3dfx Archive - Print Page

DAYTONA PCI IBM MOD @32Mb PROJECT.

This project is based on 2 MODS. One IBM V4-2 4200 PCI 16Mb DDR will be modified with 64Bits memory and the other with 128Bits memory. Both cards will be upgraded to @32Mb DDR.

[03/02/2014]
Here Gino's card. It will be mod with extra 16Mb DDR 64 Bits.

DDR memory 64Bits ready for soldering.

04/03/2013

Bios mod done and card shows 32Mb DDR once added extra ram.

Card still shows artifacts.

10/03/2014

I still having problems- trying differents settings via hack bios my card is 'dead'. Not really dead since PCI.exe detects it but seems I used a bad parameter for hacking bios and once in the flash process it failed. I have tried to flash back my card with original bios but flash utility shows a message is not possible since card is detected as unknown card.

PCI.EXE detects it as:

In the right top left where FF is present there must have 1A / 12 / 0X but SW does not let me to change it manually. :(

I have tired to reset the chip bios but not way either.

So wahat I decided, I always wanted to do it, is to add a socket bios:

I used a programmer SW tool for flashing the chip with original bios but not way either. Card not wake up.

My question is if VSA-101 has small rom in the die chip with info about ID VENDOR from the card. How is possible once added a new chip PLCC with original bios card continue without boot up. Seems as I changed something related ID VENDOR card.

Today my tests.

11/03/2014

I found the problem and I fixed it. Seems hacking the bios with an HEX I modified some id string ans flash utiliy detected the Card as a Voodoo but bios is Naplam/Napalm2. Across the PCI.EXE modified some parameters taking as reference the other Daytona (Not all the values can be changed) but it was enough for flashing tool detects again as a NAPALM/NAPALM2. My error was the first time I used a V5 as reference thinking it was a Napalmd2.

Finally with a new hack biios the card works perfect.

BOOTING UP

IN BIOS

HERE SOME MS-DOS TESTS BASED IN DIFFERENT VGA MODE :D

Now I need to test both DAYTONAS in Win98SE but at the moment seems all is correct and not glitches.

14/03/2014

Installed Win98SE with reference driver using a KT333 elitegroup motherboard.

Not artifacts in 2D mode. Info driver detects correctly the card with 32Mb.

In 3DMode there are not artifacts either but in OpenGL (Quake3) card locks up and in 3DMark2001SE I got a black screen in first test.

I think the problem is based on drivers. I need more testing.

19/03/2014

I found the problem why card freezes and downclock speed @143MHz and now is very stable.

Quake 3 1024x768

3DMark2000 1024x768 16Bits

I found dirver is not optimized for Daytona cards. 3DMark2000 has glitches if depth color is set to 32Bit, 16Bits is perfect. 3DMark2001SE has artifacts and only work more or less well with depth color 16Bits.

Today I will do some comparative between both daytonas and different memory size.

Regards,
Oscar.

Looking good Osckhar, will you be doing benchmarks before and after??

And compare em to a regular 4500 when ram is 32MB??

ultima wrote on 05.02.14 at 07:48:36:

Looking good Osckhar, will you be doing benchmarks before and after??

And compare em to a regular 4500 when ram is 32MB??

Yes, I will do different test before and later of the rework

Just little of delay here with a problem of different brand memory and waiting to fix it.

Regards,
Oscar.

I'm really curious if the 4200 @ 32MB performs the same or better then a regular 4500.

128bit SDR vs 64bit DDR...in theory it should be the same....but that is ofcourse in theory.

Great Projekt Osckhar, as always :)

Thanks at all! :)

In this project the most important is check what is the performance of a Daytona 32Mb DDR 128Bits instead of 64Bits. A V4-4500 32Mb SDRAM 128Bits is little more fast than a Daytona 32Mb DDR 64Bits at same clock speed.

Regards,
Oscar.

I thought Daytona only supported 64bit DDR?

Yes, 64Bit memory controller only since VSA101 was cost reduced version of VSA100, as far as I know.

Loeschzwerg wrote on 07.02.14 at 07:19:36:

Yes, 64Bit memory controller only since VSA101 was cost reduced version of VSA100, as far as I know.

True, but we all know the "facts" about the 3DFX hardware can be not 100% true..used to be fact that a vsa-100 supported max 64MB per.chip....so....never hurts to try ofcourse

Since there is not datasheet for VSA-101 we will not know if dyatona chip is only 64Bits interface memory controller or it is dual 64/128Bits.

Since DAYTONA chip is thought for a low end card could be not compatible with 128Bits.

Regards,
Oscar.

Beside all the reports back in 2000 it is very unlikely the VSA-101 does have a 128Bit DDR memory interface. The performance of the architecture itself wasn't competitive anymore back in the days and it would've made the DIE too expensive for what it was meant. Transistorcount and DIE size needs to stay very low for low-end market segment.
Also 3dfx didn't have the ressources to waste on a "complexe" VSA-101, Rampage was the way to go :)

But you can surely give it a shot.

Updated first post.

- Oscar.

Great work as usual Oscar!

Well done Osckhar !
Is the card stable ? Please post more details :)

Thanks Gary and Radu!

I am having problems. First post updated.

Regards,
Oscar.

Thats strange !
Check the bios socket solder points again...

Radu, all was perfect just problem with id vendor.

First post updated!!! 8-)

once again Osckhar.....YOU DA MAN!!!!

how many cards have you brought back from being "defective" to fully functional again now?

You must be a true Voodoo Witchdoctor :)

Great work !!!!!
Glad to see it is running again and with 32mb :)

I want one daytona toooooooooo !!!!

Thanks for words guys! :)

@ultima, I have fixed a lot of cards since Rampage, V5-6000, Daytona agp/pci, V5/V4, v2, so on... Well, I only fix 3Dfx prototypes since time spent on reworks is important and not a worth for me fix common cards. :)

FIRST POST UPDATED WITH MINOR INFO.

- Oscar.

Great results !

Please bench the card and compare it with a 16mb card :)

I am waiting for the 256mb v5 6000 mod too !

Hi Radu,

First I need to find a stable motherboard and driver and later I will post different benchs. Smiley

Just thinking about V5-6000 @256Mb modded card- if a common V5 @128Mb already has many incompatibility issues motherboard I dont want to imagine the problems can have the V5-6000. For other side with only soldering and adding again smd resistor in his original stated V5-6000 will be again stable as in the beginning.

Yes, I think I will! Cheesy

Regards,
Oscar.

I am sure you can mod the v5 6000 :)
It will be a succes !
The first ever with 256mb !

First post updted. Now Daytona works perfect. :)

@Radu, yes, I will...

Hi Oscar,

Very nice progress with this Daytona card. It's good to see it running 3D application in a stable manner.
Did you reflash the card with v1.05A 143Mhz BIOS or just reduced the clock speed in Windows via V.Control ?
Did your card with newer VSA100 chip revision work any better @166Mhz speed?

Thank you,
G1nX

3dfx Archive
http://www.falconfly.de/cgi-bin/yabb2/YaBB.pl General Section >> General Discussion >> DAYTONA PCI IBM MOD @32Mb PROJECT. http://www.falconfly.de/cgi-bin/yabb2/YaBB.pl?num=1391509569 Message started by osckhar on 04.02.14 at 11:26:09