Hello,

The unique thing I have clear is that the first vsa-100 is working!
For such reason the card works ok in single chip, it means master chip is ok!
I would bet there are some problem in chip 2 or 3 or 4, but I am not 3Dfx engineer!

When some chip is dead then the card show lines in the monitor because card can't render...

Repairing card now:

Hint chip unsold/resold...

http://img4.hostingpics.net/pics/447944V5k6halfdead1.jpg

http://img4.hostingpics.net/pics/391096V5k6halfdead2.jpg

http://img4.hostingpics.net/pics/923150V5k6halfdead3.jpg

http://img4.hostingpics.net/pics/339916V5k6halfdead4.jpg

http://img4.hostingpics.net/pics/825283V5k6halfdead5.jpg

http://img4.hostingpics.net/pics/688230V5k6halfdead6.jpg

3 new chips vsa-100 are perfect soldered...

http://img4.hostingpics.net/pics/790083V5k6halfdead7.jpg

http://img4.hostingpics.net/pics/429896V5k6halfdead8.jpg

Just there is one little problem now.

http://img4.hostingpics.net/pics/749637V5k6halfdead9.jpg

http://img4.hostingpics.net/pics/841286V5k6halfdead10.jpg

http://img4.hostingpics.net/pics/760881V5k6halfdead11.jpg

Smd parts need to be resolded.
For such rework we use a hot air gun. Seems in the proccess some smd parts was unsoldered. They did not see it.
All is perfect, GPUs chips are very well soldered.
The card booted up ok, just I stoped because I dont want to test it until all smd parts are soldered again.

To be continued...

Although I don't care much for your custom cooling setups, I do appreciate (and I'm sure others do as well) you spending the time and money to fix these legendary cards.

gwb wrote on 08.06.12 at 15:18:51:

Although I don't care much for your custom cooling setups, I do appreciate (and I'm sure others do as well) you spending the time and money to fix these legendary cards.

Hi gwb,
I want to progress.
Thank you.

The soldering looks great, and professional !

Michael, don't forget that the v5 6000 cards are still prototypes.

Mine is working best only in win98se !
I sugest to check first if your card works in win98se, before you change something else  ;)

Best of luck with this one, I hope you succeed. :)

Good luck with the repair - hope you manage to rescue another of these legendary cards! 8-)

Hello,

Thanks for your support.

I tested the card under win98 with same problem.

Here is the smd reworks:

http://img15.hostingpics.net/pics/662606V5k6halfdead12.jpg

Card continues with sli broken! :(

I would go to change master vsa-100 and see what happens...

can you make a movie with the card behaviour ?
i want to see the errors.

What driver did you use for win98 ? you should take the original 3dfx release.
And please check the card under win98se, not win98.

Some things I would do:
1) re-flash the bios with a common version for rev 37000 cards
2) down clock all VSA100 to 143MHz or even lower
3) apply the PCI rework
4) change the first VSA100 (you mentioned that already)

Compare the card also to other common rev 3700, maybe there are some differences.

Could you do some ultra highres shots? Not those small pictures here ;)

Keep it up.

Please post a close shot from the revison tag too ;)

m14radu wrote on 10.06.12 at 07:23:17:

can you make a movie with the card behaviour ? i want to see the errors. What driver did you use for win98 ? you should take the original 3dfx release. And please check the card under win98se, not win98.

I only have some screens:

http://img15.hostingpics.net/pics/697094Voodoo5k6tenthtest.jpg

On Windows 98se I use original 3dfx release.
On Windows XP I use SFFT1.9 and Koolsmoky betas drivers.

Loeschzwerg wrote on 10.06.12 at 11:29:44:

Some things I would do: 1) re-flash the bios with a common version for rev 37000 cards 2) down clock all VSA100 to 143MHz or even lower 3) apply the PCI rework 4) change the first VSA100 (you mentioned that already) Compare the card also to other common rev 3700, maybe there are some differences. Could you do some ultra highres shots? Not those small pictures here ;) Keep it up.

My friend Osckhar do the smd reworks.
A very good friend from Osckhar unsold and resold the Hint chip and VSA.
All tips you said are already done.
All ultra highres shots are in my brinkster website like these 3:

http://img11.hostingpics.net/pics/579630Voodoo5k6tenth1.jpg
http://img11.hostingpics.net/pics/973208Voodoo5k6tenth9.jpg
http://img11.hostingpics.net/pics/283734Voodoo5k6tenth3.jpg

trevormacro wrote on 10.06.12 at 17:21:43:

... All tips you said are already done. ...

I don't see the PCI rework attached, at least when I look at the picture in your post today at 00:33:30 -> http://www.falconfly.de/cgi-bin/yabb2/YaBB.pl?num=1339157722/6#6

Add 4.7k resistors on R540, R589, R717 and R714. Change R734, R735, R743 and R745 to 0 Ohm.

Everything else does look fine to me, but the pictures on your site are a bit blurry. I hope changing the first VSA100 helps and that there is no damage on the PCB.

Loeschzwerg wrote on 10.06.12 at 18:04:16:

trevormacro wrote on 10.06.12 at 17:21:43: ... All tips you said are already done. ... I don't see the PCI rework attached, at least when I look at the picture in your post today at 00:33:30 -> http://www.falconfly.de/cgi-bin/yabb2/YaBB.pl?num=1339157722/6#6 Add 4.7k resistors on R540, R589, R717 and R714. Change R734, R735, R743 and R745 to 0 Ohm. Everything else does look fine to me, but the pictures on your site are a bit blurry. I hope changing the first VSA100 helps and that there is no damage on the PCB.

Pci rework is not the solution. It is used when card lock up in 2/4/8xFsaa (card comes without pci rework. Anyway, it is not the solutuion since this rework is added when cards works ok but with lock up when fsaa is enable).

Mainly you are right, the rework was meant to fix the FSAA lookups. It changes the timing so the PCI bridge does have more of it to synchronize the signals. No matter how it changes the timing of the analog SLI itself and this is what's not working on your card right now. So I think it would be a chance ;)

Hi,

Here picture when sli is enabled.

https://dl.dropbox.com/u/4383540/Michel/6k_02/IMG-20120611-WA0001.jpg

I was talking with Hank about this problem and he told me the probem can  come from 2, 3 or 4 chip, probably there is a memory issue and chip never returns back and holds all the chip back.

So, the way would be to have a good tool to isolate the failure and check command registers  and see what chip fail.

Any tip?

Note- I was thinking if could be possible to change via HW or SW the master VSA-100 chip and run de card in single mode.

Best regards,
Oscar.

I have a stupid question:

Do the VSA100 render the screen sequentially?

I mean, from top to bottom:

1st VSA100 - 1st row
2nd VSA100 - 2nd row
3rd VSA100 - 3rd row
4th VSA100 - 4th row

..and then again from the beginning..

If it was so easy, I'd say the second chip seems to be affected in some way, no?

Memory sounds possible to me too. Chip 2, 3 and 4 have been changed, so these should be good.

I noticed a similar behavior of the image of my 5500 with a defective memory on the second VSA100.

It would be nice to know how the signaling between the VSA100 works, like goriath wrote. Maybe it is possible to deactivate/unsolder one of the slave VSA100 and the V6k works with 2 or 3 chips only. This way we could identify the defective memory cluster and would only have to change those. 

Yes, it could be but check the picture. If second vsa-100 is the blue row and  it is sequential. The space between blue rows should be more width since that the second row blue would correspond to the first vsa-100 now.

It should be
1
blue
3
4
1
blue

But for the image seems:
1
blue
3
4
Blue

About memory, I have noticied different artifacts when one tsop is in bad condiction though.

ps. Perhaps in the bios is set the master chip. it would be the best way to check each vsa-100 in single mode.

Regards,
Oscar.

Loeschzwerg wrote on 11.06.12 at 11:57:10:

Maybe it is possible to deactivate/unsolder one of the slave VSA100 and the V6k works with 2 or 3 chips only. This way we could identify the defective memory cluster and would only have to change those.

Oh well, this is really a good point.

A Voodoo5 5500 with a physical damage to the second VSA can still work. The system sees half of the card, like it was a V4 4500.

I was thinking we have a PCI2PCI Bridge device involved; dunno how this could affect the test. BTW it's a good idea.

osckhar wrote on 11.06.12 at 12:14:00:

Yes, it could be but check the picture. If second vsa-100 is the blue row and  it is sequential. The space between blue rows should be more width since that the second row blue would correspond to the first vsa-100 now. It should be 1 blue 3 4 1 blue But for the image seems: 1 blue 3 4 Blue

Honestly, at the beginning I thought the same. But actually paying more attention to the image, it looks like the space between two blue rows is too much for being like that, but also too little for being like I said.

Oh well, I can't say, it's too difficult only looking at picture. It was only a supposition.

EDIT: BTW, still looking at picture. If we are assuming that the first black row is a single row rendered by the first VSA, the last black row at the bottom is wider than the first row on top, but thinner than the rows between the blue ones. So that would suggest:

1°
2° - (blue)
3°
4°
1°
2° - (blue)
...
...
2° - (blue)
3°
4° (since it has to end even)

So, each VSA renders 12 rows. It also depends by the resolution...is there no way to know what res the card was attempt to run?

EDIT: I'm trying to figure out if could be plausuble. Assuming 1024x768 res, it should look like this: http://img40.imageshack.us/img40/3668/v56ksli.png

Yes, 1024x768.

Anyway I will try to take better picture. It was taken with the iphone. LoL

http://img59.imageshack.us/img59/3451/angepasst.jpg

So it looks like a problem with the second VSA100. A shot in the dark, just change the memory cluster that leads to the second VSA100.

check if the behaviour is the same as here:
http://www.youtube.com/watch?v=xyu7KLdwrPE

m14radu wrote on 11.06.12 at 21:50:53:

check if the behaviour is the same as here: http://www.youtube.com/watch?v=xyu7KLdwrPE

Hi Radu,

Nope, when card launchs a 3D application the image is like I showed above.

It is time to reemplace the tsops and master vsa-100.

I keep you informed.

Best regards,
Oscar.

i would suggest to start with the memories first.
I have a hunch :)

Any update on this one?