Sunday, March 20, 2016

Electronic Recycle Bin Rescue Super Socket 7 System

My town has a shipping container where you can rid yourself of old electronics like TVs, computer systems and monitors and printers.  The town charges a small fee per item.  Even in my small town there seems to be no lack of residents with old PCs.  Usually the container will have three to four PCs dumped in it by the time it is picked up and sent to a recycling center.  I have been known to explore the leftovers and have taken the occasional graphics or sound card from PCs.

Last week, I saw a system that interested me.  Usually I see stock Dell or HP machines from the mid-2000s, but this system was inside an AT-style enclosure.  My interest was naturally peaked.  The system did not have a cover and I could see a socket 7 motherboard.  There was no heatsink on the CPU, which identified itself as an AMD K6-2 350MHz.  I did not need any more advertising, I had to have this machine.  Three discrete trips to the bin later, I had the CPU, the motherboard on a removable tray, the CPU heatsink and all the header connectors I saw.  I did not take any disk drives, I do not need floppy or CD-ROM drives and there is an unwritten rule about taking hard drives (to protect residents against identity theft.)  I left the case, drives and power supply for the recyclers.

The unknown motherboard
Motherboard Overview :

As this is a late AT class motherboard, there are headers and ribbon cables for just about everything.

Identifying the Motherboard :

Not to busy looking from this angle
The VIA Southbridge chip is the VT82C686A (labeled as HT82C686A on this board).  Although the identity of the Northbridge chip is obscured by the headsink, the motherboard has integrated video but no external chip to provide that functionality.  According to the Southbridge reference manual, it was designed to integrate with three VIA Northbridge chips, VT82C598, VT82C501 and VT82C691.  The VT82C598 (Apollo MVP3) is a Super Socket 7 chip which provides AGP, the VT82C501 (Apollo MVP4) is a Super Socket 7 chip has integrated 2D / 3D graphics and the VT82C691 (Apollo Pro) is a Socket 8 / Slot 1 chip.

Obviously, the VT82C691 is not the chip and the VT82C501 looks better than the VT82C598 because the VT82C501 advertises integrated graphics, which fits with my board. Therefore, the Northbridge is a VT82C501.  The MVP4 is otherwise identical to the MVP3.

There is no official name of the motherboard or who manufactured it on the board itself, but judging by this thread it is highly likely a PCChips product.

The heatsink on PC Chips' boards is very distinctive.  Other manufacturers like ASUS and Gigabyte proudly identify themselves, but PC Chips left their identity to their golden heatsink. It uses a VIAGRA PC100 chipset.  No chipset was ever more unfortunately named and a Google search does not really work well.

This board is not documented in TotalHardware '99.  However, there are board shots of many of PC Chip's motherboards at  Even in low resolution, these photos correspond almost exactly to my board :

Therefore, my motherboard is a PC Chips M585LMR.  Its PDF manual can be downloaded from here :

There is a thread about the motherboard here :, but without pictures it would be difficult to associate this board with the thread.

Integrated Functionality :

This chip incorporates the Trident Blade 3D AGP 3D Accelerator.

The LAN is provided by the Davicom 9102F chip.  It is a 10/100Mbit PCI-based LAN card, and Windows 9x has drivers for it.  It cannot be disabled, but there is no reason you have to use it except that wasting a PCI slot on a LAN card is silly.

The audio and modem is provided by the SoundPro HT8738AM chip, which has dual functionality.  The audio portion is provided by a core functionally identical to the C-Media CMI8738.  The modem capability is provided by a core fundamentally identical to the PCtel HSP56 MicroModem.  This is a Winmodem, so do not expect good performance.


The only jumpers on this board are JBAT1, JP2, JP4 and JP5.  JBAT1 clears the CMOS memory, JP2 enables and disables the onboard audio and modem functionality and JP4 enables and disables the onboard LAN.  JP5 selects the SPDIF voltage output for the SPDF I/O connector, allowing for 5 volt or .5 volt.  5 volt is suitable to power an optical or TTL output, while 0.5 volts is appropriate for a coaxial peak-to-peak output.  There is also an audio and modem disable in the BIOS, but hardware disable seems better than software disable for completely eliminating the device from Windows.

There is information silkscreened for JP6 (CPU multiplier selection) and JP9 (front side bus speed selection), but no pins or dipswitches on the board to access these selections.  These settings are entirely controlled by the BIOS.

Headers :

Then you look on the other side of the board
This board has lots and lots of pin headers.  Some of them appear to be standard, others are not.  

COM1 : Serial Port, standard 10-pin header (pin 9 is key)
PRINT1 : Parallel Port, standard 26-pin header (pin 26 is key)
VGA : Monitor Display port, standard 16-pin header.  Uncommon but not unknown, pin 16 has no connection)
LAN1 : RJ-45 port module, non-standard.  
DAA1 : Winmodem 2 x RJ-11 module, non-standard.  This does not have a ribbon cable, it attaches directly to the motherboard and its header is positioned so the module aligns with a case slot.  This can be found by searching for PCTel Fax/Modem DAA module
SOUND1 : Game/midi port and line in/out/mic, 26-pin non-standard header.  The first 16 pins of this header are for a gameport, and that is standard (pin 16 not connected).  The subsequent 10 pins are used for the audio jacks, but do not follow the standard for front panel audio designs.  Instead, 4 pins are used for the stereo line in, 4 pins for the stereo line out and 2 pins for the microphone input.  This header adapter appears to be identical :
ATX FORM : PS/2 mouse, Infrared Port, and 2 x USB ports, 18-pin non-standard, pin 14 is the key.  I do not have this particular connector, but it appears identical to the one sold here :
USB1 : USB1.1 x 2 ports, 8-pin semi-standard.  The standard ribbon cable has 10 pins, but pins 9 & 10 are no connected.  A typical adapter should work.  This board supports 4 USB ports between this header and the ATX FORM header.
CD-ROM : Although not labeled as such, there are what appears to be two CD-ROM headers.  One of them has a white snap-in box around the pins.  The white box should have the pins in GLGR format, while the no-box header should have the pins in the LGGR format.  
SPDF I/O : 8-pin header, pin 3 is key.  Provides coaxial spdif input and output and stereo line input with an adapter which I cannot find.
J7 : 2-pin header for CD-ROM SPDIF input.  This cannot be used if the external SPDIF bracket is being used.
J4 : 8-pin header, pin 8 is key.  Not documented in the manual

There is a header, J11, for the switches, LED and speaker.  This header is a little odd.  The 4-pin header for the speaker and the 3-pin header for the Power LED are connected horizontally, but the other LEDs and the Power/Suspend switch are connected vertically.  They all use 2-pins, so this will work.  The silkscreening on the motherboard does not make this crystal clear.  

Special Features :

The motherboard theoretically supports up to 768MB of RAM in its three DIMM slots.  However, not all of this RAM is cacheable.  2MB cache systems can cache up to 508MB, 1MB cache systems up to 254MB and 512KB boards only 127MB.  If for some reason you need more RAM than this, the performance of the extra RAM will suffer.  My system reports itself as having 512KB of RAM, so a single 128MB stick of PC-100 SDRAM will suffice.  PC-66 SDRAM is also supported.

Front side bus speeds supported are 60, 66, 75, 83, 90, 95 and 100MHz.  If the silkscreening is to be believed, this board supports multipliers from 1.5x to 5.5x.  This pretty much encompasses the world of the Socket 7 and Super Socket 7 CPUs, from a Pentium 90 (1.5 x 60) to a K6-2+ at 550MHz.  These CPUs do not have locked multipliers, so for example, my K6-2 350 can be unclocked to 90MHz.  The system can run CPUs with 3.3v to 2.0v core voltages.

The Southbridge chip supports UDMA/66 IDE.  This limits you to 28-bit LBA, so hard disk drives should be no more than 120GB.  Windows 9x also has similar limits.

The Trident Blade 3D integrated graphics cannot be disabled.  The integrated graphics will use 4MB or 8MB of system RAM for a frame buffer.  Discrete graphics cards using this chipset may use SGRAM instead of SDRAM.  There is Direct 3D support for 3D accelerated graphics as well as an OpenGL ICD for Quake-engine game support.  There is also hardware support for DVD decoding.  This graphics controller is AGP based, but does not take full advantage of the AGP specification.  When it came to 3D accelerated graphics, an nVIDIA TNT card smoked the Trident, as shown in the Anandtech article linked above.

The Trident adapter supports 640x480x32bit @ 160Hz, 800x600x32bit @ 160Hz, 1024x768x32bit @ 120Hz, 1280x1024x16bit @ 100Hz and 1600x1200x16bit.@ 85Hz.  DOS VGA and SVGA compatibility appears to be solid.

The CMI8738 chip appears to be a quite decent all-rounder.  It supports DirectSound/DirectSound 3D, EAX 1.0 and A3D 1.0 with updated drivers.  It provides Head-Related Transfer Functions for 2-speaker 3D positional audio.  It can record and output 24-bit SPDIF and will output 5.1 AC3 from DVDs.  The SOUND1 bracket supports using four speakers by repurposing the line-in connector.

The CMI8738 will emulate a Sound Blaster 16.  It possesses an excellent FM synthesis core.  I do not know if it was licensed from Yamaha, but it sounds very close to a Yamaha YMF-724 FM recording.  If you want to use an ISA card for DOS games and keep the PCI audio, you can move the I/O ports of the Sound Blaster, MPU-401 and/or FM Synthesis to ports that will not be used.  Interestingly, the VIA Southbridge also supports Sound Blaster Pro emulation, but this may be disabled on the motherboard or the chip itself to allow the CMI8738-based chip to be the only on-board audio system.

Expansion :

Meet your expansion options
A typical AT case will have seven or eight card slots.  The motherboard included five header brackets, sound, modem, LAN, VGA and Serial/Parallel.  The manufacturer offered the ATX and SPDIF brackets as optional purchases.  You may not be able use all the headers and both expansion slots, one must be sacrificed in a seven slot case.  Typically the modem is the best choice to go without, especially in today's world.

This board is on the edge of the AT to ATX transition.  Most Socket 7 and even Super Socket 7 boards used the AT form factor, while virtually all Pentium II boards used ATX.  The AT keyboard connector has purple plastic, which is a PC97 requirement.  With the ATX FORM bracket and an ATX power supply, this system becomes a fully ATX compliant computer with its various suspend and sleep functions.  Use of AT power supplies makes fewer of these energy saving options available.  I personally find sleep, suspend and ACPI functions a nuisance.  I prefer to turn a system on and turn a system off rather than worry about whether it will wake itself from sleep.

The board is very small for the period, not much larger than a mini-ITX board.  Even so, very little that an average user may need is left out.  A parallel port and one serial port is provided.  The one serial port is intended for a mouse, and instead of a second serial port you get a built-in (if crappy) modem.  Use the ATX FORM bracket and a PS/2 mouse and you can rescue the serial port for something else.  By adding basic 3D accelerated video and audio and a LAN, there is not a heck of a lot of need for expansion for most people.

The ISA slot is in line with the switch and panel headers, so using long cards with this system may be tricky. The PCI slot is in line with the secondary IDE port, so caution is advised with using a long PCI card.

If you want to use this board for 3D gaming, you may want to add a 3D accelerator.  A Voodoo, Voodoo 2 and PowerVR PCX1 or PCX2 are add-on boards without 2D capabilities.  They should be able to co-exist without difficulty with the on-board video.  Because Voodoo 2 SLI requires 2 PCI slots, the only way you will be able to use Voodoo 2 in SLI is by obtaining a Quantum3D Obsidian 2 X-16 or X-24, which combines two 8MB or 12MB Voodoo 2 boards onto one board.

Building the System :

The 486 motherboard had to get the boot
The K6-2 is recognized as a K6-2 at 350MHz.  You can set the speed, multiplier and core voltage in the BIOS.  For multipliers, you can select 2.0x through 6.0x in 0.5x increments.  For core voltage you can select 2.0-2.5v in 0.1v increments.  Front side bus speeds supported are 60, 66, 75, 83, 90, 95, 100, 105, 110, 115, 120, 124MHz.  Perhaps a Pentium or a Cyrix CPU would have different settings available.

There is no way this board will fit inside an modern ATX case.  The mounting holes for the motherboard are in different places in an ATX case and there are not enough common holes to give this AT board sufficient support.  Even if you remove the ATX I/O panel to allow you access to the AT keyboard connector, the expansion slot holes in an ATX case do not correspond to the expansion slots on the board.

I'll win no prizes for my cable routing
Getting all the ribbon cables to cooperate with the bracket placement was a bit of a challenge.  Even though no ribbon cable had more than 24 pins, ribbon cables are not the most flexible of connectors.  The floppy drive ribbon cable also connects in the area where the on-board peripherals connect.  I could not get the modem daughterboard, which connects directly to the motherboard without a ribbon cable, to fit in the AT case I have, so I just left it out.

Lots of ribbon cables
I originally could not get the board started.  I tried an AT power supply and an ATX power supply, but neither seemed to work.  I thought I had a dead board and was beginning to believe I had wasted my time.  I looked at the trace side of the board and it appeared that the ATX traces did not make any sense where the AT traces did.  The AT connector is not keyed, making inserting the connectors a bit tricky.  Once I realized that the ground wires (black) are always together and in the middle with the AT connectors, I was able to start the board.

The ATX connector still did not work.  The silkscreening for the switch/LED block was a little confusing.  I did not realize until I found and read the manual that the 2-pin connectors were to be inserted vertically, not horizontally.  I figured they were inserted vertically just as the other connectors, which is how they are inserted in my ASUS P3B-F.  Once I inserted the power switch connector vertically in the last pair of pins, the board fired right up.

So much for that AWE32 I wanted to put into the ISA slot
If you want to use an ISA card that extends over the switch/LED pins, you will have difficulties with the LED and power and reset switches due to the lack of clearance.

The system did not come with any RAM, so I grabbed a 64MB and later a 128MB stick of PC-100 SDRAM I had lying around, the system did not have any trouble with the RAM.

Installing the OS and Drivers :

I was only prepared to use a 540MB hard drive for experimentation.  I decided to use Windows 95 for a change, since it is smaller and faster with lower end hardware than Windows 98.  I have a Windows 95 OSR2.1 CD, which I found out was not bootable.  A truly fresh install will require a boot disk with a DOS CD driver and MSCDEX.  Installing the operating system was fast and easy.  Getting the drivers to work, well that was not easy.

A streamlined boot screen
The chief difficulty is finding the right drivers for this board.  Originally the board came with a drivers CD with drivers for the LAN, Video, Sound, Modem and IDE Bus Master.  Of course I did not have this CD, so I had to search for the drivers online.  I managed to find them, thanks mostly to Driver Guide.  Installing the Sound and Bus Master drivers was easy.

Only 512KB of External Cache, I want my money back!
Installing the LAN driver was not so easy.  Even though the chip used is a Davicom 9102F, there is more than one driver set for it.  The driver set that finally worked for me was not easy to find.  It has folders for win95, win95osr2 and win98.  Even with this driver, I still got a yellow exclamation mark in Driver Properties.  What finally worked was deleting the "Unsupported Device PCI LAN" after installing the driver. Once the driver was installed properly, the LAN was well-behaved.  Connecting to a Network Drive was not a problem.

The OS installed in the exposed system
Finding a good video driver for this motherboard has been a more difficult trial.  The Blade 3D does support Direct3D, so hardware accelerated games should be possible.  The first drivers I tried refused to give me anything but standard VGA capabilities.  The second set gave me full 2D capabilities at first, but not 3D capabilities.  The first set identifies the video chipset as a Trident Blade3D, but the second set identifies itself as a Trident CyberBlade i7 AGP.  It seems that like the LAN adapter, different drivers may be required depending on whether you have a standalone expansion card or an embedded solution.  Even Windows 98SE is unlikely to have video drivers for this adapter as implied here :

You have to install the card using the setup program, not the Add/Remove Hardware, in order to obtain full 3D capabilities.  I had a non-working Diamond Monster 3D in the system, and it seems that it was blocking the Trident setup program from working.  Unfortunately, the pickiness of the drivers means that they really cannot be upgraded.

Windows 95 stripped of all its nonsense
The graphics adapter was able to run 3dmark99, so it should be able to run Direct 3D and OpenGL games from the mid-to-late 90s reasonably well.  Glide-only and Glide-preferred games have another system to run on.


Bob Jimms said...

Wow! This is the exact board that my first DOS/Linux machine had!

The VIA chip had a really hardware bug, which with Windows, needed the exact driver version and DirectX to fix.

It was PC Chips—but from memory, sold under another brand. Something like, Lucky Tech.

Anonymous said...

I keep a similar super 7 board of the same era in good working order, mine is SiS530 based and I use it in a home lab.
PCChips was then a lousy company and it was common for them to install fake cache on these low end boards even if they reported up to 1MB in the BIOS text. In my case CPU-Z and MemTest both report zero 2nd level cache; performance and reputation of the K6s suffered as a result.
Despite this, it is surprising that such old equipment is still able to operate today. Clearly the amounts of power used and associated thermal stresses were not so high then as would later become.