Thank you, Rottentreats; the pics are ritght what I wanted!
OK... Have a look: to duplicate the Voodoo5 5500 AGP mod to 128 Mb, you need to replace the orriginal 8Mb RAM chips with 16Mb memory modules, then follow the instructions described in the next pics:
No BIOS mod is required if you use this hardware strapping.
Happy soldering!
Chosen_One wrote on 04.10.08 at 14:26:14:Yes, in order to get all the information in this thread, you could copy and paste all the information from the "group-buy-thread" in this thread. So noone have to crossread in order to get all the information about this really impressive piece of work!
Allright then. Bellow I will copy/paste what I think of as relevant pieces from the other topic:
I had the ideea that it would be fun to play with the strapping setups to initialize different values for the usable RAM on the card.
Anyone interested in an elusive Voodoo5 5000 AGP (16MB memory for each GPU for a total of 32Mb onboard)? Well, not realy the true one, but my test card with only a quarter of its installed RAM activeted:
Ofcourse, the card runs OK with 2x16MB RAM, but with much reduced performance.
The 128Mb modded Voodoo5 5500 is now working with no problem on Single Chip setting.
You all are familiar with the next screenshots... They are taken when using the orriginal Voodoo5 5500 AGP on Windows ME and x3DFX drivers.When running 3DMark2001SE on a 64Mb Voodoo5 5500 AGP with Single chip setting, I got 1687 points.
Time to enter the "mutant" 128Mb Voodoo5 5500 AGP:
As you can see, I got 1815 points, that is 128 3DMarks increase in performance, so the extra RAM is realy being used.
And it looks as it is working in full 128Mb.
3DMark 2001SE 64Mb 4x Anti-Aliasing: 733 points
3DMark 2001SE 128Mb 4x Anti-Aliasing: 823 points
Not much, 90 3DMarks extra performance, but clearely better.
In Single-Chip mode, there is a clear difference in performance.
In Fastest-Performance mode, the scores are virtualy the same order.
In 4x Anti-Aliasing mode, the 128Mb is visible better than 64Mb.
In 2x Anti-aliasing mode, see below:
3DMark 2001SE 64Mb 2x Anti-Aliasing: 1662 points
3DMark 2001SE 128Mb 2x Anti-Aliasing: 1774 points
That stands for a difference of over 110 3DMarks, so I say it clearely shows that the extra memory is being used.
The other conclusion is that we need to find another much more texture intensive benchmark than 3DMark 2001SE, another test that could realy make use of more video RAM.
Thanks for the congrats and cheers, friends!
Edit:
The RAM used for modding is:
http://www.esmt.com.tw/DB/manager/upload/M12L128324A%2040%2085C.pdf