VOGONS


First post, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie

Hi all!

I've aquired a QDI V4S471/G motherboard recently and resurrected it into life. The BIOS chip was toast as it had been on the board backwards.
Though, it's a curious case - i can read the chip and burn the rom on new chip, but the board doesn't boot up. I also tried other bioses from the internet for SIS 471 based boards. In the end, really, it has AMI and Award bioses and only Award boots up...
So one thing i would love to have is an original BIOS rom.

I've used so far other SIS 471 chipset based BIOS roms and they have worked beautifully so far. (And a bit of foreshadowning: it works well with 256kB of L2 cache)

So the next step was to upgrade the board from 256kB of L2 cache to 1024kB of cache. I have bunch of 1024kb chips so should work great. The board detects it, but when it comes to booting from hdd and i expect a message of "Starting MS-DOS...", it locks up. When i disable "External Cache", it boots up properly.

Based on documentation, the TAG chip should be 512kb instead of 1024kb. So i tried to bring the last address line of that chip to VCC through 10k resistor, but that didn't make the board work. I've ordered some 512kB chips, too, just to be sure, but it takes time to arrive.

Right now, I'm in the middle of messing around with the BIOS ROM on register 72, that contains bits of how some of the caching lines should work.

I'm doing all my testing with the slowest settings in BIOS and just 25MHz bus speed just to eliminate any speed issues. Later, when stuff works, i can tune it up.

But do you know anything that could help? What could i check?

Last edited by GigAHerZ on 2021-11-10, 20:33. Edited 1 time in total.

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 2 of 25, by Disruptor

User metadata
Rank Oldbie
Rank
Oldbie

Your board also supports 512 kB L2 cache in single bank configuration.
Please try 4 modules in U31-U34 according to your manual and use a smaller tag chip (the one you've used in the 256 K configuration).

Depending whether this works or not, exchange the 4 modules and draw a conclusion.
Please note that a single bank cache configuration may not work with the tightest cache timings in some motherboards.

Reply 3 of 25, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie

I must have been too sleepy yesterday or something, but i got the board working EXACTLY as i wanted. I must have mixed up the jumpers somehow. The cache is working now! All 1024kB and i'm using 1024kb tag sram.

But i have quite a bit of nice information to share.

First, the AMI bios. I tried it and it works.
It's the AMI winbios, which i absolutely hate - it's slow, it's missing features, etc.
At the same time, the latest Award bios for SIS 471 chipset i have is from 1995-11-28, which i use on my other SIS 471 board. I tried different values for the register 72 bits 1 and 2. For one value, it recognized my 32MB memory sticks consistently as 16MB ones. 😁 With the other option - it works! So now i have an Award bios from the end of year 1995, that has all the options visible and all the caching mechanisms are working and enabled, too. I've attached the Award bios to this post.

I also had a problem to configure my CPU properly. I can configure it as AMD-DX4 or as Enhanced AMD-DX4. I have the Enhanced CPU (really meaning that the L1 cache can work in WB mode) but setting up the jumpers for it causes the same problems described in original post - if L2 cache is enabled, the board locks up.
So i started to figure out every jumper on board. Manual just tells you, how to set the jumpers, but not what they do. I wanted to set up the board in AMD-DX4 mode with L1 cache working in WB mode. And i found the jumper! It's J11 - Open(WT); Closed(WB).

So i think i have the best i could wish - A non-PCI 486 board with 1024kB L2 cache and a bios with every setting that could exist, really existing. Now off to set it up properly and do some comparisons with my previous 486 board - Soyo 025P2. (Had quite a saga with that as well: Re: 486 cache/ram speed issue with write-back)

NB! If someone has a Sis471 Award bios with later datecode than 1995-11-28 (In award bios string formatted as 11/28/95) i would be interested to have that latest and greatest bios and i will make it suitable for this QDI board.

Thank you all for pointing me to the right directions until i've arrived to here, where i am! 😀

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 4 of 25, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie

So, i was benchmarking it and found that it has really low RAM read performance compared to my other, soyo 025p2 board. As a result i found that i had to do the same modification to the bios that i did for the soyo board so that it would use 7+1 bits in tag for tag and alter bit, instead of all 8 bits for tag bits. That solved the ram issue.

I also found a slightly newer sis471 bios from december of 1995. So i modified that with all necessary changes and i'm a happy puppy now. 😀

Amongst other things, i also found that JP44 position 1-2 enables 4x multiplier.

I attached the newest bios with all the fixes and whatnot.

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 5 of 25, by Warlord

User metadata
Rank l33t
Rank
l33t

Cool project

Reply 6 of 25, by Disruptor

User metadata
Rank Oldbie
Rank
Oldbie
GigAHerZ wrote on 2021-11-12, 00:15:

... i'm a happy puppy now. 😀

Amongst other things, i also found that JP44 position 1-2 enables 4x multiplier.
...

Fine. I should consider to keep a bone for you.

Which CPUs did you use?
AMD 5x86 ?
AMD 486DX4 SV8B ?
AMD 486DX4 NV8T ?

(first 2 should have similar jumper settings)

Reply 7 of 25, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie

I have a CPU with glued on heatsink, but it has 16kB of L1 cache capable of write-back mode and can do 4x multiplier. So it must be some kind of 5x86-P75 type of cpu.

The programs, that i'm aware of, recognize the cpu based on (basically) motherboard's jumpers, not by the real cpu itself. Even the CPU ID is changing depending on jumpers, so it must be generated from settings instead of being read from the cpu itself.
Therefore i don't really trust any software too much and have to rely on what i've found the cpu to support and being capable of...

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 8 of 25, by dj_pirtu

User metadata
Rank Member
Rank
Member

Great I found this thread! I have the exact same mobo with winbios and need to run that tag 7+1 "hack" in DOS to get RAM speed where it should be. Let's flash this one...

Last edited by dj_pirtu on 2022-12-24, 11:22. Edited 1 time in total.

Reply 9 of 25, by dj_pirtu

User metadata
Rank Member
Rank
Member

And works, no need to use that 7+1 tag -program anymore.

Results are with Am5x86-160, Cirrus 5434 VLB and CIRBOOST is used. Without CIRBOOST 3DBench gives 90.

If I set Local Bus Ready to Transparent and same time use CIRBOOST I can get 98,8 in 3DBench but there's graphical glitches all around.

Reply 10 of 25, by dj_pirtu

User metadata
Rank Member
Rank
Member

Does anyone know if this motherboard can do Ti486SXL-40 or Ti486SXL2-40? Manual tells nothing about it.

Reply 11 of 25, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie
dj_pirtu wrote on 2023-03-12, 07:11:

Does anyone know if this motherboard can do Ti486SXL-40 or Ti486SXL2-40? Manual tells nothing about it.

Aren't those 486-core CPUs for 386 boards?
This QDI is a 486 board with no 386 sockets. Purely physically it would not fly.

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 12 of 25, by dj_pirtu

User metadata
Rank Member
Rank
Member

I meant the PGA168 versions of those CPUs. Tried yesterday with Cx486 -settings but nothing, won't even post. Seems like the SiS471 don't support those at all.

Reply 13 of 25, by dj_pirtu

User metadata
Rank Member
Rank
Member

Swapped Cyrix 5x86/100 in this mobo and overclocked it to 120MHz (3x40). Damn the CPU is bad overclocker. Needed to raise voltage around 4V to get it stable with BTB on, in DOS. BTB needs to be turned off when in Windows.

Resistor value is wrong in the photo, it's just for the show. Been using 47-60ohm in parallel with the original. With 51ohm it's ~4V but there's some Vdrooop.

Reply 14 of 25, by dj_pirtu

User metadata
Rank Member
Rank
Member

Got myself a Pentium Overdrive 83MHz. Works on this mobo nicely but I can't get L1 in WB mode. WB worked with Cx5x86 and Am5x86 but because there's no jumper instructions for P24T I tried DX4/100 settings but no WB...

Anyone?

Reply 15 of 25, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie

Some time ago i was debugging this board quite a bit and i mapped out some jumpers on what they actually do. The information is in retroweb: https://theretroweb.com/motherboards/s/qdi-v4s471-g

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 16 of 25, by dj_pirtu

User metadata
Rank Member
Rank
Member

Oh yes, I remember I have seen that document before but totally forgot it existed.

Edit: so the WB/WT pin just works opposite way than with Am5x86/i486DX4WB, I need to swap J11 and J14.

Reply 17 of 25, by dj_pirtu

User metadata
Rank Member
Rank
Member

No, nothing helps, tried all combinations but WT mode is all I got.

Reply 18 of 25, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie

Only thing i can suggest is to download datasheet for the chipset and datasheet for the CPU. Figure out, how they should be connected and what connections to "high" or "low" must be made to either CPU or chipset and then in the end search out specific connections on the motherboard and apply that configuration.

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 19 of 25, by dj_pirtu

User metadata
Rank Member
Rank
Member

Figured it out, POD has it's own WT/WB pin which is out of the 486 pinout. I think it's intentionally shorted to ground in socket.