You want 0 waitstate operation at 25Mhz ? That would require 40ns simms, yes? Fastest I've ever seen is 45ns and I only found 2 of them. But you never know. Sometimes you find magic chips that exceed their specs.

Your best upgrade might be to grab a Cyrix 486DLC. Those are available and your board may be MR BIOS eligible, which would help with the Cyrix support:
https://docs.google.com/spreadsheets/d/1vtPso … #gid=1809868176
Might be worth checking out V019B301 .

Hmmm yeah, I was wondering if the 4 way interleave and some "small buffer we're not actually calling it cache" inside the chipset would alleviate the access time requirements, since it would seem the mem controller can fire one request off, and have that chugging along a couple of cycles while it does another. Meaning I guess that for an atomic single cycle, zero wait state is a lie, but effectively en masse, it has same effect. Orrrrr, they claimed something it could only realistically do with a bottom of the barrel 12Mhz CPU at the time of design when 80ns RAM was fast.

Thanks very much for the MR Bios deets. I am wondering if this BIOS might be late enough to be Cyrix aware.

I was thinking of the multiplied chips, and having imperfect support, like the upgrade modules/cards, since I thought it would be cool to have a 25Mhz "486" which performs about like 386-33, then when you run the cyrix util the multiplier and enhancements kick in and you get dx2-50 and approx 486SX33 performance.

I did some tests a couple years back for a 40Mhz 386 & 486DLC running w/ cache or without.

Re: 486DLC cache coherency blues and headaches

TLDR:

• MR BIOS 1.65 detected the Cyrix chip, but I had to use cyrix.exe at startup to get best performance
• Doom likes fast memory more than synthetic benchmarks like speedsys
• A Ti486DLC with just the onchip cache performed about the same as a 386 with a 256MB motherboard cache
• Had to use the "barb" cache coherency method to make the Cyrix cache work with my motherboard cache

1CPU        Mobo   Cyrix    Doom    Speedsys   Read   Write   Move    Avg
2           Cache  Cache  Realtics  CPU Index  MB/s    MB/s   MB/s   MB/s
3Ti486DLC   256KB  Enable    6830    10.29    30.63   37.48  25.53  31.21
4Ti486DLC   32KB   Enable    7616    10.20    30.54   37.47  25.48  31.16
5Ti486DLC   No     Enable   10550     9.88    18.57   37.05  13.48  23.04
6AM386-40   256KB  N/A      10600     6.88    25.53   30.07  38.30  31.30
7AM386-40   32KB   N/A      11140     6.88    25.47   30.05  38.16  31.23
8AM386-40   No     N/A      23654     5.19    16.53   29.87  16.49  20.96
9Ti486DLC   No     Off      25448     5.76    18.57   37.05  16.49  24.03

Very good info, thanks.

Thought I better look see what was actually sitting in it right now in terms of RAM. I am 99.9% sure that's the modules that were sitting in it when I bought it, and it had no CPU then either. I'm a bit foggier about the clock module, whether I "borrowed" it or not. Anyway it arrived to me, definitely prior to 2005, probably prior to 2002. Now one module, has late 1997 marked chips on it, so someone was upgrading or repairing it in 1998, so not all that long before it ended up with me really. Well, I was just assuming it had eight single megabyte chips in there, though at the latest look I had a suspicion that one bank was 4MBs just from general looks. Bank 1 is a nice matched set of NEC 4MB 60ns, Bank 0 is a mishmash of 2 hynix 60ns, one micron 60ns and one with F with a bar over and under it, 70ns, all 4MB. ThirtyTwo MB! Was some madlad trying to put win98 on this thing??? At todays reinflated prices that's a few bucks worth of RAM. I was thinking it might get 16 if I was feeling generous. Not sure what the F is, Fairchild ??

So basically 60ns RAM installed as last working... maybe the 70 kept up, maybe it didn't work and got it retired.

I may not be pursuing use of this board in the very near future, gonna stick it back in "cold storage" as it were. Leave it to chance as to whether I get an Intel or other CPU for it. This is due to this "unplanned purchase" which kinda eats my "money I thought it was sensible to spend on 386 crap" for a bit and also, although these are more SX, they might be same performance as what I was gonna do here, Re: Bought these (retro) hardware today

Those are also scratching this itch..
AMD 386SX25 or other speeds, PQFP-100, how well do they overclock?
Which I was gonna try to scratch with a DX40/486 combo board which is more like a 2/3 or 3/4 baby AT so not as small.... and haven't turned up that 25mhz half baby AT yet, which I may want to have handy when those others arrive, for comparisons, possible part borrowing and all that.

Anyway, that leaves me kind of wondering what out of all that, if any, suits the full AT format 5170 style case. Though I do have a suspicion that that half baby AT board is hanging out with a 20Mhz 286 board, so that might be Super286 fodder. Has everyone lost track now? Yeah me too, I have to look back at these posts to figure out where I was going, breadcrumbs or something.

In trying to get clues about the NEAT chipset and the function and malfunction thereof, because of NEAT chipset memory problem on PB800/900 rev D 286 board.... I happened to come across the DTK tech-1234 manual for their NEAT implementation... here http://www.dosdays.co.uk/media/286/DTK%20TECH … rs%20Manual.pdf
Which to me, seemed to read as if during the 286/386 era, manufacturers absolutely intended interleave to be required for the zero wait state mode of their chipsets.

Anyway, I am taking this as evidence that period claims of "zero wait state" are for averagely zero, needing interleave rather than single cycle zero wait state.

It's still in limbo whether project full AT Case = project TOPCAT or not and nothing has materialized on the CPU for it front, which I am kind of leaving to chance rather than pursuing right now.