watz wrote on 2022-08-30, 17:06:

Hi! […]
Show full quote

Hi!

I recently got a good deal on a 486 board from a Compaq 7100 series Desktop PC. It is based on the UMC8881/8886 chipset and comes with decent sound and crappy VGA on board. It is this model:

https://theretroweb.com/motherboards/s/trigem-pl4600c

The board came with an AMD X5-133 installed. I had a board with the same chipset and the same CPU back in '96. So I want to recreate my old setup. I was able to get it running @160MHz again after hacking the VRM to 3.6V. 15ns cache is also installed and working.

However performance is not quite as good as I remember. I only get ~13fps in Quake timedemo demo1 where I used to get at least ~16fps back in the day. The crappy GD5430 VGA is certainly part of the problem, but I don't have the matching riser card yet to install an alternative.

The other thing I suspect are pretty bad memory/cache timings. The Phoenix Bios on this machine has no options at all. Once I have the riser card I'll try Award bios images for other boards with the same chipset. Hopefully I can mod those to include the VGA bios. I have little hope for the on board audio though. Until then, I thought I'd try my luck here. Maybe you guys have some ideas.

- did anyone here have any luck with extracting/hacking an early Phoenix Bios? I tried various cbrom versions without luck.
- are there maybe any chipset specific tools to set mem/cache timings in nvram from DOS?

I've attached a bios dump in case anyone wants to take a look.

Afaik, editing early phoenix is diving into the unknown. While AMI and Award do have utilities like AMIBCP and Modbin/Cbrom that generally let you modify the bios, i don't know of any such utility for 4.0x phoenix. Maybe it does exist but i don't know about it or nobody has a copy of it anymore.
Using Cbrom won't work as that was always meant for award, and this bios is long before phoenix acquired award. As for chipset specific tools, those may exist but i'm not much of an expert in that type of stuff so i wouldn't know.

watz wrote on 2022-08-30, 17:06:

Hi! […]
Show full quote

Hi!

I recently got a good deal on a 486 board from a Compaq 7100 series Desktop PC. It is based on the UMC8881/8886 chipset and comes with decent sound and crappy VGA on board. It is this model:

https://theretroweb.com/motherboards/s/trigem-pl4600c

The board came with an AMD X5-133 installed. I had a board with the same chipset and the same CPU back in '96. So I want to recreate my old setup. I was able to get it running @160MHz again after hacking the VRM to 3.6V. 15ns cache is also installed and working.

However performance is not quite as good as I remember. I only get ~13fps in Quake timedemo demo1 where I used to get at least ~16fps back in the day. The crappy GD5430 VGA is certainly part of the problem, but I don't have the matching riser card yet to install an alternative.

The other thing I suspect are pretty bad memory/cache timings. The Phoenix Bios on this machine has no options at all. Once I have the riser card I'll try Award bios images for other boards with the same chipset. Hopefully I can mod those to include the VGA bios. I have little hope for the on board audio though. Until then, I thought I'd try my luck here. Maybe you guys have some ideas.

- did anyone here have any luck with extracting/hacking an early Phoenix Bios? I tried various cbrom versions without luck.
- are there maybe any chipset specific tools to set mem/cache timings in nvram from DOS?

I've attached a bios dump in case anyone wants to take a look.

I'm not sure anything is wrong really with the CL-GD5430. That's an on-par card for a PCI 486 PC. What are you planning on replacing it with?

I've looked at PhoenixBIOS 4.x before. Portions of it are compressed. Compressed blocks start with a header containing the magic number 42 43 D6 F1 followed by two unknown bytes, a flag byte, a header length byte, a destination segment word, an uncompressed length dword, and a compressed length dword. What follows is compressed data. Yours uses LZARI compression from the Haruhiko Okumura set of sample LZ compressors published in 1989. Mine was a bit older and used LZSS.

The setup program appears to be in the second compressed block. I see signs of a cache speed 3-2-2-2/3-1-1-1/2-2-2-2/2-1-1-1, and ES1688 enabled/IRQ/DMA/I/O. Do you not see these settings in one of the tabs of setup? I don't think PhoenixBIOSes tend to have a ton of hidden options like AMI and Award do. You could also try explicitly restoring defaults which can change hidden settings.

Have you tried CACHECHK to compare with other users of UM8881 boards?

The 8881/8886 have no public datasheets; the only source of information about them is the BIOSes and drivers. So you won't find much in the way of tweaking utilities. The 8886AF has some features to speed up IDE transfers that are not standardized and only activated when you install the UMC drivers.

What about a mr bios?

Can you take a good picture of your board ? The one at theretroweb is not a 586 board so no ESS chip.....

Hey so much feedback 😉 It turned out the RAM and L2 timings are accessible in device specific config registers of the UM8881 PCI device. I wrote a little DOS prog today to read them. RAM was set to 1 WS read and 3 (!) WS write. L2 was set to 3-2-2-2 as expected. I fiddled with writing to these registers to turn off L2 for a start, but I had no success. I'm pretty sure it is possible to control timings this way but I have too little experience. I found a bare file "UM881.CFG" somewhere in the net that had the necessary register info. I only figured out later that it belongs to the free "ctchip34" tool distro, which probably can do these timing changes already. I have to try that.

But a couple hours ago the riser card arrived anyway. I flashed the Octek HIPPO 15 096017 bios. As expected, onboard video and sound stopped working, but otherwise this bios runs beautifully. Timedemo is back at 16.1 (from 12.9) and 3DBench sits at >90 (awful 52 before). Here is some Speedsys for you:

Before:

After:

I have no idea why the GD5430 is so abysmal on this board. I think I have a DX2-80 VLB setup with the same chip where it has more than twice the transfer rate.

The ES1688F maybe runs with its original config tool. I don't know why unisound doesn't detect it. I thought I had other cards with the same chip that unisound would detect.

Now I want to try to add the GD5430 option ROM into the bios. And I can try to add additional cache sockets for 256k of L2. There's a bunch of unpopulated jumpers that are obviously related to the L2 cache.

explorerdotexe wrote on 2022-08-30, 20:46:

Using Cbrom won't work as that was always meant for award, and this bios is long before phoenix acquired award. As for chipset specific tools, those may exist but i'm not much of an expert in that type of stuff so i wouldn't know.

Thats some useful information. It explains at least why cbrom is useless in this case.

jakethompson1 wrote on 2022-08-30, 23:19:

I'm not sure anything is wrong really with the CL-GD5430. That's an on-par card for a PCI 486 PC. What are you planning on repla […]
Show full quote

I'm not sure anything is wrong really with the CL-GD5430. That's an on-par card for a PCI 486 PC. What are you planning on replacing it with?

I've looked at PhoenixBIOS 4.x before. Portions of it are compressed. Compressed blocks start with a header containing the magic number 42 43 D6 F1 followed by two unknown bytes, a flag byte, a header length byte, a destination segment word, an uncompressed length dword, and a compressed length dword. What follows is compressed data. Yours uses LZARI compression from the Haruhiko Okumura set of sample LZ compressors published in 1989. Mine was a bit older and used LZSS.

The setup program appears to be in the second compressed block. I see signs of a cache speed 3-2-2-2/3-1-1-1/2-2-2-2/2-1-1-1, and ES1688 enabled/IRQ/DMA/I/O. Do you not see these settings in one of the tabs of setup? I don't think PhoenixBIOSes tend to have a ton of hidden options like AMI and Award do. You could also try explicitly restoring defaults which can change hidden settings.

Have you tried CACHECHK to compare with other users of UM8881 boards?

The 8881/8886 have no public datasheets; the only source of information about them is the BIOSes and drivers. So you won't find much in the way of tweaking utilities. The 8886AF has some features to speed up IDE transfers that are not standardized and only activated when you install the UMC drivers.

I had an ATI Mach64 back then, specifically for the YUV video acceleration. Thats what I'm planning to put in given I can find the card in my pile. I still have the two VideoCD discs I bought in '95 on an Amiga fair in Cologne to view on this setup (Total Recall and Star Trek 2). I remember I ran some software MPEG decoder on Windows 3.1 before I switched to 95.

Thanks for the info on the Bios structure. Perhaps I can extract the original 5430 option rom with this information.

Sphere478 wrote on 2022-08-30, 23:23:

What about a mr bios?

Is there one for this chipset and IO controller?

I love Mr. Bios. My first repaired 486 board had some extreme battery damage and no bios ROM. I found a Mr. Bios that ran perfectly. It gave me IDE transfers >5MB/s out of the box.

Here is a pic. Take a close look at the AT bus controller chip where the battery used to be. That board is running fine with a DX2-66@80Mhz and perfectly stable. A miracle 😉

Horun wrote on 2022-08-31, 00:35:

Can you take a good picture of your board ? The one at theretroweb is not a 586 board so no ESS chip.....

You can't see ES1688F chip in that pic. It is hidden by the power cable.

I really would like to upload pics and the original bios file there. No idea how contributing works there though.

How about a MR-BIOS ROM file repository?

Ask here 😀

Sphere478 wrote on 2022-09-01, 01:30:

How about a MR-BIOS ROM file repository?

Ask here 😀

Worth a try, thanks! I don't have much hope though. There seems to be much 386, early 486 and early Pentium stuff there, but not much in terms of late 486.

jakethompson1 wrote on 2022-08-30, 23:19:

The setup program appears to be in the second compressed block. I see signs of a cache speed 3-2-2-2/3-1-1-1/2-2-2-2/2-1-1-1, and ES1688 enabled/IRQ/DMA/I/O. Do you not see these settings in one of the tabs of setup? I don't think PhoenixBIOSes tend to have a ton of hidden options like AMI and Award do. You could also try explicitly restoring defaults which can change hidden settings.

btw, the sound settings are visible in the original Phoenix bios. But the cache timings are not. They have to be hidden somehow.

Do you have full 2 MB VGA memory ? It is quite possible that with 1 MB memory bus is halved to 32bit.
According to Alpine family tech. ref., all chips (5430, 5434, 5436, 5440) supports 64 bit memory bus.
What 5430 lacks is 64-bit BitBLT (bit block transfer) engine providing only 32.

Edit.
Well, above statement may be not true as a while ago I read in mentioned tech. ref. that:
The CL-GD543X1'4X is designed with the following performance-enhancing features:
• 64-bit display memory data bus for faster access to display memory (CL-GD5430/,40 has an effective 32-bit display memory data bus)

So 5430 is crippled by design 🙁

BTW. Technical reference guide is available to download from: http://www.vgamuseum.info/index.php/cpu/item/ … logic-cl-gd5430

jakethompson1 wrote on 2022-08-30, 23:19:

I've looked at PhoenixBIOS 4.x before. Portions of it are compressed. Compressed blocks start with a header containing the magic number 42 43 D6 F1 followed by two unknown bytes, a flag byte, a header length byte, a destination segment word, an uncompressed length dword, and a compressed length dword. What follows is compressed data. Yours uses LZARI compression from the Haruhiko Okumura set of sample LZ compressors published in 1989. Mine was a bit older and used LZSS.

The setup program appears to be in the second compressed block. I see signs of a cache speed 3-2-2-2/3-1-1-1/2-2-2-2/2-1-1-1, and ES1688 enabled/IRQ/DMA/I/O. Do you not see these settings in one of the tabs of setup? I don't think PhoenixBIOSes tend to have a ton of hidden options like AMI and Award do. You could also try explicitly restoring defaults which can change hidden settings.

I wrote a little prog to extract those three blobs. Works as you said. How the hell did you figure out the compression algorithm used?

The first block has the CL-5430 VGA option rom., albeit with a strange header. Now if I manage to add that VGA ROM to the Award bios, will it still auto detect an external VGA adapter? The old Phoenix bios supported external VGA cards. It must have had some code that copied the internal VGA rom to 0xC000 only in the case there is no external adapter. I would expect that Award also supports this.

zyga64 wrote on 2022-09-01, 08:36:

Do you have full 2 MB VGA memory ? It is quite possible that with 1 MB memory bus is halved to 32bit. According to Alpine family […]
Show full quote

Do you have full 2 MB VGA memory ? It is quite possible that with 1 MB memory bus is halved to 32bit.
According to Alpine family tech. ref., all chips (5430, 5434, 5436, 5440) supports 64 bit memory bus.
What 5430 lacks is 64-bit BitBLT (bit block transfer) engine providing only 32.

Edit.
Well, above statement may be not true as a while ago I read in mentioned tech. ref. that:
The CL-GD543X1'4X is designed with the following performance-enhancing features:
• 64-bit display memory data bus for faster access to display memory (CL-GD5430/,40 has an effective 32-bit display memory data bus)

So 5430 is crippled by design 🙁

BTW. Technical reference guide is available to download from: [url]http://www.vgamuseum.info/index.php/cpu/item/ … logic-cl-gd5430[/u]

Interesting info, thanks! Maybe that explains why my VLB GD5428 card is more than 2x faster. Only 1MB is installed. Looks like Compaq really cheaped out on this board big time (like no L2 installed by default).