The Unofficial Tyan 1836 GX/BX Thunder 100 DLU/DLUAN Web Site

Last Update 28 January 2005

The Tyan 1836 DLU/DLUAN Thunder 100 BX/GX

Coppermine CPU's work on the 1836


In short, this Dual CPU high-end server and workstation mainboard with on-board dual channel fast wide SCSI (Adaptec), audio (Creative), network (Intel) and was Tyan Computer Corporation's flagship product for a short time (too short !). Unfortunately, it is now considered "archived" by Tyan and support for it is minimal (no official BIOS support for Coppermine CPUs and ACPI under Win2000 - but there is a "2.0 beta" BIOS floating around), and that is the reason for the existance of this web site. I bought my Revision A BX S1836DLU (no network/sound) in 1999.
To help identify the exact revision of my board, there are some chip details:

Spread Spectrum Motherboard Frequency Generator (Controls FSB clock) : IC-Works W48S101-04x.
IC-Works was acquired by Cypress so the web site may change in the future. Critical info for over-clocking with SoftFSB .. see below.

Programmable Synchronous DC-DC Controller for Low Voltage Microprocessors (controls CPU Vcc (core)) : Raytheon RC 5051M.
Yes, my chip is a Raytheon but the web site is Fairchild, but its the same chip. This regulator supports a Vcc of as low as 1.3V, which theoretically means my board can handle the Vcc (core) for Coppermine CPUs.

Low Drop Out Positive Regulator (Adjustable) (1.25V to 8V 5 Amp for Voltage Regulated Extension, VRE) : Linfinity LX8384-00CP.
Nothing really useful with this one :)

More information about the 1836, especially the support for Intel's Coppermine CPUs is available at Tyan s1836 CuMine Support.

Incidentally, I've seen 1GHz Intel Pentium III SLOT-1 (SC-242) Coppermine 1.7V Vcore CPUs (Boxed Retail with Fan) on sale in retail outlets here in Malaysia. The price is higher than the equivalent FC-PGA CPU by 10%. Anyone had any experience with these ? I couldn't believe my eyes until I double checked their availability with a friend who works with Intel as well as downloaded the datasheet from Intel's developer's web site and got another surprise...the datasheet is for PIII's from 450MHz to 1.13GHz. Intel's site doesn't mention anything but does that mean Slot-1 is here to stay ? Another funny thing is the FC-PGA datasheet doesn't have the 1.13GHz part ! Is this related to the withdrawl of the FC-PGA 1.13GHz CPUs from the market during the 4th Quarter of 2000 ? Not long after that the Slot-1 CPUs reappeared.
Any ideas ?

Coppermine CPU's will work on your S1836 (added 16th July 2001)

 Scroll down for info on FC-PGA coppermines (added 29th November 2001)
Thanks to Eric (Eric (at) cdc.net) who took the risk and tried it.
Eric's Post.
I also got myself a Pentium III 1GHz 100 MHz FSB CPU from Ebay, S-Spec SL4KL it works fine.
What You Will Need To Do - Important Details:
What follows here is what I did myself which works.
The BETA BIOS WAS from ftp://ftp.tyan.com/bios_beta/1836b30.exe. Unfortunately Tyan has removed the file. If you want the BIOS let me know and perhaps we can arrange something :). Flash this version, this comes from Tyan's own FTP server not the newsgroups or 3rd party web servers.
You need the right kind of Coppermine CPU. There have been attempts to use the FC-PGA CPUs with 133 Mhz FSB with various FC-PGA to Slot-1 adaptors with varying success (some work - single and dual, some don't, I don't know much about this)....but what definitely works is this:
The SL4KL RETAIL P3 1GHz Slot-1 (SECC-2) CPU, 100 Mhz FSB. It needs Vcc(core) to be 1.7V.
This is what I have, it works fine. Eric's report above was with an ealier version of the 2.0 beta BIOS so this should also work for Dual CPU set ups. I only have one unit so I cannot confirm this personally.
My S1836 is a Revision A board (the very earliest) and it is able to provide the Vcc(core) of 1.7V to keep the SL4KL happy. So this *should* work in all later revisions.
So, to summarise, my success formula is - 1836B30 Bios, SL4KL P3 CPU

FC-PGA Coppermines (added 29th November 2001)

I have confirmation that with the right FC-PGA to Slot-1 converter (SLOTKET or SLOCKET) you can get FC-PGA coppermines to work. Unfortunately I do not have details (best to ask on the mailing list ... info on how to sign up is below). However, I do know that you will need the right BIOS. The one I mentioned above (from Tyan's FTP server) will work properly on Slot-1 CPUs. You will need the one modified by Midnight Angel (Thanks, MA, for the BIOS and the explaination !). If you want this BIOS, perhaps we can arrange something :). If you need to contact MA, ask me. Anyway, here are the details of the Midnight Angel modified BIOS in his own words :)

Midnight Angel said:
I modified the BIOS I got from you by inserting the microcode patches for
Coppermine CPUs (FC-PGA package) and dropped some microcodes for CPU
steppings that never were released publicly.
Later, I got another 2.00 beta (which, for example, displays the CPU speed
correctly, and already has the microcode updates for SECC Coppermines.
Since the FC-PGA packaged CuMines seem to need another patch, and the BIOS
only recognizes the CPU by its stepping, disregarding the package
information, I had to throw those SECC settings out of the table,
replacing them with the patches for FC-PGA. This version can be identified
by the string 'FC-PGA patch by Midnight Angel' that is displayed at boot
time.
However, I CANNOT guarantee ANYTHING, especially when things come to dual
processing. It works fine for me (dual CuMines on PowerLeap slockets
(which contain a voltage regulator)), but I heard that the Thunder with
CuMines is prone to noise on certain CPU pins, much as the older versions
of the Tiger are.

So, there you are, even with the right hardware and BIOS, its still pretty much a try it and see if it works for you thing.
Slot-1 Coppemines are getting RARE, so this may be your only option.
There is a thread on the mailing list (29 November 2001) discussing the possibility of dual Tualatins on the S1836 (1.2GHz P3 Tualatin CPUs with 512K cache). Nothing conclusive yet ... no one has tried em so far. Updates here as soon as someone can confirm whether they work or not. Chances are there will be "microcode errors" even with MA's modified BIOS, these are not coppermines after all.

What to do if your board dies and your warranty is over (added 5th June 2001)

On the 1836 are four IRL3103 MOSFETs (Metal Oxide Semiconductor Field Effect Transistor) from International Rectifier . These are power transistors and have something to do with supplying power to the CPUs. There are two per CPU. You can identify them as 4 black TO-220 package devices which look like this: TO-220 Package
One of mine burnt out recently, one of the two beside the CPU2 slot. What happened was when I left the machine running unattended it started smoking and someone turned it off. When I got back, I found the power supply had died, a capacitor had been cremated. I replaced the power supply and it ran fine for a few days when suddenly the system lost power. I left it for a while and powered up again and this time it lost power during POST and never recovered. I opened it up and had a look ....a MOSFET got very hot, burned a 1 inch diameter part of the board which turned the green mask brown, and fractured its plastic case and killed the board. Strangely enough another 1836 owner also reported the same problem except that TWO of his MOSFETs fried .... power supply died, replace PSU, MOSFET(s) died (thanks Snowghost).
At this point I can't really say which caused the other.
Before I start, do this if and only if:
1. The cost of shipping insurance and repair are too high for you to consider and you were going to junk the board anyway.
2. The device you are replacing is a discrete device (on the 1836 only power transistors in general) - Surface mount devices are NASTY for the hobbyist to replace !
3. If you have surface mount desoldering/soldering tools, hey, be my guest, give it a try :)
4. You are reasonably competent in desoldering/soldering - if you are a bit rusty, practise on some other stuff first, like an old 286 motherboard or something :). I was an electronics hobbyist (no time nowdays) and made my own PCBs and stuff, and worked more than a year in a electronics manufacturer making prototype PCBs, so soldering isn't a problem for me. Never done any soldering in your life ? Fixing your 1836 isn't the time to learn !
5. You have the right tools - especially the soldering iron - pen type, 1-2mm bit, earthed (grounded), 15-25 Watt power is about right. The 100 Watt pistol grip monsters are not generally useful on computer motherboards :) You have been warned !
6. You can identify the dead component reasonably well - if you can't see what died, how do you know what to replace ? Mine was a dead giveaway, with a big brown spot on a green PCB, discoloured heatsink and fractured component package ! This may be the case for power components, but logic components don't normally show any signs of death.

OK, that out of the way, this is what I did (it doesn't mean this is the best way).
I unscrewed the MOSFET first, straightened the pins (so the chip is upright instead of lying flat and removed the heatsink.
I then cut off the chip (leaving the 3 pins) to see if removing it would leave at least my CPU 1 working ... no go, board refused to power up. In hindsight this was a good way to do it, it made the job a desoldering much easier.
I then went to look for a replacement, it took me a week to get a replacement, the IRL3103 is a single source chip (only one manufacturer), no equivalents (my replacement was shipped in from Singapore and cost me US$10).
I then desoldered the 3 pins. It turns out this is the best way to do it....take the pins out one at a time rather than trying to remove the chip will all three pins attached. Why ? The PCB holes are plated through and there is track on both sides, so you need to make sure when you desolder that all the solder melts all the way through to the other side of the board.... otherwise when you pull the pin, some circuit board will come off with it. This is a TWO person job (I did this single handed but I don't recommend you try), one person to apply the soldering iron and the other to pull the pin out (with pliers :).....don't force it, if it doesn't come out easily, the solder hasn't melted. Don't be afraid to keep the iron on quite a long time, 10 secs (25 Watt soldering iron).
Once you get the pins out, you need to clear the holes for the new chip, same thing, 2 people, one with the iron and the other with the vacuum desolder tool.
Make sure you bend the pins of the new chip at a 90 degree angle before you put in and slide the heatsink underneath (check pin to heatsink clearance) and screw it in before you solder it in. If you bend the chip after soldering (like me) the pins can get really close to the heatsink...too close for comfort. Make sure you get the chip the right way around, there are 3 other examples on the board, if yours doesn't look like those, something is wrong !
Check your work, look for dry joints and solder bridges !
That's it, plug in the power supply, turn it on and watch for smoke :). If it powers up and passes the 30 second smoke test, then plug in video cards, disk drives etc and test it properly.
Good Luck !

UPDATE 9 October 2001

There is something more serious wrong with my board besides the fried IRL3103 MOSFET. It turns out that the MOSFET frying is a symptom of the problem and not the root cause (of which I have no idea, any suggestions will be greatly appreciated). When I tried dual 550Mhz P3s it seemed ok for a while (30 seconds), until I got the smell of very hot electronics. Quickly turning off the power, I suspected the the MOSFET and touched the heatsink, thereby burning my finger. Ok, not good. I removed the 2nd P3 (the one supplied by the hot MOSFET) but 1836 refused to power up for quite a while, I almost gave up on it but gave it one last try after everything had cooled down and it powered up. It is happily running FreeBSD 4.x (4.11RC as of Jan 2005) now albeit on one CPU.
Incidentally I'm having a funny problem with FreeBSD on the 1836. Any FreeBSD gurus reading this, I would be most grateful if you can give me your opinion on this problem:
This only happens during my source tree CVSUP (run using make update in /usr/src). I keep getting the following error messages to the console:

kernel trap 19 with interrupts disbled

/kernel: NMI ISA 3c, EISA ff

/kernel: NMI ISA 2c, EISA ff
My 1836 has one narrow SCSI drive, 256MB RAM and two Intel 82559 NICs. (Incidentally, this error has appeared on similarly configured systems on OpenBSD as well). My guess is that CVSUP generates a lot of disk and network activity simultaneously and this is causing some unknown conflict. Heavy disk and CPU activities like make world and make buildworld and heavy network activities like tcpdump do not generate any errors.

The errors are not usually fatal but rapidly fill up syslog. Occasionally CVSUP will terminate with errors.
The code that flags the NMI is a catch-all for an unknown cause that doesn't generate a panic:
/*
 * Handle a NMI, possibly a machine check.
 * return true to panic system, false to ignore.
 */
int
isa_nmi(cd)
        int cd;
{
        int isa_port = inb(0x61);
        int eisa_port = inb(0x461);
        if(isa_port & NMI_PARITY) {
                panic("RAM parity error, likely hardware failure.");
        } else if(isa_port & NMI_IOCHAN) {
                panic("I/O channel check, likely hardware failure.");
        } else if(eisa_port & ENMI_WATCHDOG) {
                panic("EISA watchdog timer expired, likely hardware
                failure.");
        } else if(eisa_port & ENMI_BUSTIMER) {
                panic("EISA bus timeout, likely hardware failure.");
        } else if(eisa_port & ENMI_IOSTATUS) {
                panic("EISA I/O port status error.");
        } else {
                printf("\nNMI ISA %x, EISA %x\n", isa_port, eisa_port);
                return(0);
        }
}

Selected Archives of alt.comp.periphs.mainboard.tyan

These are selected and edited archives of USENET postings to the alt.comp.periphs.mainboard.tyan newsgroup. Flames have been removed as are other non-useful postings. Email addresses have been modified to prevent harvesting by spammers but are retained to give credit to the posters. If you wish to have yours removed please email me. The purpose of this is to function as a reference for the 1836 until such time a proper FAQ can be compiled.
Suggestions or better still, help, to improve this site will be more than welcome ! :)
Disclaimer : All Trademarks are properties of their respective owners.
Go to the Archives (23 Jan 1999 to 21 November 2000) - Warning 600K+ Text File. Thanks to Mike Stewart for this section.
I'm still in the midst of editing a much larger text archive. 30% done as of 19 Feb 2001. Work has stalled, volunteers ?

The Unofficial Tyan 1836 GX/BX Thunder 100 DLU/DLUAN Mailing List

UPDATE January 2005: A server disk crash killed the mailing list. I do not intend to restart the mailing list unless there is a great demand for it, which I doubt. Alternatively check out postings in alt.comp.periphs.mainboard.tyan
or email me at
The purpose of this list was purely for 1836 owners to discuss issues and disseminate information directly relating to the Tyan 1836 mainboard and unofficially exchange BIOS images.

Overclocking the Tyan 1836 DLU/DLUAN Thunder 100 BX/GX

Although the BIOS settings allows you to overclock the board by changing the FSB to 103-124 MHz, the board will pause during startup to display a warning message which requires the user to press a key to continue .... obviously unacceptable, especially for servers. A BIOS option is available to disable the message BUT it also disables the overclocking by forcing the default FSB ! Naughty trick by Tyan !
Fortunately, for those of you who want to overclock your board, you can have your cake AND eat it. Help comes in the form of a Win32 program called SoftFSB H. Oda (SoftFSB is no longer available for download here but he has other other cool programs.

SoftFSB can be downloaded here. (21st February 2001).

SoftFSB allows you to change the FSB on the fly so that you can have this in your startup. I have tested SoftFSB successfully, overclocking my Celeron 300A to 463 MHz with 103MHz FSB on my Thunder 100 (since my upgrade to dual PIII-550, I've since stopped overclocking since there was a 2D to 3D mode switch lock-up problem with my AGP Voodoo3 2000 and 3000). Details of how I got SoftFSB to work on *MY* Tyan Thunder 100DLU S1836BX Revision A mainboard is below:

WARNING - DISCLAIMER

Whatever you do with SoftFSB is entirely your own responsibility. It worked on my system but it does not mean it will work in yours and if your machine dies an untimely death (permanent or otherwise), that is not my problem or the author of SoftFSB :). It would be wise to shutdown all other programs before playing with SoftFSB. You can lock up the system into a state where a HARD POWER DOWN is required to bring it back. My system has also had a situation that even after a HARD POWER DOWN (that means turning off the power at the wall socket and not just holding down the power button for 5 seconds), the FSB remained at the overclock setting ! Read the documentation before trying anything !
The Clock Generator is an IC-Works W48S101-04x PLL (Phase Locked Loop) chip. There is a setting for selecting the S1836 in SoftFSB or the clock generator chip itself, use whatever works for you. Another IMPORTANT thing, the clock generator settings are WRITE-ONLY, that means you can set the FSB but you CANNOT read the FSB reliably. Make sure SoftFSB is set up to WRITE-ONLY, if you allow SoftFSB to attempt to READ from the chip, you will need to do a HARD POWER DOWN. Using these settings I successfully over-clocked my Celeron 300A as well as the current dual PIII-550s (just to see if it worked).
Page Created on February 14, 2001 and Last Updated on January 28th, 2005