stamasd wrote on 2022-02-06, 00:12:

I have a Tyan Tomcat IV board (model S1564) which is very similar to the one here except the second CPU socket is not installed: […]
Show full quote

I have a Tyan Tomcat IV board (model S1564) which is very similar to the one here except the second CPU socket is not installed: https://www.ultimateretro.net/en/motherboards/5131 So it's a single CPU version of this otherwise dual-socket7 mobo.
I've had it for some time but never got around to use it in anything yet. I used the additional jumper settings in the docs on the page above, kudos to whom made that research about voltages etc.
It looks like the particular VRM on my board errs on the low side of those voltages: with jumpers set for 2.15V, I actually measure 1.97V; 2.25V measures 2.04V and so on. So for the actual CPUs I used, I set the jumpers according to the measured voltages, not the theoretical ones.
I used 2 CPUs in the board, both at the same FSB and multiplier settings, the only difference between them is the core voltage. The FSB was set at 66, and multiplier 3.5 (both are the max this board provides).
First CPU, AMD K6-233 ANR, Vcore=3.2v (measured voltage set actually 3.19V on the 3.3V setting).
Second CPU, Cyrix MII/400 (95x3 so 285MHz) with Vcore of 2.2V. I ran that one at 66x3.5 so 233MHz because this board cannot provide more. Vcore measured was 2.21V on the 2.45V position. So it's clearly underclocked, but for now I was testing it like that.

Both CPUs POST and boot properly in DOS. However when I try to run programs on them, they crash. For instance, Phil's benchmark pack: I can run the Doom benchmark on the Cyrix, not on the AMD. Neither runs the Quake benchmark. Neither runs Norton SysInfo. Neither will run the Speedsys test. Neither will run cachechk. All of these crash.

The other components I used in this test build: video card Matrox G200 PCI; 4GB CF card through adapter; 16MB of FPM 72-pin RAM; all of them are known good and running perfectly fine in another motherboard. Same for the two CPUs, they both run fine in another motherboard with the tests above not crashing.

I'm not sure what to do next.

(one more weirdness, the motherboard will refuse to use any ISA video card; I tested 3 or 4 of them and neither gives a video output; PCI video cards, all I tested work fine).

Edit: the BIOS is revision 1.01. I see there is a revision 1.02 and a patched 1.02, but I don't want to flash a new BIOS while the board doesn't act properly and many programs crash.

FWIW, I downclocked the CPUs even further, at 66x2=133MHz. While there is some marginally better stability (for instance the K6 is able to complete Speedsys 4.71 without crashing), most of the other unstable behavior continues.

That would be me who did that research 😀

Also, you may find one of the links in my signature interesting. Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

There are a number of things that could explain the difference in voltage between what I measured on mine and what you get on yours. A a slight difference isn’t impossible. Heck, you may have a different voltage chip than mine but it should be similar. As yours is.

3x is actually the max the board provides, 3.5x is the 1.5x setting. But that’s simantics. Though by that logic 6x is the max setting because on k6-2/3 2x=6x

That cyrix chip may be able to set a 4x or 4.5x setting edit: yes but only to 4x, but you’ll have to do a mod or see if that top jumper goes to bf2
(Solved) Cyrix 400gp/366gp multiplier settings

I’ve had issues with cf cards on the ide on this board and several others. Even the latest bios doesn’t fully fix the glitchyness of the onboard ide, But booting from a xt-ide seems to work very well. (Though it is slow) personally a promise tx4 sata II card and a ssd is the best way to go imo on this board. (Or a better card if you know one)

Okay, now on to your problem, first off, I’m like 99% sure that you can just flash jan’s s1564d bios to that board and it would work on the s1564s so do that, and if it didn’t work for some reason just install the APIC and remove the resistor (described in my other thread) and that should fix it. (Apic is like 10$ Shipped) but like I said, I believe it should work without the APIC. I think your 400gp isn’t fully supported by the bios version you have also btw. (Even if the 400gp is detecting). If nothing else, jan’s j2 bios version for this board is well established to be stable, spare the as of yet unresolved ide glitches. (Just leave the detection in bios to auto and it should be fine for most things.) (none of the bios versions have fully patched IDE)

Anyway, my bet is your problem is bios or more likely the way you installed your ram. Not that you have told us what your ram is, just experince with this board with issues like this, it was almost always the ram configuration, other possibilities are which slots which cards are in. Sometimes one card will like one slot but not another.

Remove all ram except a matched set in space 1,2 and try again.

I have many pci vid cards that don’t work on this board, it is quite picky. (Along with most socket 7 boards) Consider yourself lucky all yours work haha.
The best vid card I have tested so far is the low profile radeon 7500 64mb i haven’t tested any isa video cards on it as I don’t have any.

Ps: cmos reset is in the same jumper block as the voltage settings, don’t mess with the jumper by the bios chip., the easiest way to flash this is with a chip programmer. Which is how I usually do it. When first I got it, I believe I did do a flash through either awardflash or uniflash. If your bios version isn’t listed on the ultimate retro (for the s1564s a different page than you linked), then please dump and upload it so we can have it.

https://www.ultimateretro.net/en/motherboards/5132

Ps, more than one link in my signature is relevant to this board, may read through them.

Btw, 66mhz is indeed the most you can hope for from the fsb, I have looked into this quite extensively.

The fastest processor you can install in this board is a k6-3+ 400-550 (running at 400mhz) which actually speeds it up SIGNIFICANTLY and adds another level of cache to the system for a total of three caches.

There is zero chance of running that in dual, (another thing I have looked into extensively)

The max ram is 512mb even though 1gb will fit, and 768 will detect, 512 is the most that can actually work. 768 is unstable, (will actually give issues like what you described)

Nice, thanks for the long write-up. Actually, I had had this thought for a few years (when I acquired this board) "couldn't I solder a socket there and..." 😀 I happen to have a few spare socket7 zif parts.
But for that, I'll have to have this board running stable first. I have read through your other threads, really good job there.

The RAM I use is two 8MB FPM modules which I know work well in other motherboards. They are inserted in SIMM slots 1 and 2.

I have gone ahead and flashed the Jan BIOS in my TL-866; it did not make a difference, I see the same unstable behavior. I did save the original BIOS but it's identical to the one listed as "05/06/97-i430HX-2A59FT59C-00 v1.01" at https://www.ultimateretro.net/en/motherboards/5131#downloads

I have to get the RTC going before I go any further, as I have to eliminate that as the source of instability. I didn't do it yet because I was out of CR2032 battery holders, I should have a batch delivered today and then I can modify the Dallas chip.

One more thing I'd like to mention, while swapping the CPUs yesterday for tests I once forgot to put the jumpers back to 3.2V from 2.2V when dropping in the K6 again. I was amazed that it booted DOS at 1V less core voltage than it should have had. 😀 But it was even more unstable, crashing even when doing a simple "dir"

So I just did more testing.

J21 does not seem to do anything. I think it's not connected at all. I tried with the Cyrix MII all combinations of J19+J20+J21, and it makes no difference whether J21 is on or off, the other 2 set the multiplier. I think they put J21 there in anticipation that it may be used for BF2, but in the end they didn't actually connect it.

As for the instability, surprisingly it seems related to Phil's benchmark pack. The crashes still occur when I run his scripts and the benchmarks from within. But I tried some other programs separately from that pack, including some fairly demanding late DOS games like Frontier Elite 2 and Frontier First Encounters, and they run perfectly fine without crashing. I have to install a separate copy of quake to test with it also.

Well I've solved one problem. Quake timedemo runs fine after setting a PATH variable (I was booting off a bare floppy with no autoexec, and getting the "ompt=: cannot open" error. But I've set an autoexec.bat with a PATH variable and now quake runs).

But that doesn't solve the other weirdness. Some other programs like Norton System Info and pcbench won't start properly with a "divide overflow" error and then the whole system locks up, needs to be power cycled. (edit) this is still without a working RTC installed.

DS12887 prepped for the mod. C'mon Amazon deliver those battery holders faster. 😀 (Yes I know some pins aren't straight, I'll fix that before reinstalling).

stamasd wrote on 2022-02-06, 12:13:

Nice, thanks for the long write-up. Actually, I had had this thought for a few years (when I acquired this board) "couldn't I so […]
Show full quote

Nice, thanks for the long write-up. Actually, I had had this thought for a few years (when I acquired this board) "couldn't I solder a socket there and..." 😀 I happen to have a few spare socket7 zif parts.
But for that, I'll have to have this board running stable first. I have read through your other threads, really good job there.

The RAM I use is two 8MB FPM modules which I know work well in other motherboards. They are inserted in SIMM slots 1 and 2.

I have gone ahead and flashed the Jan BIOS in my TL-866; it did not make a difference, I see the same unstable behavior. I did save the original BIOS but it's identical to the one listed as "05/06/97-i430HX-2A59FT59C-00 v1.01" at https://www.ultimateretro.net/en/motherboards/5131#downloads

I have to get the RTC going before I go any further, as I have to eliminate that as the source of instability. I didn't do it yet because I was out of CR2032 battery holders, I should have a batch delivered today and then I can modify the Dallas chip.

One more thing I'd like to mention, while swapping the CPUs yesterday for tests I once forgot to put the jumpers back to 3.2V from 2.2V when dropping in the K6 again. I was amazed that it booted DOS at 1V less core voltage than it should have had. 😀 But it was even more unstable, crashing even when doing a simple "dir"

Try running only the video card in the lowest pci slot and see if you have problems still. Which ones do you have btw? I forget if you said.

If those sticks are matched, then I agree probably not the issue. (But bios ram settings may be)

Try using a add in card for your hard drive in the top slot.

I’ll have to post the bios settings I am using, set your ram settings to the highest numerically, and enable all the other settings under chipset, the recovery time settings can be n/a but set system bios cachable to disabled, video cachable can be enabled.

I’ll try to post up my bios settings when I can.

Be sure to cut the old battery out of that circuit on the dallas, or the new battery will be drawn down a bit by the old battery.

stamasd wrote on 2022-02-06, 12:53:

So I just did more testing.

J21 does not seem to do anything. I think it's not connected at all. I tried with the Cyrix MII all combinations of J19+J20+J21, and it makes no difference whether J21 is on or off, the other 2 set the multiplier. I think they put J21 there in anticipation that it may be used for BF2, but in the end they didn't actually connect it.

As for the instability, surprisingly it seems related to Phil's benchmark pack. The crashes still occur when I run his scripts and the benchmarks from within. But I tried some other programs separately from that pack, including some fairly demanding late DOS games like Frontier Elite 2 and Frontier First Encounters, and they run perfectly fine without crashing. I have to install a separate copy of quake to test with it also.

Oh heck, well then there ya go!!

Maybe doesn’t like your choice in video card or corrupted?

stamasd wrote on 2022-02-06, 13:19:

Well I've solved one problem. Quake timedemo runs fine after setting a PATH variable (I was booting off a bare floppy with no autoexec, and getting the "ompt=: cannot open" error. But I've set an autoexec.bat with a PATH variable and now quake runs).

But that doesn't solve the other weirdness. Some other programs like Norton System Info and pcbench won't start properly with a "divide overflow" error and then the whole system locks up, needs to be power cycled. (edit) this is still without a working RTC installed.

Your issues may disappear once you can set proper bios settings.

I can confirm that all of the weirdness was caused by the non-functioning RTC. This must be one of those motherboards that almost works without a RTC, but not really. Mod completed, everything works perfectly fine now.

One last mod I have to do today to this board - is unsolder the AT keyboard connector, and instead solder PS/2 kb+mouse connectors. The pads are there under the AT kbc.

And that is done too. Keyboard and mouse confirmed working in the new configuration.

stamasd wrote on 2022-02-06, 23:02:

f.jpg

g.jpg

You know what, I’ve considered adding the mouse port. I like it!! Seems like a crime to remove the AT port though 😂

https://www.vogons.org/download/file.php?id=129873&mode=view […]
Show full quote

😖 ahhh!! Empty socket footprint! 😖

Question,

Will the keyboard work in both ports?

I was thinking about maybe getting a green/purple port for the header by the AT port and leaving the AT port.

There isn't enough space to mount a PS/2 mouse connector without removing the AT one. If the AT had been 1-2mm more to the left it would have been possible, but like it is they would overlap.
No, the PS/2 connectors for mouse and keyboard are not interchangeable. A PS/2 keyboard will not work if plugged in the mouse connector. The communication protocols they use are different.

stamasd wrote on 2022-02-06, 23:02:

f.jpg

g.jpg

I love it

Sphere478 wrote on 2022-02-07, 03:33:

😖 ahhh!! Empty socket footprint! 😖

I've ordered a few APIC chips, but it will be a few weeks until they get here as they're getting shipped from Poland to the US.

stamasd wrote on 2022-02-07, 17:31:

Sphere478 wrote on 2022-02-07, 03:33:

😖 ahhh!! Empty socket footprint! 😖

I've ordered a few APIC chips, but it will be a few weeks until they get here as they're getting shipped from Poland to the US.

The ones I ordered came from the usa

stamasd wrote on 2022-02-07, 12:45:

There isn't enough space to mount a PS/2 mouse connector without removing the AT one. If the AT had been 1-2mm more to the left it would have been possible, but like it is they would overlap.
No, the PS/2 connectors for mouse and keyboard are not interchangeable. A PS/2 keyboard will not work if plugged in the mouse connector. The communication protocols they use are different.

Interesting, reason I asked is because I have encountered boards where the keybord worked in both ports

Also there is such a thing as a ps/2 splitter for running a keyboard and a mouse from one port

Do you know of a atx I/O shield that will work with two ps/2? That could be ordered?

Sphere478 wrote on 2022-02-07, 19:41:

The ones I ordered came from the usa

I didn't order mine from ebay or any other such place. I found them cheaper elsewhere. I can wait.

stamasd wrote on 2022-02-07, 19:55:

Sphere478 wrote on 2022-02-07, 19:41:

The ones I ordered came from the usa

I didn't order mine from ebay or any other such place. I found them cheaper elsewhere. I can wait.

Btw, did you happen to figure out the pinout for the ps2 header in doing research for this project? I can check the manual, if it’s not there, we should add it to Ultimate Retro

Edit: it’s in the manual