Out of curiosity (since I'm not sure the BIOS is being picked up by the system, I assume that is how it detects if the card is present or not?), I looked at the VDD input to the 39VF512 ROM on the board. According to the datasheet should be 2.7-3.6 volts.

Looks like it's fluctuating between 3 and 3.4 volts. Seems okay?

Also looked at CE# and OE# lines just to see

CE#

OE#

The only curious thing there is you can see where the voltage where it is cycling on and off towards the right swinging past the 3.2V (ish) range and also going below 0V.

Guld wrote on 2025-03-20, 01:13:

Looks like it's fluctuating between 3 and 3.4 volts. Seems okay?

It depends.
If the manufacturer saved money and ran 3.3V directly from the connector, then it's so-so.
If there is a 1117 regulator with 5V to 3.3, then it's a lot.
It should be somewhere around 25mV peak-to-peak.

When you turn on the video card, the AGP mode (pin Detect) and logical levels are immediately determined.This is done on resistor dividers (and transistors for AGP4x-8X).
In parallel with this, the voltage regulators are started.
After this, the Power good signal from the regulators is compared with the Reset from the AGP.
However, Power good may not be processed.
If there are no power failures, then the reset is removed from the video card and the VBIOS is initialized.
The VBIOS is read by the CPU (the GPU here is just a bridge to the BIOS chip) via AGP, so by this time the card should be accessible via the interface.

Thanks Shevalier, I'm not all that up on how AGP works (at all really 😀 ), so I appreciate the info!

Today I compared my good and bad boards to try to make sure the caps I redid didn't have bad connectors or similar.

I was booting my AB-BX6 board up and was going to check some voltages to compare with the card on the board it works on....and.....the stupid thing decided to boot up like nothing was wrong. Got video, looked stable, etc. The CMOS was completely reset as I replaced the CMOS battery in some of my testing. However, that doesn't appear to be the fix (as mentioned above it was recommended to reset BIOS to safest/most stable values). I tried it in my BX-133 Raid system (P3 1 GHz)...and it also booted without issue!

Obviously I'm not confident, but I'm also not sure why it is suddenly working. All I've done since last time is check the capacitors with my multi-meter, and wiped off a little thermal grease that I got near the edge connector in the process (it was not there during previous testing)....so either my meter probes "cleaned" something just enough to work (seems unlikely), me cleaning near the connector was enough to clean something off the edge that was causing issues (also doesn't seem likely but....possible), or possibly a trace is marginal/damaged and is playing with me and just happens to work right now.

Will report back after some more testing.

But here is my test matrix with all my motherboards, which unfortunately doesn't really lend to any clues that I see:

1Combo-Z               AGP 1.5V          Works with both 5V 20A and 36A
2K8T Neo               AGP 1.5V          Does not work with either PSU
3AB-BX6 (P3-450)       AGP 3.3V          Does not work with either PSU
4BX-133 Raid (P3-1GHz) AGP 3.3V          Does not work with either PSU (tried additional load of 2 hard drives, no change)
5KT-7 Raid             AGP Universal     Does not work with either PSU

Probably not related to your problem but maybe worth a trying checking the same BIOS option with your K8T Neo

Re: MSI K8T Neo / VIA K8T800 not posting with a newer AGP card

I had an X800 PCI-e, which was detected only on one motherboard when cold.
After heating - nowhere.
I replaced all 1117 and TL431 on this card, and all discrete logic.
The card worked fine only after replacing the regulator for PCI-e terminator.
But only for a couple of months, then even replacing this regulator did not help.
Apparently, the GPU was partially damaged.

Your GPU may be in the same state.

I'm starting to wonder about the caps that I replaced on the board.

So I had replaced the through hold parts with a new part...but what I didn't do is check the ESR of the old part first. Turns out with my replacment part (ESR ~34 mOhm), the board has issues running any tests. 2D is okay, but 3D locks up and reboots fairly quickly (similar to the problem I had early on). The old parts measure ESR ~10-12 mOhm. Put them back in...and those problems go away....at least at lower resolutions. Still have similar issues past about 1600x1050 (tried with both PSU, so not having sufficient 5V current available as mentioned above does not appear to be the issue).

Board is still working on the P3 systems, so I still have no good explanation for why that suddenly started working.

So now I have to ask, when I replaced the surface mount caps, maybe I did not put in ones with sufficiently low ESR?

Can anyone help identify if any of the caps on the picture are low ESR? This is a picture of the original, unmodified board.

Parts are:
2 x 1500 uF 6V
2 x 1000 uF 6V
1 x 470 uF 6V
2 x 330 uF 6V
1 x 82 uF 16 V

I had replaced with the following (ESRs listed where I could find them, for some reason the Panasonic S series don't list them on the data sheet?!)

1EEE-0JA152P  1500 uF 6.3V ? (Panasonic Series S)
2EEE-FN0J102P 1000 uF 6.3V 160 mOhm
3EEE-0JA471P   470 uF 6.3V ? (Panasonic Series S)
4EEE-0JA331XP  330 uF 6.3V ? (Panasonic Series S)
516SVP82M       82 uF 16V  40 mOhm

The good news is that it's at least usable up to about 1600x900 as best I can tell although that resolution seems about on the edge of when it has issues. So I wonder if it could be as simple as I picked caps with too high an ESR?

zuldan wrote on 2025-03-22, 20:58:

Probably not related to your problem but maybe worth a trying checking the same BIOS option with your K8T Neo

Re: MSI K8T Neo / VIA K8T800 not posting with a newer AGP card

Yeah, I saw that option, I'll take a look the next time I have that board out.

shevalier wrote on 2025-03-23, 12:54:

I had an X800 PCI-e, which was detected only on one motherboard when cold. After heating - nowhere. I replaced all 1117 and TL43 […]
Show full quote

I had an X800 PCI-e, which was detected only on one motherboard when cold.
After heating - nowhere.
I replaced all 1117 and TL431 on this card, and all discrete logic.
The card worked fine only after replacing the regulator for PCI-e terminator.
But only for a couple of months, then even replacing this regulator did not help.
Apparently, the GPU was partially damaged.

Your GPU may be in the same state.

Will keep an eye out, but cold it seems to be working on everything again (although I really wish I knew why it went from working -> not working -> working on the P3 systems?!).
Just ran it for a while on the Combo-Z board and switch it over to the P3 board while still warm and it still seems happy to work there again. So doesn't appear to be temp related.....but I'll try to keep more notes and see if I notice any trends. I hadn't been noting board warm/cold when I run tests.

Guld wrote on 2025-03-23, 18:56:

I'm starting to wonder about the caps that I replaced on the board.

That's why I always say - dudes, don't waste your time, just replace the capacitors to polymer ones right away.

Unfortunately, the FX5700 came to me with a cracked chip, so it died soon after. 🙁

Guld wrote on 2025-03-23, 18:56:

I'm starting to wonder about the caps that I replaced on the board. […]
Show full quote

I'm starting to wonder about the caps that I replaced on the board.

So I had replaced the through hold parts with a new part...but what I didn't do is check the ESR of the old part first. Turns out with my replacment part (ESR ~34 mOhm), the board has issues running any tests. 2D is okay, but 3D locks up and reboots fairly quickly (similar to the problem I had early on). The old parts measure ESR ~10-12 mOhm. Put them back in...and those problems go away....at least at lower resolutions. Still have similar issues past about 1600x1050 (tried with both PSU, so not having sufficient 5V current available as mentioned above does not appear to be the issue).

Board is still working on the P3 systems, so I still have no good explanation for why that suddenly started working.

So now I have to ask, when I replaced the surface mount caps, maybe I did not put in ones with sufficiently low ESR?

Can anyone help identify if any of the caps on the picture are low ESR? This is a picture of the original, unmodified board.

The attachment Ti4200 4x.jpg is no longer available

Parts are:
2 x 1500 uF 6V
2 x 1000 uF 6V
1 x 470 uF 6V
2 x 330 uF 6V
1 x 82 uF 16 V

I had replaced with the following (ESRs listed where I could find them, for some reason the Panasonic S series don't list them on the data sheet?!)

Copy code to clipboard

1EEE-0JA152P  1500 uF 6.3V ? (Panasonic Series S)
2EEE-FN0J102P 1000 uF 6.3V 160 mOhm
3EEE-0JA471P   470 uF 6.3V ? (Panasonic Series S)
4EEE-0JA331XP  330 uF 6.3V ? (Panasonic Series S)
516SVP82M       82 uF 16V  40 mOhm

The good news is that it's at least usable up to about 1600x900 as best I can tell although that resolution seems about on the edge of when it has issues. So I wonder if it could be as simple as I picked caps with too high an ESR?

I have a similar issue with my Ti4600 Gainward Ti4600 - No signal when booting into Windows. I've just worked out that the card will only run in 640 x 480. As soon as I increase the resolution I get a black screen. Need to reboot. I'm wondering if I need to replace all the capacitors (with low ESR ones).

Okay, finally got low ESR caps and recapped the entire board (except for a few that tested okay....but will likely try them next anyway).

And it does generally seem to run longer before encountering an issue...but not always...Also got set up with the same benchmark (3DMark03) running on my BX-133 system and it died just 2 minutes into the same test 😁. What's interesting is that it fails in the exact same way.

Here is what I see on the screen when it does. And it always seems to occur when the screen resolution is changing to set up for one of the benchmarks.

This is what I see for a bit of time just before the system reboots.

I'll try replacing the few caps that are left that tested good last time I had them out with some low ESR ones I also have but not optimistic that will change anything since the others seemed to have low loss %, low ESR, and still at rated capacitance...but...never know.

So, good news! It's fully working now and had run for multiple hours on multiple machines with 3DMark03 with no issues!

I did end up replacing the 4 through hole 510 uF 4V capacitors with new ones. Closest I could find was 510 uF 16V. It's possible that they were the culprits all along, but I'm not 100% sure, I did replace them at one point but it wasn't low ESR and made no difference before replacing the surface mount. The caps seemed to test okay, but apparently were just problematic enough to cause issues.

In case it helps anyone, here are the parts I installed in place of the old caps:

1 x 82uF 16V low ESR Aluminum Organic Polymer SMT- A766KE826M1CLAE030 1 x 470uF 6.3V low ESR Aluminum Organic Polymer SMT- A765K […]
Show full quote

1 x 82uF 16V low ESR Aluminum Organic Polymer SMT- A766KE826M1CLAE030
1 x 470uF 6.3V low ESR Aluminum Organic Polymer SMT- A765KH477M0JLAE016
2 x 1500uF 6.3V low ESR Aluminum Organic Polymer SMT - A765MU158M0JLAE010
2 x 330uF 6.3V low ESR Aluminum Organic Polymer SMT - A765EG337M0JLAE017
2 x 1000uF 6.3V low ESR Aluminum Organic Polymer SMT - APA0812102M006R
4 x 510uF 16V low ESR Aluminum Organic Polymer radial, through hole - EEU-FS1C511 [note: only need 4V on these but 16 was the lowest voltage I could find in low ESR]

Guld wrote on 2025-03-23, 18:56:

So now I have to ask, when I replaced the surface mount caps, maybe I did not put in ones with sufficiently low ESR?

since nobody bothered to point this out: yes, those s-series caps are a general purpose type, which you can already see by the maximum ripple current being listed for 120 hz as opposed to 100 khz, and the ESR not being listed. definitely not meant for something like this. they are also just 85°C rated parts, not 105°C.

of the original caps, the purple ones are polymers, so definitely low ESR, as should be anything around regulators, at the minimum. however if by replacing 4 thru-holes you mean the four purple caps on the top left, those are polys that could have been going bad - at least i seem to remember reading on badcaps that this is the case for certain types of polys from the time. so if swapping those changed something, that is one possible explanation. what are you using to test those capacitors? a bit odd that it works after (properly) changing the caps now despite none of the original ones showing any out of spec values.

auron wrote on 2025-04-15, 00:34:

Guld wrote on 2025-03-23, 18:56:

So now I have to ask, when I replaced the surface mount caps, maybe I did not put in ones with sufficiently low ESR?

since nobody bothered to point this out: yes, those s-series caps are a general purpose type, which you can already see by the maximum ripple current being listed for 120 hz as opposed to 100 khz, and the ESR not being listed. definitely not meant for something like this. they are also just 85°C rated parts, not 105°C.

of the original caps, the purple ones are polymers, so definitely low ESR, as should be anything around regulators, at the minimum. however if by replacing 4 thru-holes you mean the four purple caps on the top left, those are polys that could have been going bad - at least i seem to remember reading on badcaps that this is the case for certain types of polys from the time. so if swapping those changed something, that is one possible explanation. what are you using to test those capacitors? a bit odd that it works after (properly) changing the caps now despite none of the original ones showing any out of spec values.

Thanks for the additional info. Still slowly learning how to identify caps. I had replaced the surface mount with low ESR thinking that was the issue (when I thought the through-hole were still good). It's entirely possible that replacing the 4 through hole purple ones would have been sufficient.

The surface mount purple one (possibly polymer by your notes) is the 82uF 16V, I had wondered why it was a different color.

The purple through hole are Sanyo 510 uF 4V. In general I've heard and experienced that the Sanyo caps tend to hold up quite well (mostly).

For testing, I've been using a GM328 transistor tester/meter. Relatively cheap but has seemed to work for me for the most part. Open to other suggestions of course.

I want to say the old ones showed ESR around 0.12-0.14 Ohm which didn't register as being that high to me...although for low ESR that could be high? The new ones are 0.056 Ohm. Could just be they were all out of spec. I would have expected some more variance though but am still learning, so wouldn't be at all surprised if I was incorrect 😀.

the purple ones are all sanyo OS-CON, panasonic makes those now. they are generally well regarded indeed but there is the odd report of them failing/aging, so it's not clear to me what was the case with your card, someone else may offer a better guess.

for the wet SMD caps, while i don't know what brand or line that is, you could have referenced some low ESR SMD series of same size/capacitance/voltage rating to compare the ESR. for GP caps, there are charts out there on expected ESR, though SMD caps usually have somewhat worse ratings for ESR and durability.