AaronAsh wrote:

So I've tried PCBONEZ suggestion of running the system for a few hours then letting it cool a few times, and with two different Voodoo3s, and in neither case did it seem to fix or have any effect on the problem.

Bummer. I was pretty sure it would work.
If it did I was going to buy a few.
.

Or it could be that the issue is on the motherboard and all cards are fine to begin with. Guess there is no way to tell until your NVidia card arrives.

alexanrs wrote:

Or it could be that the issue is on the motherboard and all cards are fine to begin with. Guess there is no way to tell until your NVidia card arrives.

Three plus members have had the same problem with the same cards from the same seller.
I don't think all their motherboards are bad.
.

Yeah, but it could always be that this specific case is about the motherboard, and the others would be able to bring the Voodoos back by leaving them on for a while. The only way to be sure is to have more data.

alexanrs wrote:

Yeah, but it could always be that this specific case is about the motherboard, and the others would be able to bring the Voodoos back by leaving them on for a while. The only way to be sure is to have more data.

I would like to know what the ESR is on the caps.
A big lot of NOS boards that don't work sounds a lot like they were stored someplace too hot.
Time (with no DC applied) and heat are both hard on caps and these cards are ~15 years old.

In storage the oxide layer (the actual dielectric) slowly dissolves back into the electrolyte solution.
Heat accelerates that.
DC voltage applied reverses that process and rebuilds the layer. (This is called "Reforming".)
The power-up for 2 hours I suggested is a red-neck ghetto version of reforming but it usually works fine for low voltage (under 50v) caps.
For 15 year old caps 2 hours should be enough.

The one possible problem is if the layer got so thin that the leakage current is too high that could destroy the cap rather than reforming it.
Not usually a problem but it might be with caps that have been improperly stored for 15 years.

The "official" way to Reform caps is to put them in series with a 1k ohm resistor (which prevents the possible problem of excessive leakage current damaging the cap) and apply a low voltage (1 or 2 volts).
Then slowly raise the voltage in steps until the cap has the rated voltage across it.
Then it is held at the rated voltage for a period of time.
- There are as many variations as to how large the voltage steps should be and the times as there are people who have written down the procedure.
An example would be: Start at 2 volts. Go up 2 volts every 5 minutes until the rated voltage. Hold there for 20 minutes + 15 seconds for each month it has been stored.

Some capacitor companies recommend capacitors be reformed if they have been unused for more than 1 year.
Obviously that's a bit excessive for caps used in computer equipment.
Their recommendation is general for ALL caps they make including 400-450 + volt caps and those the size of a trash can.

[Edit][ I just checked at Nichicon's site and they have added the following to one of their tech docs since last I read it.]
When an aluminum electrolytic capacitor is stored under no load conditions for a long period of time, its leakage
current tends to increase slightly. This is due to a drop in the withstand voltage of the dielectric caused by the
reaction of the anode oxide layer with the electrolyte. When the voltage is applied to the capacitor, the leakage
current returns to its initial level because of the re-forming action of the electrolyte (called voltage treatment).
If the storage temperature is high, the leakage current will increase substantially. Therefore, it is desirable to store
capacitors at normal temperature level with no direct sunlight. A voltage treatment is recommended when using
a capacitor stored for a long period of time. The treatment for an individual capacitor is accomplished by charging
up to its rated voltage through a resistance of about 1k Ω and applying the voltage for approximately 30 minutes.

When a capacitor is already built into an appliance, the appliance must undergo aging. If the input voltage is
adjustable or the power supply that supplies power to a module, first set the voltage to a low value (approximately
half the rated voltage) and let it run for about ten minutes. Then, increase the voltage to the appropriate value little by
little while monitoring the working of a device.
If the voltage is not adjustable, turn on the switch and let it run for about thirty minutes while confirming if the device
complies with the specifications. Then turn off the switch before using the capacitor for practical applications.

Generally, if the capacitor has been stored within 2 years in the storage temperature range of 5~35˚C, the
capacitor can be used without voltage treatment.
.

Thanks for the information PCBONES, very interesting stuff even if it hasn't been helpful in my specific case. I would also be really interested to know if anybody else who has one of these cards from combinationinc is able to revive them via that means - I've done enough hardware troubleshooting to never assume there is only one problem! I could just be unlucky and have bad cards and a bad motherboard.

I did a bit more testing over the weekend in that machine using the S3 with 3D Mark 2001 and it worked pretty much as well as could be hoped for an S3. One thing I did notice was it suffered occasional texture corruption on things like text and specifically the HUD in the helicopter part of the benchmark - to me it looked like perhaps the S3 was struggling to render features it didn't fully support (textures too high res?), but worth mentioning.

TNT2 hopefully arriving today or tomorrow.

I'm sorry to hear about your troubles, but I'm relieved that others have had the same issues with the cards from this seller. It's possible that the ones I returned got sent out again without testing, but konc stated that his were tossed, so there are at least two bad batches even if the seller resold my return. The seller was rude and borderline abusive when I said that all three cards were bad, so this thread is a vindication of sorts. I did eventually get my money back.

boxpressed wrote:

...so there are at least two bad batches even if the seller resold my return...

I'm sure that there are waaaay more that two, the whole batch seems to be problematic. We all bought cards in different dates, it just can't be the case that the 4-5 faulty cards ended up in this thread. For me every card of this seller's stash is problematic. Unfortunately (and inexplicably) not on first boot.

So TNT2 Pro arrived (specifically it's an Elsa Erazor III Pro).

Good news! The system posts and gets into Windows with it.

Bad news! I seem to have run into either a K6 AGP problem or an issue with my monitor, as after I install the driver for the graphics card it only gets as far as the Windows splash screen then just goes to an unresponsive black screen.

Doing some reading this seems to be one of the many problems suffered by non-3Dfx AGP cards with the Ali Aladdin V chipset. Suggested solutions thus far have been BIOS tweaks, registry tweaks and installing the latest 1.9 AMD AGP driver, none of which have worked; I've also tried a variety of vintages of Detonator drivers, and the official Elsa drivers.

I actually suspect what might be happening is the card is loading into a refresh rate my LCD monitor cannot cope with, as I had a few issues like that with the S3 when changing resolution. If I start the system up in safe mode and try to change the resolution/depth options it says it is going to restart again to let me pick them in normal Windows, does so, but then only displays a really limited selection of options (640, 800 and 1024 for resolutions, locked to 16 colours at anything but 640) - selecting anything there then applying and restarting just ends up with me back at a black screen after the Windows splash screen. Is there some trick I should be using to get my LCD monitor to enumerate its modes correctly to Win98? Currently it is just using the standard PnP driver, though Windows does actually correctly read the monitor model from it.

Any other suggestions for possible AGP related issues greatly appreciated.

@bospressed As it looks like I might be returning my Voodoo3s to the same seller, do you mind if I ask what you said to them to get your money back?

Do you have a CRT you can test with to confirm your theory?

Unfortunately no, although a switch to a different LCD shows me a little more; it occasionally gets past the black screen, loads the windows background pattern and then immediately freezes with a few box shaped areas of random noise corruption on the background.

AaronAsh wrote:

Unfortunately no, although a switch to a different LCD shows me a little more; it occasionally gets past the black screen, loads the windows background pattern and then immediately freezes with a few box shaped areas of random noise corruption on the background.

I don't like the sound of that. 🙁

@OP.
At this point I suspect both bad 3DFX vid-cards and a bad motherboard. (Probably due to old caps.)
Unless they work in a different mobo I would return the cards while you still have the option. (Unless you just want to tinker with them.)

Can we get some pretty photos of YOUR mobo?
In particular I would like to see the caps and IC chips.
.

If I had not already blown through my "PC Play Money" allotment for this month I would buy one of combinationinc's bad cards just for the challenge of figuring out what their problem is.
.

AaronAsh wrote:

So TNT2 Pro arrived (specifically it's an Elsa Erazor III Pro). […]
Show full quote

So TNT2 Pro arrived (specifically it's an Elsa Erazor III Pro).

Good news! The system posts and gets into Windows with it.

Bad news! I seem to have run into either a K6 AGP problem or an issue with my monitor, as after I install the driver for the graphics card it only gets as far as the Windows splash screen then just goes to an unresponsive black screen.

Doing some reading this seems to be one of the many problems suffered by non-3Dfx AGP cards with the Ali Aladdin V chipset. Suggested solutions thus far have been BIOS tweaks, registry tweaks and installing the latest 1.9 AMD AGP driver, none of which have worked; I've also tried a variety of vintages of Detonator drivers, and the official Elsa drivers.

I actually suspect what might be happening is the card is loading into a refresh rate my LCD monitor cannot cope with, as I had a few issues like that with the S3 when changing resolution. If I start the system up in safe mode and try to change the resolution/depth options it says it is going to restart again to let me pick them in normal Windows, does so, but then only displays a really limited selection of options (640, 800 and 1024 for resolutions, locked to 16 colours at anything but 640) - selecting anything there then applying and restarting just ends up with me back at a black screen after the Windows splash screen. Is there some trick I should be using to get my LCD monitor to enumerate its modes correctly to Win98? Currently it is just using the standard PnP driver, though Windows does actually correctly read the monitor model from it.

Any other suggestions for possible AGP related issues greatly appreciated.

@bospressed As it looks like I might be returning my Voodoo3s to the same seller, do you mind if I ask what you said to them to get your money back?

3DFX AGP cards typically don't use as many AGP features, which is one of the reasons they have good AGP compatibility (less supported AGP features == fewer things that could go wrong because of this).

Since both V3 and TNT2 use more power than the AGP S3 Trio card (the AGP Trio didn't even have a heatsink iirc), could one of the problems here be an AGP slot which simply has problems supplying enough power to the TNT2 and V3?

Tetrium wrote:

Since both V3 and TNT2 use more power than the AGP S3 Trio card (the AGP Trio didn't even have a heatsink iirc), could one of the problems here be an AGP slot which simply has problems supplying enough power to the TNT2 and V3?

"Could it be...." - Yes. Possibly.
The Vcc for AGP slots is 3.3v and in most cases there is a 3.3v POL step-down regulator near the slot that provides it.
That regulator almost always has a significant cap associated with it.
If that cap is degraded the Vcc may be insufficient for full load or it may have issues with passing excessive noise. (Making some cards unstable.)
.

So I didn't have a chance to take any photos yesterday, but in the mean time here is a photo of my board from the Ebay auction - not the best quality but enough to see the layout clearly.
http://i.ebayimg.com/00/s/MTIwMFgxNjAw/z/fd8A … pWXV5O/$_57.JPG

As you say PCBONEZ, There is a suspicious looking capacitor right next to the AGP slot.

From that angle I don't see any obviously bad caps but like I (probably) said earlier not all bad caps show visual signs.
A couple near the CPU may be bloated (a little) but I can't really tell.
None of the major chips (MOSFETs, chipset and so forth) show any signs of overheating or taking a voltage spike.
From that angle it looks really good all around.

I don't see a VR dedicated to the AGP slot so that board probably takes 3.3v for AGP straight off the PSU and that cap (+ some SMD parts near it) provides some extra filtering before the slot. That approach may have been common with early AGP boards like this. I wasn't paying that much attention back then.
If that cap is bad/degraded the result would be the same with or without the VR.

I have my $2 in on Power Ball. (Which I don't usually do).
If I win the 1.5 billion dollars tomorrow I'll send you an ESR meter just so you can tell me what the ESR is on that cap is.
.

PCBONEZ wrote:

I don't see a VR dedicated to the AGP slot so that board probably takes 3.3v for AGP straight off the PSU and that cap (+ some SMD parts near it) provides some extra filtering before the slot. That approach may have been common with early AGP boards like this. I wasn't paying that much attention back then.
If that cap is bad/degraded the result would be the same with or without the VR.

This board can take AT PSUs, and and those do not provide 3.3V in any way... so unless that slot is entirely disabled when using AT PSUs (highly unlikely), then there must be a VR somewhere. I -HOPE- it is not just using that regulator near the RAM slots. Its interesting that, unlike many AT S7/SS7 motherboards I have seen (which, sadly, aren't that much) this seems to use a switching VR instead of a linear regulator - I couldn't find any big heatsinked regulator near the CPU. The easiest way to check if it even used the 3.3V line from the PSU is to get a multimeter and try a continuity test to the slot's pins.

alexanrs wrote:

PCBONEZ wrote:

I don't see a VR dedicated to the AGP slot so that board probably takes 3.3v for AGP straight off the PSU and that cap (+ some SMD parts near it) provides some extra filtering before the slot. That approach may have been common with early AGP boards like this. I wasn't paying that much attention back then.
If that cap is bad/degraded the result would be the same with or without the VR.

This board can take AT PSUs, and and those do not provide 3.3V in any way... so unless that slot is entirely disabled when using AT PSUs (highly unlikely), then there must be a VR somewhere. I -HOPE- it is not just using that regulator near the RAM slots. Its interesting that, unlike many AT S7/SS7 motherboards I have seen (which, sadly, aren't that much) this seems to use a switching VR instead of a linear regulator - I couldn't find any big heatsinked regulator near the CPU. The easiest way to check if it even used the 3.3V line from the PSU is to get a multimeter and try a continuity test to the slot's pins.

Yup, It's one of those cool AT-ATX powered boards.
The manual states it takes 3.3v(Std) or 5v(EDO) SDRAM so probably that regulator does handle 3.3v for both the RAM and the AGP slot.
There is nothing in the manual that says ATX power is needed for AGP.
.
Max amps for 3.3v on a 1x/2x AGP slot is 6 amps.
It's just under 5 watts for two sticks of SDRAM (I don't have data for modules smaller than 256Mb though), so ~1.5 amps or less.
Unless there are other 3.3v loads the regulator would only have to handle 7.5 amps.
.