my half-functional agp 4x - 2x adapter..

Started by IceFire, 01 December 2006, 20:07:27

Previous topic - Next topic


hi guys!

since some time i'm trying to build an AGP 4x/8x to AGP 2x adapter.

by replacing the power supply lanes on an agp riser card, i was able to plug a voodoo3 in an i845 chipset based system.

sry, pic removed through traffic problems..
this is how the adapter looks like:
sry, pic removed through traffic problems..

i've tested the adapter on some modern agp 4x/8x boards... here my results:

-no post screen
-pci graphics card: device (v3) is recognized in windows, but when tying to install the drivers, i get a blue screen (page fault in unpaged area...)

i845 chipset based SIEMENS oem board:
-no post screen
-with a pci graphics card, i was able to install the drivers of a v3 (a v5 has not been recognized), but when trying to intizialise the v3 as a secondary card, the system freezes...on next boot it shows in the device manager: "could not launch device (code 10)".

i875p chipset based board:
-no post screen
-pci card freeze in BIOS (on v3 and v5)

well, hank had posted some years ago:


So far a bit of news on AGP 4X slot support. I did some looking arround, and found AGP wear extenders. This might make the AGP 4X slot possible for most people. I was going to cut a new key in the existing V5 AGP connector, but internal PCB planes extend into the area. If you cut into the PCB you will expose power planes and can cause a short.

I'll see if I can get an extender with 4X key cut on one side and universal socket mounted on the other. Most of the cases can support the aded height on the V5, but the bracket will have to be modified. VGA connector is low enough on V5, there would be no problem in connecting monitor to the shifted V5. The cost of the extender from 3rd party is about $50 US from what I have found so far.

Next VREF to VSA100 must be modified and a couple of straps. I have to check on AGPCLK, INT and RESET if the pins are still 3.3V on AGP 4X.

through this post and some intel pdfs i know that i have to replace the agp VREF signal (pin 66, side a & b).

so i did.
i assume that the VREF signal is dependent of the delivered voltage to the graphics card, so i connected the VREF-pins on the riser card with a matrox G400 (AGP 4x), isolated the pins there, plugged it in a dummy system (Agp universal, on witch the matrox should run at 3,3V due the A3 & A2 pins are not connected) and started both system at once.

it didn't work, but on the 1000x try the led of my monitor switched on , stayed 3 seconds on und switched off again.

well, hope that you where able to understand all that I've wroted there even through my english is the worst..

now, my problem is:

how can i simulate an agp VREF signal on a v5 so that the mainbaord will deliver 3,3V on the signal lines?


hank once said:
"Hi Guys,

VSA 100 is capable of supporting 1.5V signalling required for AGP4X slot. The IO voltage ring on the chip is isolated from 3.3V supply. There is voltage refference on V5 that will support 1.5V, but it is disabled through a pin called Sense. The mechanical problem of plugging V5 into AGP4x/8x slot can be gotten arround by using a riser. The MB will not burn since the IO voltage is isolated (AGP VDDQ Plane).

You will be limmited to PCI 66 speed, there will not be DMA! You can afford to send more virtex info and do texture paging upto the bandwidth limit of pci 66.

Hail to the king.....Baby (I think this one is better from Ash :-P)"

well, as far as i know, the agp bus delivers the signal amplitute that will deliver most compatibility when no vref signal is received - a agp 4/8x bus will send 1,5V and the voodoo, witch wants 3,3V is not able to understand anything.
but, hank said that the v5 support a voltage reference that would make it possible to have the v5 run at 1,5V. it would be helpful if i would know witch pin is meaned with "sense"

thanks, guys!

the thread in the german voodooalert forum:



Hi IceFire,

That's are very good news.
I'm sure that Hank can give the answer...
In this way there will be no need for an AGP2PCI riser anymore !
( Just the modified AGP Protector )
Congratulation and keep up the good work.
"Du bist auf dem richtige Weg....weiter so ! "


Maybe a closer look to a Voodoo4 ( AGP4x ) will help.
Follow your dreams !


no, the VREF lanes are disappearing after some condensators (?), resistors in the PCB :-(

hope hank will write back soon... i wrote him an email 2 weeks ago...


3dfx tricks, tests and wallpapers


wow, thank you!

i will need some time to understand the important things... ^^


can you post here exactly whitch pins do u cut and supplied with 3.3v ? [:p]
maybe a drawing or something will be easy to do because in the pictures i can only see around 4-5 Wires ..but i cant see where they are soldered...
I'm willing to try also :D

Check also this:
Follow your dreams !


hi guys!


i've cut every pin witch is described as a 3,3V VDDQ lane on the agp 1.0. specification but a 1,5V VDDQ lane in the agp 2.0 specification.

f.ex. pin 34A,B, 40A,B, etc.

i don't know axactly witch pins i had connected... sorry :(
you also need to connect the pins on the second agp notch. (one of them has to be connected to GND!)

well, i studied the documentation about the rampage chip...

"AGP_VREF_1_5VAGP Voltage Reference: This input supplies the switching threshold for the AGP
recievers for the 1.5V case. This is derived from a resistor network consisting of a 210
ohm 1% resistor to VDDQ and a 158 ohm 1% resistor to ground. There is a 0.1 uF
capacitor in parallel with the resistor to ground.

AGP_VREF_3VAGP Voltage Reference: This input supplies the switching threshold for the AGP
recievers for the 1.5V case. This is derived from a resistor network consisting of a 210
ohm 1% resistor to VDDQ and a 158 ohm 1% resistor to ground. A second RC network
cocnsists of a 210 ohm 1% resistor in series with a 560 pF cap to VDDQ and a 158 ohm
1% resistor in series with a 560 pF cap to ground. This is total of four resistors and two

i didn't understand much, but i think the vref lanes are a network of resistors and capacitators connected to the VDDQ lanes. that means that the vref signal is dependent of the vddq voltage. due i have no idea at all what i habe to construct to imitate this network, i've got an idea:
afaik the ati rage fury maxx has an agp 4x connector but is only able to run at 3,3V. as i was able to se on some other fotos, the vref lanes are existing on the pcb. as you can see in the documentation of  the rampage, the vref signals on the 1,5V and 3,3V cards are different (well... i'm not sure). due some users where able to get the rage running on an agp 8x board, that would mean that the vref on the board  is forcing the chipset to deliver a 3,3V signal. now, if we would connect the vref lanes on a rage fury maxx with the (empty) vref pins on the voodoo5, would plug the rage fury maxx in an agp 2x board (in order to have it run correctly), we should be able to tell the agp 4x/8x bus that he should deliver his damn 3,3V...
well, i need to buy a rage fury maxx and a second riser card...

or i need help of somebody who understands exactly what i have to build for a resistor network mentioned in the rampage documentation ^^

i appologize for my bad english (one more time)... but right now, i'm in holidays and my last english lessen has been 4 weeks ago...


Ps: and a happy new year to all voodoo users ;-)


m14radu: thank you for the documents!
in fact they say exactly what i had assumed. it should now be much easier for me to build the resistor network for the pins 66a,b - the vref pins...


i found the vref resistor network on a voodoo4...^^ it really looks close to the configuration describled in those 3 documents... but on a voodoo4, the resistor network is kind of the agp 4x speed mode. we need a card on witch the agp 2x speed mode is build on. well, i dont know what kind of resistor network is build on the rage fury maxx :/
-> but we're getting closer :D



Have u contact Hank ?
He is the only who can help us !
Follow your dreams !


hm, he still did not replied me...

but i think we should be able to solve that problem. it is not impossible....well, tomorrow, i will buy some resistors an capacitators and will try to build this ..."network".


so do u want to build the schmematics on pacge 40 ? :D
( second document )
Follow your dreams !


no, i want to build the one on page 39... this one is for the agp 2x mode, so might work with this adapter!
well, if this also doesnt work, i will try to build the one on page 40....


everything bought ;) ! i think i will build this construction in the following days...


hm, there's the usual god and bad news...

the bad: i still dont get a post screen with the vref modification for the agp 2x standart (i connected vref to pin 66A - i dont know what 2 do with 66B - the documents says that this pin is used for the signal from motherboard to graphic chip... i dont think that this one has to be used....).

the good one: on the i875p motherboard, i get a vga beep when installing the v3 with adapter, but i still dont see any post screen.

what i will do now:
-> make more test with a pci video card, try to get it running as secondary device, etc.etc.etc..................

-> if that will not work, i will try to build the kind of vref modification used for the agp 1,5V...

good night!
greetings from the cold southern germany ^^


I made some test, unfortunately i don't know if the i875p chipset based board had made the vga beed without the vref modification. the i845 sys still freezes when trying to initialise the voodoo3... damn.... %/(§$)/(§%%$"

i had checked the resistors and i found out that the deviation is almost 8%.... i will buy some new ones in the next time...


Quoting Hank, untill he has the time to explain it in details himself: "if you run the mother board at 1.5V IO and V5 at 3.3V IO, it will self destruct anfter a while"


In theory this will work for AGP V5 5500 and V4 boards. VSA 100 VDDQ can be set to 1.5V.

Do not try this with V5 6000 the hint bridge will burn up. This also will damage boards, if some yahoos improperly designed V4 boards with VDDQ and 3.3V planes shorted.  I think Powercolor was one of the outfits that did this.

VSA100 has 2 voltage references for 3.3V and 1.5V system. The comparator in the parts autodetects the VDDQ level and switches between the pins. On V5 the references were shorted together, I'll check on this, since I have one card taken appart with chips off. The design files were lost for a very long time.

In theory it might be possible to just cut out the 1.5V key, on V5 5500, and modify the card to take 1.5V directly. I'll have to check it again on the board.



hank, as far as i know, the agp vddq lanes are supporting the VRAM on the voodooo5. the vddq for the vsa 100 is reached by the power transformators on the upper right side of the card (at the molex connector).

on the other hand, my heavy damaged voodoo5 was able to run some time (bios post screen and bios)without connected molex. up to today, i don't know exactly how the hell this was possible (?).. that would mean that on this card the vsa-100 vddq was delivered over the agp bus.. or maybe through some condensators, i don't know.
the thread to this card in the german valtert forum:

a short summary in english:
i bought a defective v5. the second (right one) vsa 100 overheated itself right after some seconds, while the primary vsa 100 stayed almost cool. i had no post at all.
then i removed the slave chip with a hot air fan, damaged the pcb a little bit, but the card did work. it was a bit slower then a voodoo4, but run stable. then, i was able to get a tray of unused vsa-100 chips. i installed one on the damaged socket with a fastening clamp, forget to plug the power cable in, but the card powered up.. for 3 seconds, then the pc switched off. i was able to see the post screen.
when i plugged the power cable in, nothing happend (no post screen).
wehen i removed the vsa-100, the card powered up like always..

as u can see on the foto, many pins has been ripped of the surface and many lanes are ripped out... i did not work very clean :-(

sry, pic removed through traffic problems..
some more photos of the card:
sry, pic removed through traffic problems..
well, hope hank knows what to do now ^^.