VOGONS


First post, by 0xCats

User metadata
Rank Newbie
Rank
Newbie

So this will be a relatively straight forward post.
I'm searching for a bios for a rather rare board, and I'm hoping one of you here maybe also has this board and is able to dump the contents of the BIOS chip.

https://theretroweb.com/hardware/motherboards … ion-mb-p5n-dpv3

The board is 430NX chipset based and supports two Socket 5 Pentium processors.
As far as we (Ultimate Hardware 2019) know it is made by Sprint Manufacturing Corporation sometime in 1994.
http://www.win3x.org/uh19/motherboard/show/4725
I have recently restored this board and gotten it running again (sort of) the original bios is corrupted and through a lot of bios hacking I have managed to get the BIOS of the Chaintech 5IDL0 working on this board. This BIOS however is also partly not suited to this board, it does not setup the internal or external caches, among many other things, it also cannot read any of the GPIO board jumper pins and as a result of various quirks runs absolutely terrifyingly slow. In addition PCI/IDE setup is a mess and it will not boot those either. Only thing I have managed to boot is PXE network devices. But there also things like iSCSI fail too due to complex reasons with how award bios handles disk boot and iscsi.

The BIOS version 0208 of the Asus P4P54NP4D also works but PCI and LBX setup fails, and ISA video cards are also not initialized, yet I can tell that the board is actually running since it will attempted to boot disks and shows the relevant post codes.
Originally the Sprint board came with an AMI Core bios dated 10/10/94.
Here is a link to the bad one:
https://theretroweb.com/uploads/mobos/sprint- … 01bx-tdip32.bin

Here is a picture of this rare board and it actually running to some extent.

I hope someone can help, if they have a such a board or even a dual socket 5 that looks similar enough if you could read the bios contents out and send them to me.
a6103fda1bcd613.94711492

slowspeed ahead.jpeg
Filename
slowspeed ahead.jpeg
File size
502.57 KiB
Views
667 views
File license
CC-BY-4.0
chaintech bios.jpg
Filename
chaintech bios.jpg
File size
150.23 KiB
Views
667 views
File license
CC-BY-4.0

Thank you
The Cyber Cat

Last edited by 0xCats on 2021-08-09, 16:59. Edited 1 time in total.

Reply 1 of 11, by Horun

User metadata
Rank l33t
Rank
l33t

Interesting ! It looks like a OEM Micron motherboard. Cannot find anything...

Hate posting a reply and then have to edit it because it made no sense 😁 First computer was an IBM 3270 workstation with CGA monitor.

Reply 3 of 11, by 0xCats

User metadata
Rank Newbie
Rank
Newbie

As a progress update I have now managed to get cache and various other elements functional.
Also disk controller enumeration etc is working.
I'm now using an AOpen DP5 motherboards BIOS which has been adapted to run on this board.
Some chipset logic configuration is still required however with the documentation at hand it's going somewhat ok.
The next trick will be to get the board to consistently set the masked [7:6] bit chipset registers init values for 52h.
Currently I am using a reset quirk of the chipset to set those bits (essentially bug programming).
Essentially when cold posting the board the cache register is programmed with incorrect cache configuration parameters for the AOpen platform and as a result cache is not configured correctly.
So what I do is post, then I remove the cache config jumper, (bit 6 high) and then reset and I get the inverted logic sampled and bits 7:6 are configured as 10.

Screenshot_20210809_184021.jpeg
Filename
Screenshot_20210809_184021.jpeg
File size
92.05 KiB
Views
555 views
File license
Public domain

As you can see this has given good results 😁

Screenshot_20210809_183741.jpeg
Filename
Screenshot_20210809_183741.jpeg
File size
153.56 KiB
Views
555 views
File license
Public domain
Last edited by 0xCats on 2021-12-13, 15:39. Edited 1 time in total.

Reply 5 of 11, by 0xCats

User metadata
Rank Newbie
Rank
Newbie

No pictures this time, but finally I can report I have managed to restore full speed and full cache functionality using a custom crafted AwardBIOS 4.50PG adapted from the AOpen DP5. Itself yet another undocumented unseen dual pentium motherboard.

Will post more once I have the time and a proper PCI IDE card haha. PXEBooting a P586 architecture via iSCSI over gigabit is a bit of an alice in wonderland adventure worthy of it's own thread. 🤣
Next up will also be modding the board with a stronger VRM and getting higher multipliers, plus syncing the APIC for fast dual processors.

And just for fun let me give the laundry list of faults this board had:

1. Both 3.3V Linear LDO voltage regulators for CPU Vcore had failed (likely cooked to death), the heatsinks on them are pathetic.
2. 430NX had loose pins on two sides (bad terrible solder)
3. S82379 SIO had loose pins on bottom edge.
4. Keyboard rom socket had a spider nest under it and corroded pins.
5. One of the 433 LBX chips had loose pins all around. (yet again crap solder)
6. two address lines of BIOS chip were cut
7. 1 CPU address pin on primary socket had a bad corroded joint.
8. 2 ISA address and 2 ISA data lines on back were scratched through
9. AT header power pins had cold joints (reflowed)
10. Original BIOS is corrupted and will more often not post than post, large part of chipset setup logic is FUBAR
11. degraded SMD capacitor on Address Line 30 (used for strapping cache config after PWR-OK edge) was causing cache config flipping between 256/512K.

Last edited by 0xCats on 2021-12-24, 14:47. Edited 1 time in total.

Reply 6 of 11, by 0xCats

User metadata
Rank Newbie
Rank
Newbie

So as an update. It turned out that the cache I originally though working was a fluke. A bunch of chips and the buffers for them were degraded.

A while ago I actually went about testing the cacheless wonder. Well here are some more images.
When did you ever see Vintage Edition CPU-z with the Socket Dropdown active? 😜

image(1).jpg
Filename
image(1).jpg
File size
66 KiB
Views
349 views
File comment
Vintage Edition with more SOCATS???
File license
Public domain
image.jpg
Filename
image.jpg
File size
65.11 KiB
Views
349 views
File comment
COARS!
File license
Public domain

Windows 2000 installed (without L2 cache active) -oh horror
The memory subsystem performance was astonishingly dreadful.
After some testing done using a custom build of the chipsandcheese memory latency tool we came to suspect the reason for this is that the OS partly uses the L2 cache to speed up page table lookups.
Without this, memory address generation and lookups quickly become absolutely dreadful beyond about the 256KB size.
Here is a cache enabled 430TX compared using a somewhat different processor.

image(6).png
Filename
image(6).png
File size
54.08 KiB
Views
349 views
File comment
Memory latency plot vs 430TX PMMX 233
File license
Public domain

The memory latency here shockingly is worse than the worst Nvidia GPU OpenCL could still be run on in 2021.
For a GPU to outperform a CPU's memory latency in any way is pretty hefty.

unknown.png
Filename
unknown.png
File size
20.22 KiB
Views
349 views
File comment
430NX vs 9500 GT
File license
Public domain

Anyways, I've finally received some new cache chips for this thing today, finding 3.3V Supply Low Voltage TTL rated SRAM is a PAIN.

Last edited by 0xCats on 2021-12-13, 15:38. Edited 1 time in total.

Reply 7 of 11, by 0xCats

User metadata
Rank Newbie
Rank
Newbie

Next modifications that are about due:

- Already modified BIOS for very low AT and PCI clk ratios
- Replace all the slow 74F244/245/373 series logic chips with ultra fast 74ABT series. (MOAR MHZ!!!)
- Replace the crystal oscillator can driven FSB with a custom PLL PCB and inject any desired FSB frequency.
- Wire all the multiplier options from the socket to an addon switchblock.
- Split the 3.3V Voltage plane at the sockets and install a custom VRM PCB ontop of the linear Vreg 3 pin footprint via headers. (Pentium MMX SMP!)
- Design a new cache interface PCB that goes ontop of the existing async SRAM sockets but allows the use of a single 512K Pipeline Burst SRAM chip. (512K is the max 430NX supports)
- Attempt 100MHZ FSB speedrun

Reply 8 of 11, by PC@LIVE

User metadata
Rank Member
Rank
Member

Hello and congratulations on the work and the motherboard.
I have seen the excellent work done, to adapt the BIOS, and to improve the performance, I guess the voltage regulators are linear, in that case going up the frequency would have more heat to dissipate, so I would recommend 166MHz. Perhaps we can go further, but to produce less heat it would be necessary to use cpu for laptops with reduced voltage, this should guarantee operation at lower temperatures and therefore greater reliability, these are ideas that I made myself (maybe am I wrong?). In several VXs with linear regulators I have had problems with 200 or 233MHz CPUs, perhaps changing the heatsink on the mosfets (with a bigger one) and adding a fan would work better.
On a VX and a TX with switching regulators, on the other hand I don't have these problems, and I have AMD K6-2 CPUs at 360 and 400 MHz, the limit is due to the FSB that reaches 83MHz at the maximum, but since you want to reach 100, maybe it would be possible that it works more stable if you used a 90MHz FSB, especially if you can make (divider 1/3) that the PCI goes to 30Mhz, instead of 45MHz with divider 1/2.
With fsb 90 a 233MMX would go at 315Mhz, and it would be a very good frequency for that CPU, then having two would be even better, but reaching and exceeding 300 is possible, but it is not guaranteed that it will be stable and reliable over time.

I will gladly follow the developments of your work, I am very interested in this thing of the FSB 100, in my opinion those who have tried have failed because the divider 1/2 makes the PCI work at 50MHz.

AMD 286-16 287-10 4MB HD 45MB VGA 256KB
AMD 386DX-40 Intel 387 8MB HD 81MB VGA 256KB
Cyrix 486DLC-40 IIT387-40 8MB VGA 512KB
AMD 5X86-133 16MB VGA VLB CL5428 2MB and many others
AMD K62+ 550 SOYO 5EMA+ and many others
AST Pentium Pro 200 MHz L2 256KB

Reply 9 of 11, by 0xCats

User metadata
Rank Newbie
Rank
Newbie

After much toil and suffering I have a new update.
L2 cache is working again FINALLY!!

cpu-z cache functional.png
Filename
cpu-z cache functional.png
File size
1.1 MiB
Views
269 views
File license
Public domain

Here is proof of that alongside a quick test of aida64 version 2.50.

cpu-z cache test no optimizations.png
Filename
cpu-z cache test no optimizations.png
File size
1.19 MiB
Views
269 views
File license
Public domain

Now that this is all working I can actually start optimization work alongside getting this bios code tweaked to the max potential.

Last edited by 0xCats on 2021-12-13, 15:38. Edited 1 time in total.

Reply 10 of 11, by 0xCats

User metadata
Rank Newbie
Rank
Newbie

Time for the real stuff!

kernel 5.15 booting.png
Filename
kernel 5.15 booting.png
File size
1003.48 KiB
Views
234 views
File license
Public domain

I've gotten the latest Linux Kernel booted and running on this thing.

cpuinfo + kernel.png
Filename
cpuinfo + kernel.png
File size
1.08 MiB
Views
234 views
File license
Public domain

So how much storage does your pentium 1 have?

Ah yeah a casual 1TB disk, nothing much...

disk size.jpeg
Filename
disk size.jpeg
File size
175.75 KiB
Views
234 views
File license
Public domain