The cyrix 5x86-120 is definately the best cpu you have, I always assumed you needed a modern PCI mainboard to run that chip as they have specific jumpers for the cpu but someone did have success running the cyrix 5x86 in a vlb board

The next best cpu is the AMD 5x86-p75 which is actually 133mhz but should easily overclock to 160mhz by changing the fsb to 40mhz, If I were you i'd run this chip as the cyrix(IMO) is better suited for a PCI board.

Generally people recommend the ET4000 but the S3 card is probably okay, in reality their wont be much difference if any at all in dos games, I would suggest running a few benchmarks like 3Dbench, pcpbench, quake timedemo and see for yourself, otherwise look at the visual quality you get from both cards

What I think are the cache chips - 8 chips in total have: AS26256AK-15 9507

Those are the cache chips, plus there's another one up behind the top two 8-bit ISA slots. Or rather, those are supposed to be the cache chips, but there's a decent chance that they're fake... the lack of a manufacturer logo, the unusual part numbers, and the fact that you're not seeing any L2 cache in your tests, it all points in that direction.

I guess the first thing would be to check the jumper settings... the picture is too blurry to tell for sure, but it looks like the relevant jumpers might be labelled up near the RAM slots, so make sure it's actually set for 256K cache. If that's all OK, then you could try the chips from your other board, and see if you get working L2 with those. (Make sure you install them the right way around.)

One other strange detail is when I put Ram in the 72 pin simms it wont POST. When I fill all 30pin simms and 72 pin simms it will POST. However If I only fill the 30 pin simms it posts fine.

That's not particularly unusual behavior on those 'transitional' 486 boards with both types of slots. The 30-pin slots were usually wired up as Bank0, and the 72-pin slots as Bank1, Bank2, etc. Sometimes that can be changed with jumpers, but usually not.

I've gotten a Cyrix 5x86 chips going in both a pure ISA 486 motherboard, as well as a VL/EISA board. There wasn't anything particularly hard about it. It will default to write through cache, so a standard 486DX jumper setting should work just find provided you can feed it the proper voltage.

rgart wrote:

I have always been a fan of the 486-DX4-100 Overdrive chip.

However Crusader No Remorse and No regret...as well as Ultima 8 and 7 do lag a little. I dont expect to be getting anything like 60fps however I'd like to get them smoother without using a PCI bus or Pentium and I have no interest in dos box or Exult.

Hi!

Ultima 7 at least shouldn't lag on a DX4-100, in fact it should be way too fast. 50Mhz is as fast as I'd go for that game.

Heres a pic of the cache jumper settings. What is JP30 for? INTLV?

Looking at JP35 this 486 board supports 128Megs of Ram? (right now its set to 0-64Megs and I have 64 megs recognized)

JP39: what is 30P or 72P?

Your right! It doesn't lag as such. I guess Ive been spoiled with 60-100fps.

I just think it can move a tad smoother 😀

CPU + FSB settings...

I guess the 25M, 40M, 33M, 50M settings are FSB?

What is the difference between AV9107-03 and MX-8315?

Picture attached.

what is JP30 for? INTLV?

INTLV = interleaving. That's something that should improve performance (once you get the cache working, anyway), but I'm not sure exactly what the jumper changes... in other words, I don't really know in what situations you'd want it enabled vs. disabled. It may be that it only works in write-back mode or something. I'd leave it as-is for now and focus instead on getting the cache to work at all.

The jumper settings for the cache look correct, so I'd look at the chips themselves next. I see a bit of corrosion on some of the pins in the cache sockets... so the first thing to try would be to clean that up and re-seat the chips. If that doesn't change anything, then the next step would be swapping out the chips for known good ones.

JP39: what is 30P or 72P?

Er, that one should be pretty obvious, especially given that it's right next to the 30 and 72 pin SIMM slots. The setting does just what the label indicates... if you only want to use 72 pin SIMMs, set it to 72 pin only.

How do you guys recommend I remove corrosion from the chips and or sockets?

Found this page:
http://www.trinityos.com/HARDWARE/mb_benchmarks.txt
It's a benchmark result of some sort and if you search "asaki" there, it has 0K Cache RAM. Probably the same board with misprinted first letter. I also suggest you search for a different cache. And don't forget the 9th cache chip near the chipset.

By the way, look if your cache sockets are even connected to anything... The traces might be going nowhere at all, in that case no cache chips can help.

Found this page: http://www.trinityos.com/HARDWARE/mb_benchmarks.txt It's a benchmark result of some sort and if you search "asa […]
Show full quote

Found this page:
http://www.trinityos.com/HARDWARE/mb_benchmarks.txt
It's a benchmark result of some sort and if you search "asaki" there, it has 0K Cache RAM. Probably the same board with misprinted first letter. I also suggest you search for a different cache. And don't forget the 9th cache chip near the chipset.

By the way, look if your cache sockets are even connected to anything... The traces might be going nowhere at all, in that case no cache chips can help.

Wow a reference to the Asaki 😀

Thank you!

Ill keep trying to get cache acknowledged on the board.

What type of cache chips can I buy? Can I source them from Ebay?

Your motherboard looks like it has support for 128K, 256K, 512K, 1024K asynchronous cache chips. Getting 256K chips is fairly easy, they should work with your system provided that your cache sockets are not fake (propperly connected). 512K and 1024K are more rare, but should also work on your board.

Cache latency (last two digits after "-") is cache response time in nanoseconds. The smaller the better. There's a formula l<=1000/FSB, where l stands for cache latency and FSB stands for your FSB in Mhz. To work at any given FSB cache latency should be less or equal to the 1000/FSB:

25ns is minimum for 40Mhz FSB
20ns is minimum for 50Mhz FSB
15ns is minimum for 66Mhz FSB
12ns is minimum for 83Mhz FSB
10ns is absolute best and should work at up to 100Mhz FSB.

There might be some info in this thread Super Fast 486 Cache - 10 ns

I hope I did not mess up the math.

Edit: You only need 8 larger cache modules, 256K is enough for the 9th (KEY). However, the 9th should be of equal speed or faster.

"Rgart" and that picture... Are you him?

"Rgart" and that picture... Are you him?

Alas no.

Oh well, it did look like it 😉

Are the cache chips backwards compatible?

For example If I have 15ns cache chips - Can I still run the bus at 33mhz?

Also where can I download a copy of pcpbench? I seem to be having difficulty locating one.

Note that the ratings given were "minimal" requirements, and it's almost a waste of time even having cache when you have to crank up the timings and wait states. I'd say this is much closer to reality (and what all my motherboard manuals agree with):

35ns - 16MHz
30ns - 20MHz
25ns - 25MHz
20ns - 33MHz
15ns - 40MHz
12ns - 50MHz

Ideally asynchronous cache is not used with 60,66, 75 and 83MHz FSBs due to the lousy performance when using loose timings. This is why Pentium didn't really take off until after pipelined burst cache became common. For the most part 10ns asynchronous cache in DIP modules is "unobtainium", and I doubt it would make it to 100MHz. Have you ever seen a SS7 motherboard using old style cache?

Yes, cache chips are "backwards compatible".

Luckily, 60 MHz L2 cache at 3-2-2 is still faster than 33 MHz L2 at 33 MHz. I find the 40 and 50 MHz FSB settings problematic for reliably fast L2 speeds. Running a Cyrix 5x86 or Intel DX4 at 60x2 makes for a pretty nice setup. And AMD X5-150 at 3x50 with 2/3 PCI is nice too if you can get the L2 at 2-2-2 or better.

I'm not sure why transitioning from 256 KB double-banked to 512 KB single-banked is more timing sensitive. And if 512 KB double-banked or 1024 KB double-banked will work as well as 256 KB double-banked, or if it is more timing sensitive like 512 KB single-banked cache. One day I'll look into this further. In general, if you are setting up a 486 motherboard, start with 256 double-banked cache and optimise from there, that is, once you've determined the system is stable. Many UMC 486 motherboards will even work OK with 40 MHz FSB and 2-1-1 if you stick with 256 KB double-banked cache, but your PCI bus may be too hyper (40 MHz) or too slow (27 MHz).

Ok guys: Thank you to everyone who replied. I acquired some cache chips out of another board and now have 256 Cache(L1+L2). It appears those chips were fake and I believe the board was manufactured in Brazil. I still can only get very limited details and mentions of the Asaki board online. It is now running much faster and super super stable with a Intel 486DX4 Overdrive 100mhz chip. I am now trying to get a Cyrix 5x86-120 or AMD 5x86-133 working in the machine but having serious difficulties. Basically there is no mention of any divider on the details silk screened on the motherboard. I can put a 100 mhz chip in there and its detected....I put a 66 mhz chip in there and its also detected. It seems I don't need to change any jumpers besides voltage etc.

I really need a manual as a few jumpers are not referenced on the board. What bus speed should I be using for the AMD 133 and the Cyrix 120? When I use an unknown jumper and benchmark tools I can actually get the individual chips recognized "Cyrix 5x86 120gp" or the AMD 133 ADW" however they are both running super super slow. 386 speed....

Any advice?

I will follow up with some details tomorrow when I'm working on it....