A 486 motherboard produced in a shady Taiwan alley? That can't possibly be true 😉

I'd like to know where the heck the data thats in TH99 came from anyway. Most of the 386/486 era stuff on there is VERY obscure and there is no way someone managed to collect all those manuals/boards to verify that data. Some of the data is inaccurate too.

I'm not sure if all this is correct, but here's the way I understand it...

TH99 is actually a pirate eBook of sorts... it was ripped off and condensed from the 'MicroHouse Technical Library,' which was a big 3rd-party reference compilation intended for computer support companies and such, kinda like a computer equivalent of what the Rider's and Beitman's manuals are to TV and radio repair.

But MicroHouse got bought out, and then that company got bought out and so on a few times, and the library was discontinued somewhere in the middle, and at some point I guess either the ownership of the copyrights got lost in the fray, or the owner just no longer gave a crap about enforcing the copyright, or something... in any case, nobody really knows who owns it anymore, and so TH99 ended up falling into a quasi-public-domain grey area.

Oh, and AFAIK, all of the online versions of TH99 are some sort of abridged version, the original release apparently had a few thousand more listings, but I have thus far been unable to find a copy of that version.

NJRoadfan wrote:

I'll doublecheck, but thats likely the Intel 82350 series EISA chipset.

Are you sure? I think the 82350 consisted of more chips than 2.

Old Thrashbarg wrote:

TH99 is actually a pirate eBook of sorts... it was ripped off and condensed from the 'MicroHouse Technical Library,' which was a big 3rd-party reference compilation intended for computer support companies and such, kinda like a computer equivalent of what the Rider's and Beitman's manuals are to TV and radio repair.

Odd that I never came across the real thing. I think the shop I worked at did have their hard drive reference book though. Sam's Photofact seems to be big in the TV repair business, heck they even made a few technical manuals for computers in the 80s.

Are you sure? I think the 82350 consisted of more chips than 2.

I've seen several EISA boards with only two chips used. Its likely the Micronics chip handles the functions the missing Intel chips would normally do.

NJRoadfan wrote:

are the memory benchmark numbers in-line for a WT L1 cache DX4 100Mhz?

You'll want to refer to the Ultimate 486 Benchmark Comparison, Appendix 1, for how well an Intel DX4-100 (L1: WB) does in Speedsys (L2: WB)

Intel DX4-100-WB scores 42.4
Intel DX4-100-WT scores 41.5 (your chip)

Intel DX4-100-WB, L1: 94 MB/s
Intel DX4-100-WT, L1: 70.5 MB/s (your chip)

Intel DX4-100-WB, L2: 50 MB/s
Intel DX4-100-WT, L2: 47.6 MB/s (your chip)

Intel DX4-100-WB, Memory throughput: 37 MB/s
Intel DX4-100-WT, Memory throughput: 33.1 MB/s (your chip)

It seems that your L2 and RAM speeds are about where they should be, and your L1 speed is lagging when compared to L1 write-back, as can be expected. The amount your L1 is lagging is very consistant with WB vs. WT for L1 cache.

I guess its fuzzy memory. I had a 486DX4/100 machine back in the day with WB caches enabled on a proper socket 3 platform (screen shots of that machine landed up on Nathan's Toasty Technology GUI site). That machine seemed to fly compared to this one. Enough that I called it the "Worlds Fastest 486", although my PCChips M919 with AMD 5x86 was technically "faster". It could be the IDE drive on ISA making this machine sluggish though. Those other machines had VLB/PCI controllers.

The board has the Intel 82357 and 82358DT chips controlling the EISA bus for those who are curious.

Details here: http://www.geekdot.com/index.php?page=eisa-indepth

Since I can't well leave enough alone. I subjected this machine to some cruel and unusual punishment last night. I installed Windows 2000 on it, and surprisingly it works decently fast. No support for the Adaptec AHA-2740W though, although I hear the NT 4.0 driver will work. Also despite what the 2000 HCL says, S3 801/805 video is supported by native drivers.

If you think W2K is surprisingly fast on the Intel DX4-100, try a Cyrix 5x86-133 with enhancements on!

What begins to slow down W2K is all those bug fixes and SP4. After installing those, W2K will consume 100% CPU usage for a good minute after reboot, then drop back down to 5-10%. On a DX4-100, this might be more like a 2-10 minute wait.

I've found some NT4 drivers to also work fine in W2K.

Can you let me know if the DX4-100 will play stereo 128 kbps mp3s in Winamp 2.x? I know the write-back version will just barely play them in NT4 at 100 MHz.

The CD I have here only has SP3, a good thing I suppose. I can give Winamp a try, but will be doing so under Windows 98SE.

Replaced the Micronics board with a MCCI/NiCE SuperEISA yesterday (BIOS POST indicates its V1.2.1). Its having some weird issues. First, my S3 805 VLB video card seems to cause the machine to run slower. Memory counts noticeably faster during POST with an ISA video card installed! Speedsys results are slower than the Micronics EISA board. Next, I can't seem to set the memory timing much above "slowest" in the Advanced Chipset config without the machine locking up or throwing random parity errors.

I'm going to play with it tonight. I've run Speedsys with an ISA card to confirm if the S3 card is causing the slowdowns. I also have a Cirrus Logic 5428 VLB card to test.

This is a pretty interesting thread. I'm surprised I didn't notice it last year when it first popped up. In any case, I'm glad I finally got a chance to have a look

I'm pretty jealous that you got Win2K installed on your EISA system before me. I was planning to do so after I got a larger SCSI disk and another 128mb of RAM. It's only been like 8 years since I said I would do it, but it'll happen eventually. That's interesting that you said Win2K doesn't support the AHA 2740W. I have a feeling the NT4 driver would work fine. I had to do this kind of thing quite a bit when Win2K first came out.

I've experienced what you were talking about with memory counts going faster or slower depending on which graphics card is installed...on basically every 386/486 motherboard I have ever owned. I don't know why this happens. Do you think all VGA graphics modes are affected by this problem? I never bothered to check. The ET4000W32P I use in my EISA/VL board is crazy fast, so I never had a reason to check. I recommend trying one of these, or even an ARK1000 (if you can tolerate the shitty windows drivers).

You really need to do something about your memory timings. AT 33MHz I have no problem running in "fastest" mode. I think at 50MHz I ran "slow". It's probably related to either your tag RAMs, cache chips, or even using low quality or mismatched memory. I have server grade 60ns SIMMs in my board, with 12ns tag RAMs and 15ns cache. If you can wait until the summer, I might be able to send you a few spare 12ns tag RAMS (they're hard to get). Though, I have to check how many I still have left (originally I bought 50). I'm saving them for my future AMI Baby Screamer Mark V 386....Oh, one other thing to make sure of is that you fill the first 8 memory banks with match SIMMs. This board supports interleaved memory. Make sure it's enabled in the BIOS. It will double your DRAM scores.

Do you know what your motherboard revision is? I wonder if there are also different versions of the AMI BIOS too. Be patient with this board. I've found it to be no more or no less quirky than other 486 VL boards from 92/93, but it could be a turnoff if you're used to plug-n-pray. This board can be a beast when it's properly tuned. It's supposedly the fastest VL board ever made.

The board is just silk screened "REV 1", just like yours, but has a blue "Intel Overdrive Ready" Socket 2 on it. The machine came with 4x1MB and 4x4MB parity SIMMs, obviously not matched. No clue as to the speed either. I'll pull the 1MB SIMMs and see if that helps with the timing issues and then search for more/better RAM. Looks like the board has the stock 256k cache still. The CPU included was a 486 DX-33 which I swapped out for an Overdrive DX4-100. Going by the stickers on the motherboard and CPU, this board was pulled out of a Tangent brand computer. I'm still pretty POed that my #9 GXE64 VLB card died, that sucker was fast. I can't seem to find the ET4000/W32p cards for reasonable prices. The CL card is a Genoa Windows VGA 24 8500VL.

I'll give Micronics credit. While their board's BIOS isn't tweakable, and the ROM shadowing setup stinks, it was pretty rock solid. I don't regret switching boards though, having a proper BIOS that can have the RAM test bypassed is nice.

Checked the chips. Cache Tag SRAM is labeled -20, while the Data SRAM is labeled -5. Don't know if that means 5ns or 50ns. I'll post pictures later since that seems off either way. The 4MB and 1MB SIMMs are all 70ns.

My board originally came with 20ns tags. I was able to to move from "slowest" to "slow" on DRAM timings by upgrading to 12ns.

The fact that you are stuck running at the slowest DRAM timings at 33MHz incidates something else could be a problem. I would try replacing all the cache and DRAM with better quality stuff.

How much for an ET4000W32P do you consider reasonable?

Here are the Speedsys results thus far:
Memory "Slowest"

Memory "Slower"

The thing thats bothering me about this board is that the CPU seems to run slightly slower and both the L1 and L2 cache are running slower despite the latter being write back.

Another project that I attempted on this machine tonight and failed is using a UMB driver that doesn't involve EMM386 and protected mode. Both HIRAM.EXE (in conjuntion with a generated HISET.SYS) and a memory manager called "The Last Byte" claim to support the SiS 85c411 EISA chipset and will detect it. Neither program will successfully load and enable UMBs however.

I finally decoded the data cache part numbers, its 20ns. Hmm, both data and tag are 20ns, that could be a problem. FWIW, the cache on the Micronics board was also 20ns.

I also figured out why the machine won't run in the "fastest" setting. One of the 4MB SIMMs is apparently bad, but the defect appears to be in one of the parity chips. Turning off parity error halting in the BIOS allows the machine to boot without a problem. I guess one of the SIMMs is marginal, and the fault only appears when it is driven at higher speeds.

Ugh, I hate transient memory problems. This problem did get me thinking, would a board this old even support 3 chip 4MB SIMMs (2 4M x 4 DRAM chips plus parity chip)?

I have a habit of using 9 chip SIMMs on all of my boards, and I didn't even bother to test 3 chip SIMMs on the NICE. I think there's a good chance it would work, but you're better to get 9 chips to be safe.

I just snagged a MAE486-33, anybody have a link for the BIOS?