Uncreative Labs

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

So I'm currently in the process of building a system out of a real POS Pentium motherboard I bought at a computer show for $15 in 1996. I just got it working today. It's pretty shite. It's not much better than a 486.

The specifications are as follows:

-TMC PAT45PVS Socket 4 motherboard

-Uses some awful OPTi chipset. I think it may be called "Viper", and it supports both 486 and Pentium CPUs. The 486 socket was not installed on the board, but the traces are present.

-Has 2 VLB slots. No PCI!!! I haven't really had much luck getting the VLB working yet.

-No integrated I/O, no integrated anything!!!

-Uses terminator packs in many places instead of jumpers. It also uses an oscillator to set the CPU frequency....just like old 386 and 486 boards.

-1993 AMI Pentium ISA BIOS identical to 1992 486DX BIOS

-Supports 512kb cache using standard 28-pin SRAMS. There are 19 sockets in total. Supposedly it supports writeback mode too!

-Four 72-pin SIMM slots that are super picky about memory specs. I tried 10 pairs of FPM non-parity RAM, and only got one set working. Supposedly this board likes parity memory, and can support up to 128mb of it.

Right now I have a Pentium Overdrive installed, which in theory allows me to run at up to 133MHz. Though so far I have only gotten it to run in 2x50MHz mode. When the FSB is set to 60 or 66, the board won't start. I have 512kb cache installed, but I imagine it's quite slow at only 15ns. I'll have to do some benchmarks later tonight. It should be interesting. My guess is that my AMD-486 133 beats the heat out of this thing.

Joined: 19 Apr 2006 Posts: 261 Location: WisConSin

I guess there was a 64bit VLB standard that looked a lot like 32bit, sadly it was never really supported

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

Oh no. It surely isn't 64-bit. 64-bit expansion buses didn't become mainstream until just last year. I think the reason the VLB isn't working is because it is operating way out of spec. I suspect that the VLB clock isn't being halved like it is supposed to.

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

So I somehow managed to get this board working at 66MHz with the CPU at 133MHz. I installed Wintune 2.0 and ran some unoptimized tests. I must say that considering the simplicity of the motherboard it is surprisingly fast!!!

The CPU certainly seems to do its job, and easily beats a first generation PCI P90 system in FPU and CPU performance. Memory is slower than late model Pentiums, but still comes in at a respectable 20MB/sec (respectable for early model Pentiums that is). Hard disk speed with 32-bit file access and 16-bit disk access managed 7mb/sec on a VLB IDE controller and 4500rpm Maxtor EIDE drive. Video drivers were not installed, so those results aren't worth mentioning. I will be using a Mach64 VLB with 2mb.

I suspect there is some room for a small increase in memory performance, and a substantial disk throughput upgrade.

I'll have to take some pictures. This computer is hilarious.

Joined: 03 Feb 2006 Posts: 488 Location: pennsylvanai

my first pentium was a pcchips pm8088.
it did have 50-75 bus speeds. the 66mhz chips mseemed to work well at 76,
the board had 4 sim and, i think one or two simm sockets the intel 430 chi[pset seems to be limited to 128megs of ram.
it has isa and pci slots and a socket for an unavailable regulator for the 233mmx chips.
I bought a bunch of similar boards in 2002 or 2003 for $1.00 each.

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

Wow Ryan, it looks like you were right about 64-bit VLB. It's also called "VLB 2.0". It seems that it came out in 1994, but nobody used it. Suppsedly they were able to implement 64-bit transfers without modifying the socket. I'm not sure how you would distinguish between 1.x and 2.0 just by looking at it. Not that it matters...I doubt any cards supported it. Though, supposely 2.0 is still better because it does a better job of handling signal noise and lets you run more cards at higher frequencies.

Anyway, I've been playing around with my weirdo VLB pentium, and here is what is in it at the moment:

-TMC PAT45PVS V2.0 w/512kb cache
-Intel Pentium Overdrive 5V133 (Socket 4)
-8MB non-parity FPM 70ns (128mb on order)
-ATI Graphics Ultro Pro VLB w/2MB (Mach64)
-QDI dual channel VLB IDE controller with I/O
-Sound Blaster AWE32 w/8.5mb wave table
-IBM 34GXP 20GB 7200RPM 2MB cache
-Panasonic 32X ATAPI CD-ROM

Joined: 19 Apr 2006 Posts: 261 Location: WisConSin

Yep, there was most definately a 64bit VLB implimentation that made it onto a couple boards (it was backward compat. with 32bit so it would be ambiguous) and there were also controllers for non-486 machines to use VLB CORRECTLY (as opposed to halfassed) Most manufacturers being cheap did not bother though. That is why I said VLB wasn't just a 486 thing, technically could be run on any machine even a 386sx with the proper support chips and it did indeed work on old Pentium machines like my Everex p60.

To be honest I wish VLB would have lived longer and been properly supported when it was #1. Most manufacturers used it as a cheap improvement over ISA when they didn't want to pay for PCI royalties.

Heck we might be using VLB cards of some sort today if they would have lingered a little longer. The 64bit implimentation was fairly impressive for the time.

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

Would you happen to have a list of motherboards and cards that supported the 64-bit standard?

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

I've been playing around a little bit with this screwball system, and I was able to tweak it a little more. After installing the drivers and configuring the BIOS, I was able to get the VLB video to do 9.5MP/sec at 800x600 16-bit. I guess that's not too bad for a mach64, but I'd really need to see wintune results for a PCI mach64 to know for certain.

The biggest problem I have been having is improving HD bandwidth. I was originally using an old 4200RPM Maxtor CrystalMax drive, but I swapped it for a nice 7200RPM IBM 34GXP. It didn't change a damn thing in terms of performance. I am now certain that my VLB IDE controller needs to be replaced. I am hoping to upgrade to a PIO mode4 controller. Wide SCSI would be nice, but I don't think there is such a card for VLB.

I have also been unable to increase memory performance. I think the cache on the motherboard is just too old. However, I have a plan. If I can install a CPU that has integrated L2 cache then I could avoid this bottle neck. The only CPU I know that could possibly work is the AMD K6 III or one of its derivatives.
While the Socket4 does not support lower voltages or split rail CPUs, there is an adapter from Powerleap that can do this. It is called the PL-54C/MMX. Officially it can go up to a 300MHz K6. In my opinion there isn't a lot of difference between a K6 and a K6 III except for 3DNow! and L2 cache. So, I am hoping that perhaps my little scheme can work....though I've heard sometimes the cache can really cause BIOSes problems. I may have located an adapter in Germany, so hopefully I can have this all done in a matter of weeks.

If it does indeed work, I will have on my hands the strangest little beast of an x86 system known to man.

Joined: 19 Apr 2006 Posts: 261 Location: WisConSin

Hmm, I never bothered to write them down, sort of like I never kept track of software that would run under windows 1.0 but here is something used to test the cards which nobody seems to have

http://198.66.233.108/vlb.htm

Best I can remember there might be a few adaptec scsi variations in 64bit vlb and a couple server type motherboards and some cheap socket 5 tawanese/korean boards that used it circa 94. sadly this is something I didn't think much of a footnote at the time it would have mattered.

Also the Cirrus Logic 5428 VLB is a VLB 2.0 card if that is the same as 64bit Vlb

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

So I ordered 128MB of 72-pin 60ns parity memory for this motherboard. It really doesn't seem to like it very much. When I enable "parity", the system won't boot reporting a "parity" error. When parity is disabled, it boots, but becomes unreliable and eventually crashes. There is nothing wrong with this memory, as I have tested it in another system.
I have all sorts of different SIMMs lying around, but so far I have come to the conclusion that this board likes FPM single sided SIMMs with or without parity. However, my only single sided SIMMs are either 4MB modules, or have EDO. This board doesn't like EDO either. So, I guess the best thing to do would be to order a set of 16MB single sided FPM SIMMs with Parity.

What a pain in the arse.

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

The more I play with this motherboard, and the more I read about the OPTi 82c571/82c572 chipset the weirder this board becomes. According to some information on the newsgroups, this OPTi chipset is the only one of its kind to allow you to install single SIMMs rather than in pairs without taking a performance hit. Somehow the chipset can "poll the memory twice" to keep an apparent 64-bit datapath. That's really about as many technical details I've managed to find so far.
Though, the board is really screwy. According to TH99, motherboards that use this chipset can use double sided SIMMs. Though, I personally believe this to be false. The system will boot to a DOS prompt, but often with a HIMEM.SYS error, and windows will crash the system. Also, the memory bank assignments for this board are really random and non-sensicle. I would really like to meet the doofus who designed the chipset and attempt to figure out what kind of crack he was smoking. It seems that you can only have one pair of chips installed of the same density at a time. Installing more than 2 SIMMs of the same type makes the system not boot. Pretty weird, eh? For those interested the memory bank configuration tables can be seen here:

http://www.thegreenhouse.us/th99/m/S-T/32694.php

What this TH99 doesn't show is that the board also has an optional 486 CPU socket and CPU selection jumer. It's a pretty screwy design. An earlier revision of this board also has 4 30-pin SIMM slots. That's right, 30-pin SIMMs on a Pentium! I think perhaps only three companies bothered to make boards similar to this one, so I expect it to be quite rare. It's certainly a nice piece of history to own, and a lot of fun to see working.

UPDATE:

I ordered a set of high quality single sized 64MB SIMMs to test out my theory. Hopefully it'll work...or I'll be angry.

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

Ugh. So my latest upgrade to this system has turned out to be another dud so far. I wanted to improve hard disk performance, so I went all out and bought a PIO mode 4 multi-I/O. I ended up with a DTC 2278E, because the other one was a Promise...and I hate those guys. Well, as it turns out neither the DTC nor the Promise are particularly good cards. I just can't seem to get the bloody thing to work properly in this VLB pentium.

I think that perhaps it doesn't like the disk overlay I am using. I am currently using the IBM Ontrack Disk Manager 9. I have had nothing but problems with this peice of software. For starters, during the installation when it requests a bootable disk it can accept a Windows 98 bootdisk, but rejects my PCDOS 2000 bootdisk everytime. You would think that IBM would have at least checked out compatibility with their own DOS version before slapping their logo on DM9. I have though to try DM10, but I haven't managed to find an IBM or generic version yet.
When I do finally managed to get DOS installed on the system, running the DTC 32-bit DOS driver seems to mess up DOS making my disk drive inaccessable. From what I read about these DTC controllers they are nothing but trouble. Is there a trick I'm missing?

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

I'm slightly annoyed. I think I ordered the DTC 2278E for nothing. As it turns out, the QDI QD6580 controller that I was using previously was already capable of doing PIO mode 4! I should have doubled checked that before I considered a new controller. I am pretty surprised to find that software is still readily availlable for the QDI. I'm going to give it a shot and see what happens. I have a feeling that I won't be able to do much about my windows performance. I think that the 32-bit disk driver that came with Ontrack was already enabling a higher PIO mode. Though, perhaps the QDI driver will be better. I'll have to test it later tonight to find out.

I also got my compaq 64-mb FPM 72pin RAMS in the mail today. My motherboard will only boot with one of them installed. It seems to be stable. I suppose I'll find out for sure after I reinstall Windows....I kind of facked up my partition when trying to install the DTC DOS drivers.

UPDATE:

I reinstalled the QDI controller and 64mb FPM RAM. The disk performance improved dramatically...though I am not quite sure the reason. I suspect it might be because of the extra RAM. I am now getting 11MB/sec average disk speed with only 32-bit file access enabled. I have yet to enable the 32-bit disk access. I think this disk is capable of around 25MB/sec sustained using UDMA33. The best I can do with PIO4 is 16MB/sec....though, I am not sure if the drive will be capable of reaching this kind of bandwidth without DMA. I guess all I can do is test and see.

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

I have been running with the single 64MB FPM SIMM for about a week now. The system seems to be relatively stable, except that it resets when it becomes too hot. I think the airflow in my case is poor, because I am currently running the AT board in an ATX chassis without a backplate. I bought an adapter to let me use ATX power supplies on AT boards. I didn't want to blow out one of my old AT power supplies, as they seem unreliable in their old age. I managed to get disk throughput to 12MB/sec. I think it won't be going any higher on this system.

In the mail, I have a proper backplate coming, an SB16 and a 1542CP SCSI controller. I am now running a 4X SCSI CD-ROM on a TMC850 8-bit card, and I have removed the AWE32 as it is not a reliable card. I was hoping to get an SB16 with a SCSI interface, but I found that I could get the two separately for the same price. The SB16 has an ASP chip...whatever that is.

*UPDATE*

It seems that the cause of the instability was the memory parity option in the BIOS. Though the memory is indeed parity, the board just doens't seem to handle parity properly. They system configuration is now as follows:

-ATX Desktop chassis with 340W Antec ATX PS
-TMC PAT45PVS V2.0 w/512kb cache
-Intel Pentium Overdrive 5V133 @ 133MHz (Socket 4)
-64MB parity FPM 60ns (running non-parity))
-ATI Graphics Ultro Pro VLB w/2MB (Mach64)
-QDI dual channel PIO4 VLB IDE controller with I/O
-Sound Blaster AWE64 Gold
-Adaptec 1542CP DMA SCSI controller
-IBM 34GXP 20GB 7200RPM 2MB cache
-Plextor 32CSI Caddy SCSI CD-ROM

Future Plans:

-Powerleap Adapter w/K6-III
-2MB module for Mach64
-100mbps Ethernet
-TV Card