VOGONS


Tyan S1562 Dual Pentium MMX succes

Topic actions

First post, by flupke11

User metadata
Rank Oldbie
Rank
Oldbie

In an older thread of mine, I mentioned a Tyan Dual Socket 7 board I had. Due to the lack of a decent AT-power supply, I never got round to testing it.

Now that I've found some time and space to relish in my hobby, I tried the Tyan with 128 megs of EDO ram, in 4 of the 8 simm slots.

The mainboard was fitted with dual Pentium 200's, and worked.

I tried installing two MMX 233's, and that worked as well! Setting the jumpers to the Pentium 100 setting gave me the full 233 mhz i wanted (I found that out on this board, thx to all).

I'll try and install WindowsNT on it, as soon as I've decided out what hard drive I'm going to put in it.

Attachments

  • Filename
    IMG_0206.jpg
    File size
    39.19 KiB
    Downloads
    92 downloads
    File license
    Fair use/fair dealing exception
  • Filename
    IMG_0210.jpg
    File size
    39.76 KiB
    Downloads
    102 downloads
    File license
    Fair use/fair dealing exception

Reply 1 of 36, by Unknown_K

User metadata
Rank Oldbie
Rank
Oldbie

Cool, always wanted one of those rare dual P1 systems (or even rarer dual 486).

Collector of old computers, hardware, and software

Reply 2 of 36, by luckybob

User metadata
Rank l33t
Rank
l33t

2 things.

Please tell me your jumper settings, I was under the impression that 200 non-mmx was the fastest this board would take.

Would you be kind enough to run a speedsys benchmark and post the result? thanks!

It is a mistake to think you can solve any major problems just with potatoes.

Reply 3 of 36, by Jolaes76

User metadata
Rank Oldbie
Rank
Oldbie

due to some changes in the split voltage but staying on 66 mhz bus allows the 1.5x multiplier to be interpreted as 3.5x for MMX processors 😎

"Ita in vita ut in lusu alae pessima iactura arte corrigenda est."

Reply 4 of 36, by flupke11

User metadata
Rank Oldbie
Rank
Oldbie

I've finally been able to do some rounds of testing:

p233 dual HX results

Cachechk W: 92,9
Cachechk R: 88,2
Speedsys Throughput: 85,42
Speedsys R: 80,73
Speedsys W: 85,31

Strangely, the writing speed is higher than the read speed.

The system beats my dual Ppro Overdrive on these memory write results. The PR440FX is faster on reads (about 100MB/s) but puts down a poorly 70 MB/s on the speedsys reads.

Reply 5 of 36, by elianda

User metadata
Rank l33t
Rank
l33t

I have a similar board running. So do you have problems with the Linear Voltage Regulator overheating after about half an hour if you push both P233MMX in NT4 f.e.
I also noticed the problem that with a full memory setup, EDO RAM produces memory errors if accessed from CPU 1. You can't detect this with memtest86, as it runs on CPU0 only.

Retronn.de - Vintage Hardware Gallery, Drivers, Guides, Videos. Now with file search
Youtube Channel
FTP Server - Driver Archive and more
DVI2PCIe alignment and 2D image quality measurement tool

Reply 6 of 36, by luckybob

User metadata
Rank l33t
Rank
l33t
elianda wrote:

I also noticed the problem that with a full memory setup, EDO RAM produces memory errors if accessed from CPU 1. You can't detect this with memtest86, as it runs on CPU0 only.

wait... WHAT? Do you have any more data on this? I benched mine because I kept getting random memory errors during mem-test86 (smp version).

I came to the conclusion that I had a bad memory socket and was going to replace them all one day. Please tell me you have more information.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 7 of 36, by elianda

User metadata
Rank l33t
Rank
l33t

Well actually I got the board fully equipped with 256 MB, Bank 1,2 with EDO and 3,4 with FPM DRAM.
So I changed the FPM to EDO DRAM, memtest86 was good (non SMP), DOS etc. ran fine too.
But as soon as I installed NT4 and it booted up first it went quickly BSOD. The typical random BSODs that point to memory errors.
It actually took me a while to realize that this must be a memory error on CPU 1, while memtest86 showed that everything was working nicely.
Then a rather tedious testing started with changing memory, starting NT4, changing memory timing, starting NT4... (a bit more than a week at least)
In the end it worked out for me, that the original config with split EDO/FPM was on purpose, because it worked. So someone who setup the system originally must have known of this issue.
It is a Tyan S1563 with two P233MMX.

Retronn.de - Vintage Hardware Gallery, Drivers, Guides, Videos. Now with file search
Youtube Channel
FTP Server - Driver Archive and more
DVI2PCIe alignment and 2D image quality measurement tool

Reply 8 of 36, by sliderider

User metadata
Rank l33t++
Rank
l33t++

For that system I'd be more likely to go with Windows 2K than NT. 2K raises the bar for DirectX support all the way to DX9, making a larger number of video cards available for use. You'll be stuck with using the DX5 beta release with NT 4. I am using 2K on my dual PPro machine and it's not heavy on that system.

Reply 9 of 36, by Xolares

User metadata
Rank Newbie
Rank
Newbie

WOW thats rare and i have never seen a P1 Duel cpu system before and i use http://imageshack.us/a/img96/3053/20130320171509.jpg

2x P3 800MHZ - 1GB PC133 - 3DFX Voodoo 5 5500 64MB - Soundblaster AWE32 28MB 32Pin ISA & Music Quest ISA MIDI I/O + Roland SC-88 Pro - 2x IDE to CF 16GB Flash HDDs-Win98SE SP3 137GB+-Windows 2000 SP4R2-17" CRT NEC MultiSync 1600x1200

Reply 11 of 36, by cdoublejj

User metadata
Rank Oldbie
Rank
Oldbie

the question on my mind is what os you will run? i'm certain it had some form of linux or unix on it back the day, oooohh or some sort of sunmicrosystems os. did they even have windows server back then?

Reply 12 of 36, by feipoa

User metadata
Rank l33t++
Rank
l33t++
elianda wrote:
Well actually I got the board fully equipped with 256 MB, Bank 1,2 with EDO and 3,4 with FPM DRAM. So I changed the FPM to EDO D […]
Show full quote

Well actually I got the board fully equipped with 256 MB, Bank 1,2 with EDO and 3,4 with FPM DRAM.
So I changed the FPM to EDO DRAM, memtest86 was good (non SMP), DOS etc. ran fine too.
But as soon as I installed NT4 and it booted up first it went quickly BSOD. The typical random BSODs that point to memory errors.
It actually took me a while to realize that this must be a memory error on CPU 1, while memtest86 showed that everything was working nicely.
Then a rather tedious testing started with changing memory, starting NT4, changing memory timing, starting NT4... (a bit more than a week at least)
In the end it worked out for me, that the original config with split EDO/FPM was on purpose, because it worked. So someone who setup the system originally must have known of this issue.
It is a Tyan S1563 with two P233MMX.

This is curious. I've noticed many pre-millennium boards don't like to have their memory size maxed out or have all banks filled and use fast timings- as if the chipset cannot cope. For your system, do you get errors in NT4 if you fill all banks with, say, only 8 MB EDO sticks? Also, what happens if you use only 4 sticks of 64 MB EDO for 256MB total, or 4 sticks of 64 MB FPM, or 8 sticks of 64 MB FPM, or how about just 2 128 MB EDO sticks? That last option is probably unsupported.

Plan your life wisely, you'll be dead before you know it.

Reply 13 of 36, by BastlerMike

User metadata
Rank Member
Rank
Member

I experienced strange behaviour when I tried running a dual CPU setup with the S1562 board.
Running one CPU with cache module installed --> no problems
Running two CPUs with cache module not installed --> no problems
Running two CPUs with cache module installed --> permanent lockups and freezes

After some research and digging around in some old newsgroups I figured out that TYAN equipped some of these boards with incompatible cache modules. The only module that works with a dual processor setup on this board is called "S1543-PCB-04" !

Reply 15 of 36, by ThresholdAU

User metadata
Rank Newbie
Rank
Newbie

Hi, I'm really keen to understand the jumper settings you used here.
Are you sure you not using the S1564 ? that has VIO ?
I can't get my Single S1562 to post at 233 for my 233mmx with any combination of jumpers.
Can you share some more info if you are still active?

Reply 16 of 36, by flupke11

User metadata
Rank Oldbie
Rank
Oldbie

I hope I am allowed to resurrect this old topic.

I've recently pulled out the MB from storage to test and install Windows 2K on a CF-to-IDE.

As I haven't installed an OS before, the system's slowness only recently became apparent. Turns out the 512kb COAST cache module is either dead or not recognised by the system. Any ideas how to test what is wrong? I'd like to have this as a working system. Although the install in the end went without any hiccups, and the system is stable with the 128 MB of ram and both P233MMX, it is too slow due to the L2 cache being absent.

Reply 17 of 36, by flupke11

User metadata
Rank Oldbie
Rank
Oldbie

Slow replies, my apologies:

- Settings for the dual P233MMX to be detected are:

J13: 1-2
J19: OFF
J20: OFF

I suppose you also need the latest BIOS. On the board I have V4.02 is installed (V4.02 - 07/09/098-i430HX-2A59FT59C-00).

- Update on the cache question.

The COAST module S1543-PCB-04 is not faulty, it simply does not make proper contact anymore in the slot where it is seated. With some fiddling, I managed to find a position in which the module is at least detected at boot.

Windows 2K (vanilla installation) is not stable with the cache detected at boot. It boots, but with all sorts of .exe crashes after which the system usually self-reboots. Without the cache module, the system is stable, but very slow as you can imagine.

I suppose a system does not like it when the L2 cache is suddenly disappearing and that this causes the .exe errors?

Reply 18 of 36, by flupke11

User metadata
Rank Oldbie
Rank
Oldbie

Update, fixed the issue with the COAST module more or less by a thread:

Webcam-1567962779.png
Filename
Webcam-1567962779.png
File size
263.01 KiB
Views
2122 views
File license
Fair use/fair dealing exception

CPU-0 gets 219 on Integer and 135 on Floating
CPU-1 gets 209 on Integer and 136 on Floating

System seems to behave now with the cache all bound up.