3DFX Voodoo5 5500 works on PCI-Express !!

[m14radu]

ok, so i had here an old hard drive, and i install on it the winXP - special to test if the adapter is working or not @66MHz.

The answer is no.
I install the Koolsmokey driver, the SFFT, and Amigamerlin: with every of this drivers, i get around 2900 points....so it's working @33MHz.
Aida64 still detects that the pci2pciecard doesn't support 66MHz !

I tested the adapter in PCIe 1x and 16x. - same result
I used two V5 5500 PCI with different bioses. - same result.
I used sell tape to isolate the pin49B on the V5 and to be sure i unsolder the pin on the pci2pcie adapter. - same result.

Btw. i install Koolsmokey and Amigamerlin R3.11 driver under Win7 too, to test them. The performance is low, they are wrking but the 3dmark result is around 1400 points - i belive that the card is working in single chip mode.

@Osckhar: Please post some pictures with your moded card, and with the adapter.
Did Aida64 detect now your pci2pciecard with 66Mhz support ?
Please check if the Pin49B on your pci2pcie adapter is conected to the ground. If not, where is conected ? ( to VDD )

If somebody has other ideeas please coment

[jix-reggio]

Mradu : Thanks for your very clear pictures and screens.

A chart of my results in 16 and 32 bits.


I got exactly the same information than you with Everest.
1/ V5500 Pci (pc version) is told to be '66 mhz capable' but this item is not mentionned as 'activated'.
2/ PEX8112 is told to be 'Master capable' but not '66 mhz capable'.

I am not a specialist, but If we want to run the bridged V5500 @66mhz we have to
take in consideration hardware, bios and software informations processed at the init of the system.

As the Pci-E link is told to negociate by software its capabilities at the init,
it is not sure there is a 'pure hardware' solution, maybe it is a mix of hard, bios and soft.

Following last Sfft post, I understood the '66 mhz capable' could to be related
to the '256 Bytes Pci configuration space'.
Everest says the pci device is '66mhz capable' when the 'status register' bit 5 is set to 1

So if PLX chip PEX 8112 would manage the 66 mhz, this bit must also to be set to 1.
Can a Software like Wpcrset update it early in the init ?

If Oschkar had find a solution, why did he not post pictures and screens of the mod ?

[m14radu]

Osckhar will post picture and more details tonight
I'm also going crazy because mine is not working @66MHz.

Let's wait till tonight!


@jix-reggio - thanx for the GREAT ideea with the pci-express adapter !

Regarding the overclocking possibilities, there is a way to overclock for sure both VSA's under Win7 - there is a modded bios, witch alows you to change the freqvency from the system start.

Let's hope that the 3dfxtools and vcontrol would be soon updated with "Win7 READY "

[jix-reggio]

Mradu14, using siv was very interesting.
If you dump (click 'dump pci' on a device) you'll see the pci configuration zone at the offset 06 :

1/ For the V5500 Pci a hx value of 80BO -> Bit 15 to bit 0  -> 1000 0000 1011 0000
where bit 5 (66 mhz capable)  is set to 1.
As sfft said, the same dump of my V5500 Pci PC version proved it was 66 mhz capable.

2/ For the Plx bridge a hx value of 0010 -> 0000 0000 0001 0000
where bit 5 is set to 0. So It doesn't prove the bridge is 66 mhz capable.

As the PLX is a bridge beetween bus x/0/0 and bus y/0/0
You where mistaken by the Bus destination y (it xas the V5500) indicated as 66 Mhz capable.

[m14radu]

@jir-reggio: Looks like i'm not the only one who is concerned about this Problem and can not sleep

Look at siv, my red marks :



Status is exactly the same in my configuration for voodoo5 and bridge.
But look at the BUS 3 status : 0x2220 - support 66MHz operation.
What this means exactly ? Do we use the wrong bus ?

I am not skilled in used Wcpregedit or Wpcrset softwares, but it might work to change the register status with them !

[osckhar]

Hello,

First fo all I must say I use another model of pcie2pci adapter . With this adapter you can enable 66MHz:

• Unsolder PIN49B directly on PCIE2PCI.  
• Isolate directly the PIN49B on V5-5500 PCI  

Here some pictures:

MOD ADAPTER @66MHz:


ISOLATE V5-PCI @66MHz


Here some test using 3DMark2001SE:

• V5 PCI pcie2pci - 3450Points
• V5 AGP pcie2pci+agp2pci - 3361Points  
• V5 PCI pci-x slot - 3356Points
• V5 AGP agp2pci+pci-x slot - 3356
• V5 PCI pci slot - 2871Points

All the settings to default. No overclock or tweeks.

Note- Important to use Beta Kool driver if you use SFFT driver the performance is low because this core has a bug when adapter (sli bridge) is used and SLI configuration does not work well.


Regards,
Oscar.

[jix-reggio]

Oschkar, could you post a dump of the pci configuration area for example
with the siv program (to be executed in Windows 2000 compatibility mode on windows XP) for the modded PLX device and the V5500 device to see how the bits are setted at the init when modded.

[osckhar]

Originally posted by jix-reggio

Oschkar, could you post a dump of the pci configuration area for example
with the siv program (to be executed in Windows 2000 compatibility mode on windows XP) for the modded PLX device and the V5500 device to see how the bits are setted at the init when modded.

Of course, I will this night.

- Oscar.

[SuperFurryFurryThing]

oskhar I would like to see comparative tests with the sfft driver, as the evidence for any bug with a bridge is questionable at best. I would like to see more compelling evidence, beofre accepting there is a problem. If people can run comparative test with both drivers it would supply us with more data.

SFFT

[voodoo5k6]

Osckhar PCIetoPCI adaptor :  with PERICOM PI7C9X110BNBE chip (in 66 mhz)

Other PCIetoPCI adaptors : with PLX Ref : PEX8112 (in 66 mhz too)

[m14radu]

Voodoo5k6: did you try to mod your adapter too ? It's working @66mhz ?

[osckhar]

Originally posted by SuperFurryFurryThing

oskhar I would like to see comparative tests with the sfft driver, as the evidence for any bug with a bridge is questionable at best. I would like to see more compelling evidence, beofre accepting there is a problem. If people can run comparative test with both drivers it would supply us with more data.

SFFT

Hi SFFT,

Here I let you some test I did.



I have used SFFT 18 because jointly with v.19 is the most fast (fastet mode). Later version 19 until v47 seems to be a problem in D3D in fastet mode (3DMark does not work, only in single mode). It is fixed in vers. 48 but the performance is penalized.

- Oscar.

[SuperFurryFurryThing]

oskhar can you run 3dmark2001 having run set3tile.exe assuming that you are using double buffering and also post your system spec for these results. I presume that you are testing using 2 way SLI (default) for the 5500 and 4 way SLI (default) for the 6000. And also doing testing with AA enabled.

SFFT

[osckhar]

Originally posted by SuperFurryFurryThing

oskhar can you run 3dmark2001 having run set3tile.exe assuming that you are using double buffering and also post your system spec for these results. I presume that you are testing using 2 way SLI (default) for the 5500 and 4 way SLI (default) for the 6000. And also doing testing with AA enabled.

SFFT

Ok, I will test it with set3file.exe.

For V5-5500 2-WAY SLI / 2xFSAA
For V5-6000 4-WAY SLI / 2xFSAA

My system:

ASUS P5WDG2-WS PROFESSIONAL
ChipSet 975x
PCI-X Bridge: Intel 6702 PXH
PCIe @90MHz
E8400 @3GHz
4Gb DDR2
WinXP SP3

- Oscar.

[SuperFurryFurryThing]

oskhar can you test with SLI enabled but AA Disabled, i.e. in fastest mode for each card, and separately with AA enabled.
Can someone test a simple bridged setup with a single bridge and a 5500 PCI card.

SFFT