EISA at video, yet he didn't even test an EISA graphics adapter. He only tested the VLB and ISA versions of the Stealth24.

The Micronics board is indeed faster than the SiS based boards, at least at 33Mhz bus speed with a DX4/100Mhz chip. Here is my speedsys with the Tyan set to "Fastest" DRAM in the BIOS and writeback L2 cache enabled. The Micronics still beats it even though it only has write-through L2 cache and the CPU even runs slightly faster in it. I guess the custom Micronics memory controller is pretty damn good. Too bad it has that stupid shadowing limitation.

Also, I have confirmed that one of the 4MB SIMMs I got is indeed marginal, they fail on the Tyan board too.

The memory results I get from my Nice SuperEISA seem to be quite a bit better than this. Is memory interleave enabled on the Tyan?

This is a bad photo of my Nice board from several years ago. The bus is running at 40MHz (can't remember if it's on faster or fastest for memory timings). The results are much higher than your Tyan. Oh, and as a quick side note, did you know the guy who founded Nice later went on to found SuperMicro?

For comparison's sake, here's one with a Cyrix 5x86

Any BIOS options I should look for in particular? The Tyan and NiCE have the same performance numbers, I didn't realize that I forgot to post the NiCE "Fastest" speedsys. I did notice the Tyan have the proper 15ns tag cache chips though. The CPU benchmark number remains the same no matter what the memory settings are. I also have an older version of speedsys.... whoops.

CPU score shouldn't change much. I think it's about right for an intel dx4 in write through mode. But, you should be able to get better memory performance on your SiS boards. I believe the Nice BIOS should have an option for "PAGE INTERLEAVE". At least on my board it is enabled by default.

I'll check when I get home. Both boards have what appear to be identical BIOS outside of the ID string.

Memory interleave option is turned on in the BIOS. The SIMMs the Tyan board came with are 60ns too.

When you get a chance, I'd really like to see the speedsys for your Nice board on FASTEST at 33MHz. Be sure to use v4.78.

That will be a rainy day project for tomorrow. Weather is too nice right now to stay all cooped up in the basement!

Test results with 4.78. Note that Speedsys doesn't seem to properly report the L2 cache on the NiCE board when BIOS set to "Fastest" (don't ask me why). It IS working with a noticeable speed drop at the 256k boundary. Also posted my Advanced Chipset BIOS page. Both boards were set exactly the same and appear to perform similar.

What happened to the graph in the test of the Nice board? Did you notice that the main memory read speed on the Nice is much higher than the Tyan? It is listed as 31.46MB/sec, vs about 15MB/sec on the graph for the Tyan. Looks to me as if memory interleave is not working on the Tyan.

Also, you wouldn't happen to have a 5x86 for testing would you?

I wish I had my EISA system here with me. I have to wait until the fall to get it back.

Nope, no 5x86s here for testing, just DX4-100s. The NiCE numbers may be skewed by the funky L2 cache non-detection. The graph was weird, read and interleave were almost completely flat past the L1 boundary with a slight dip at the L2 boundary. Also that NiCE BIOS has some weird bugs, it reports a Weitek co-processor as being installed when its turned off in the BIOS when the board is set to "Fastest"!

Something on the Nice is not right. Either you have a bad SRAM (or DRAM) or the timings are cranked up too high. Maybe the 20ns tag RAM has something to do with it.

A way you can test for DRAM interleave is to remove all SIMMs from bank 1 and just run 4MB. Performance of DRAM should drop in half.

Yeah, with slower memory timings, the L2 appears to work without a problem on the NiCE board.

Since it came up before in this thread, archived information about the AIR 486EI has been "unlocked" in the Internet Archive! They were advertising it with "Green PC" support and write back cache (I presume this is referring to L2).

http://web.archive.org/web/19961029012651/htt … .com/486EI.html

Whats interesting is this board uses the SiS406/411 chipset the NiCE board has, so I don't know how APM support was added. Perhaps AMI updated the Hi-Flex core for the SiS406/411 to somehow support APM and properly recognize newer 3.3v CPUs? Would be interesting to get a copy. Sadly the board's manuals were on their FTP site and never archived.

Revisited the machine last night, which was ironically the 21st birthday of the BIOS core. Its now old enough to drink. I managed to get UMBs working with the latest version of "The Last Byte". So now I got 618k free conventional memory with the machine running in real mode. No more stupid EMM386. If you have a SiS 85c411/406 EISA chipset, its the only UMB enabler that works.

I'm running into a really weird problem with the board that cropped up last night. Whenever I run Simcity 2000 (even on a clean boot), the PC Speaker starts randomly beeping and the keyboard locks up. Sometimes I get a Stack Overflow crash too. The problem also crops up when I reach the test stage of Creative's DIAGNOSE program. The problem isn't the soundcard though, as it still happens in SC2000 with the card pulled. Any items what may be causing this problem? I'm suspecting the keyboard controller for some reason.

Well, I pin-pointed that weird SC2000 problem. It must be some random bug or hardware incompatibility since it only happens when the program is started in real mode. With EMM386 installed and the machine in V86 mode, SC2000 works fine. I still don't understand why it happens in Creative's diagnose.exe program. I might pull out the NiCE SuperEISA board again and see if the problem crops up there.

EDIT: The NiCE board seems to work fine, weird. Might want to try 8MB of memory in it though since I only put 4MB in.

My hunch somehow proved correct. Swapping the keyboard controller fixed the odd behavior. Now I just need to find another chip. Anyone got a spare i8042 compatible keyboard controller laying around? 😜