RAMDAC's do take input from the EDO RAM. If bad RAM is common that could certainly affect it.

In your last image .. the PCI and register settings to the FBI test OK so I would leave that, and mainly look at the EDO RAM.

Hi,

Thanks for your help, Any idea which one(s) to try first? in the thread I posted above, deeper into it they talk about limiting ram to 2mb using a variable in autoexec.bat. to narrow down which 2mb chunk of ram might be the problem. I will try that to see if it allows witchery to get past the ramdac error. Even if it does I am not sure which 4 chips to remove as different cards have their memory chip chunks set up differently... Also in that last image of mine it talks about unable to detect rs2 line? Any idea what that refers to? I've included a shot of mojo output and some more images of my card.

RS2 is a register select pin (RS0-2) for internal settings. But it's possible the RAMDAC isn't ready or intialising properly.

RS0, RS1, RS2, WR, RD, CLK0... CLK0 (video clock output) is determined by internal PPL register setup. If there's no default values and it's not setup, clock output would be wrong.

In the last image, there's a TMU count error for Board 0. TMU Revision looks wrong and PowerOn Sense is "0xdead".

Maybe TMU's didn't finish setting up, so the RAMDAC wasn't initialised either.

So RAM closest to TMU's..... or TMU's...?

so I should also be looking at TMU ram and TMUs? there is a tool somewhere that checks tmu ram. I need to google for that again. should ai try and reflow the tmu chips? I think I will proceed with fb ram first.

If EDO RAM is the most common fault, and there's no visible damage to the TMU's I would do the RAM first.

FBI RAM also reports as 0MB.

Swapped out entire bank of fb ram on the top of the board with new ram, reflowed solder on TMUs. Still same errors in mojo/witchery. Gonna take one back of board fb ram chip off at a time to see if I can get past this ramdac error... I understand this can also be the clock buffer chip. Witchery does mention it can be this chip u65 on my board, it is a 14 leg 74ACT125... ordered a few from digikey.

I took off one fb ram chip at a time from the back of the board, none fixed the issue. Also, did some thinking. When SLI board is attached, Witchery reports it's ramdac working and finds only one fb ram chip bad. There is no 74ACT125 chip or xtal on the SLI board, for ramdac chip to function there both should be good on the main board, right? If so I think I am down to a bad FBI chip or a broken trace between ramdac and FBI???

IMO if the SLI board is supposed to be attached from the factory, the BIOS may stop before initialising the majority of the board. No RAMDAC setup, no TMU count, no FBI memory count...

Only read the Witchery results when the SLI board is attached....

If FBI pins are reflowed.. and RAM is all replaced, hard to think of anything else except replace the FBI not responding in Witchery

The daughterboard with secondary chipset can be removed and the rest wil work just fine as single Voodoo2 (with pci-pci bridge and TV-OUT). So if the suspected defect is on the primary board, then I suggest to debug the issues with second board disconnected.

Thanks gents,
I've done both, run witchery both ways. It finds and works with the ramdac on the SLI board fine reports one bad fb ram chip which I swapped out today. I also reflowed all of the 3dfx chips on the SLI board. Also, reflowed the Digital PCI bridge chip on the main board. Will reflow sli ramdc too.. I have a working reference design voodoo2 and I have probed all of the connections between its ramdac chip and fbi chip, wrote them down. I get VERY different results from both my Q3d mainboard fbi and SLI board fbi. Many connections not getting signal in either maybe there are signals routed through the Digital pci bridge chip? Or through caps or resistors . Hell if I know why?? But it isn't a reference design, so. Also, seems mojo isn't a reliable diagnostic tool on my system. Both cards get the 0xdead and 0 fb memory found errors when connected in sli... I'm running freedos on a core2quad ls775 system.

Yep, If I can't find anything obvious soon I will swap in a new fbi....

Hey, how critical are the big yellow electrolytic caps on these boards? can they be sourced? what would be a good alternative?

The yellow caps are critical for stability, but as they're tantalum they'll blow up if they stop working... they gradually short. If all tantalums are slightly worn they may bring voltage down to a level that's too low. They're a common part.

Another issue may be the reflow. If the max 'non-operating' temps are the same as the max 'operating' temps then reflowing might destroy the chips. The alternative is to hand solder.

Are you sure witchery can correctly diagnose a Q3D board? Does any other test work?

I found some 10 mF 10v caps that look right, will those work? originals are 16v. That is a good question about my q3d board and Witchery. Witchery mentions the ramdac being a problem, mentions the buffer chip, x65 on my board. it correctly found THE bad fb ram chip on my SLI board. It doesn't say anything about my fbi chip being bad...

I suck at hand soldering, I have been using a small brass knob as a heatsink when doing the reflows. How much heat can the chips tolerate? I have the reflow set at 350c, Medium air flow. Only takes a dozen secs or so...

The q3d design routes a bunch of the normal ramdac to fbi connectivity through RAs that are just too high a value for my continuity check to beep. But everything does eventually get to the fbi. I did reflow those RAs for good measure on the Mainboard. Also swapped in a new U65. Still haven't seen anything odd. I also verified the clock signal is getting to the mainboard ramdac. It isn't the xtal, witchery mentions it but the sli board works and they share it...

I will ask in the witchery forum what they think about the fbi chip even though the software doesn't specifically mention a possible failed fbi...

The yellow caps are 10 uf 16v. Both critical numbers. There's a 12v rail on the PCI bus and some cards may use it for volt regulation on the DAC. The width/length numbers such as 3216 are measured in millimetres (3.2 x 1.6mm).

Max operating temp would be somewhere between 75-100c.

I don't know any other suggestions... Buffer chip as you said is a possibility.

Well, I may have fried my good ramdac with hot air? I found one on ebay. I also found the 16v tant caps on ebay. Thanks! I think I can use a soldering iron on the ramdac. I'd hate to try on an fbi chip, ha ha.