yes, all components were in place.
a good news is that gddr3 has termination on die, there is no "hundreds of resistors" around the chip, like in case of ddr1

got this BFG7800GS from scrap.

with a little problem: does not start.

checking HSI bridge chip and some pci-e lines were lost + cracked smd capacitor in pci-e line.

HSI chip reflow and connection to pci-e has been restored.

[attachment=-1]7800-2.jpeg[/attachment]

Glad to see that card back in action. They're in the top of the line region for AGP. And it's always fascinating to see how different the approach of rialto vs. HSI design was back in the day.

Gainward Ti4600 - No signal when booting into Windows [FIXED] - A repair of a Geforce 4 Ti4600 - screen going blank under 3d and later on even under windows, underclocking did not yield an results. No visible damage. Recapping (especially) C1263, C1267 and C1271 solved the issue. Card is fully functional. Caps where looking okay, but one had severely degraded capacitance. Done by zuldan

got this blue FX5900 Turbo (and not so turbo) for repair.

missing VMEM, ripped memory pads...

but it is alive and working!

Nice work on those pads! Had a card once where one or two pads just went missing, fortunately they were NCs.

A successful repair of a Radeon 9700 Pro: ATI 9700 Pro Broken [FIXED]

Starting the card in a system directly prompted full screen artefacting, though it appears like a repeating pattern, indicating memory issues. After running R3MEMID in a automated fashion (so that the display issues don't matter) revealed errors on all four memory channels (MDA, MDB, MDC and MDD). This lead to the measurement of all power rails on the card, revealing that MVDDQ (used for memory I/O by both the memory ICs and the GPU) is not up to the desired voltage (1.13V instead of 2.8V). The rail resistance measured normal (card disconnected from the system), leading to the conclusion that the power delivery of the rail appears faulty. An attempt was made to investigate the ISL6522 is in overcurrent protection. During the process the dual FET of the rail got disconnected from the board. Replacing it fixed the power rail and fixed the whole card.

Repair done by zuldan

not a report but a conversion to pc version of ATI Radeon 9800XT Mac edition.

Found this piece another day on scrapyard:

tried to use in my Asus P3c-D -)))

Reflashed to x86 Radeon XT firmware and made a cut in AGP pro connector on the card to be able to fit regular Agp 8x.

An external 6-pin power connector has been attached to the board to provide external 12v power supply.

Unfortunately, only vga output is available.

Lower connector is a proprietary Apple ADC connector which is required a special ADC-DVI converter,
and the upper one in MAC version is using external SIL transceiver which is not available in PC card version bios.

I love that hacky MiniFit 6pin on that 9800. Nice conversion! 😁

not a report again but a conversion of mx4000 pci to fx5600

i have a couple of useless mx4000 pci cards an one day i though it will be nice to have unobtanium fx5600 pci.

swap the core, swap the memory, fix straps, fix nvdd... done.

added some fancy thermaltake heatsink and run 3dm benchmark.

Do I see that right that you moved the crystal to the back of the board? Neat little detail! Nice job!

Hi,
I think it was OK to add more "cold" cases here...
I'd like to add my first repair report. It's one for the beginners, but everyone has to start somewhere.
Sorry also for the picture quality, still a novice with that.

So this Radeon 9800 XL came to me end of last year.

Seller said it was working when removed from running system. Well, yeah, kind of...

R3MEMID (thank you! tehsiggy for bringing these tools up here, really a great help) showed MDC0 [0..31] fail.
According to this schema that corresponds to U56.
Diode testing around U56 revelead that VREF ball of U56 wasn't contacting the board. Pressure on the chip did change that and did also influence the picture (but not restore it entirely). So there's likely more loose balls underneath.
I suspected (more likely: hoped) the RAM chip itself would still be working. So the card was preheated and with some flux U56 got a little hot-air treatment.

After that, the card's working again.

R3MEMID showed no errors any more, 3DMark99/2000/2001/2003 were ok, played also half an hour Doom3 without any problems.
Should the errors come up again the plan is to remove the chip, reball & resolder. But this is another story I'll have to practice some more before. Till then I'll call it a day 😀

ChrisK wrote on 2026-02-16, 08:24:

Hi, I think it was OK to add more "cold" cases here... I'd like to add my first repair report. It's one for the beginners, but e […]
Show full quote

Hi,
I think it was OK to add more "cold" cases here...
I'd like to add my first repair report. It's one for the beginners, but everyone has to start somewhere.
Sorry also for the picture quality, still a novice with that.

Any report is welcome! Good job, glad to see that card back in working fashion! Any repair is a good one, even if it is not as complicated as others.

Hi there! Newbie with a soldering iron here (no hot air station yet, it's on the way though!). I recently purchased a PNY GF4 Ti4600 8x (The model shown on TPU's Ti4800 page). On the first boot, the image displayed, but had extremely severe ghosting and smearing horizontally. I did not manage to take a picture of it because I was concerned that something was going to explode if I left the card loaded for any span of time. Closer visual inspection (which in retrospect I should have done more closely upon receipt) revealed a bunch of missing capacitors on the back.

In my first attempt at repair, measured some adjacent capacitors and then purchased some 0402 mlccs from Mouser. I am pretty confident in the surrounding capacitor values, but for the vcore I wound up using a mix of 200pf and 2,000pF capacitors (because I did not buy enough 200pF caps for a first time mlcc soldering experience and did not realize how easily you can lose those things to the void 😅). In a stroke of dumb luck, the result was actually pretty good and practically identical to my third and current result (see below). Which is to say, useable but not great.

In my second attempt I tried to use some more "ideal" from a |Z| frequency response perspective (300mHZ) 680pF capacitors across the board, which did not work at all. 🙃

At this point I was getting frustrated so I measured some more caps and noticed that there are 470pF values (dark brown) mixed in. This gave me the idea of trying to color match mlccs using backside high res macro shots from other makes of this sku, but even the color patterns did not match, so that was a bust. I settled for swapping for a mix of 200pF, 470pF, and a couple of 0.022nF capacitors to bulk up the current delivery. This brings me to where I am now.

There is faint but noticeable smearing and ghosting on the right side of icons and text (identical on both vga and dvi-i (dvi as analog output) outputs); looking closely, you'll notice faint doubled offset button/ui elements. I'm 99% sure this is to do with the card because a Ti4200 that I also have shows zero such artifacts in the same pc with the same cables etc. As an aside, I also replaced all electrolytics with polymers as a desperation play to no avail.

So yeah... this is kind of ok, but not where I want to stop. I now have a whole booklet/binder of mlccs to throw at this problem, but I don't have access to a schematic, board view, or any other reference resource to go on. If anybody has such information, or information on Quadro 980 XGLs (which seem to be visually practically identical), please let me know! If it might be a software/driver related issue, please feel free to point me in a direction of inquiry! I feel like I'm so close to a good result but I don't think the components like being used as game pieces in high-temperature combinatorics guessing game. 😣

JammyPajammies wrote on 2026-02-18, 07:45:

So yeah... this is kind of ok, but not where I want to stop. I now have a whole booklet/binder of mlccs to throw at this problem, but I don't have access to a schematic, board view, or any other reference resource to go on. If anybody has such information, or information on Quadro 980 XGLs (which seem to be visually practically identical), please let me know! If it might be a software/driver related issue, please feel free to point me in a direction of inquiry! I feel like I'm so close to a good result but I don't think the components like being used as game pieces in high-temperature combinatorics guessing game. 😣

Keep it up! Regarding the schematic, there you go!

I wonder if there is any issue with memory, as the pattern with the colors in your first picture is very repetitive.. you could try to find a version of MATS for that cart RAM chips layout on Quadro 980 XGL

JammyPajammies wrote on 2026-02-18, 07:45:

Hi there! Newbie with a soldering iron here (no hot air station yet, it's on the way though!). I recently purchased a PNY GF4 Ti […]
Show full quote

Hi there! Newbie with a soldering iron here (no hot air station yet, it's on the way though!). I recently purchased a PNY GF4 Ti4600 8x (The model shown on TPU's Ti4800 page). On the first boot, the image displayed, but had extremely severe ghosting and smearing horizontally. I did not manage to take a picture of it because I was concerned that something was going to explode if I left the card loaded for any span of time. Closer visual inspection (which in retrospect I should have done more closely upon receipt) revealed a bunch of missing capacitors on the back.

The attachment asReceived.jpg is no longer available

In my first attempt at repair, measured some adjacent capacitors and then purchased some 0402 mlccs from Mouser. I am pretty confident in the surrounding capacitor values, but for the vcore I wound up using a mix of 200pf and 2,000pF capacitors (because I did not buy enough 200pF caps for a first time mlcc soldering experience and did not realize how easily you can lose those things to the void 😅). In a stroke of dumb luck, the result was actually pretty good and practically identical to my third and current result (see below). Which is to say, useable but not great.

In my second attempt I tried to use some more "ideal" from a |Z| frequency response perspective (300mHZ) 680pF capacitors across the board, which did not work at all. 🙃

The attachment badCapValues1.jpg is no longer available

At this point I was getting frustrated so I measured some more caps and noticed that there are 470pF values (dark brown) mixed in. This gave me the idea of trying to color match mlccs using backside high res macro shots from other makes of this sku, but even the color patterns did not match, so that was a bust. I settled for swapping for a mix of 200pF, 470pF, and a couple of 0.022nF capacitors to bulk up the current delivery. This brings me to where I am now.

The attachment currentAttempt.jpg is no longer available

The attachment currentSmearingGhosting.jpg is no longer available

There is faint but noticeable smearing and ghosting on the right side of icons and text (identical on both vga and dvi-i (dvi as analog output) outputs); looking closely, you'll notice faint doubled offset button/ui elements. I'm 99% sure this is to do with the card because a Ti4200 that I also have shows zero such artifacts in the same pc with the same cables etc. As an aside, I also replaced all electrolytics with polymers as a desperation play to no avail.

So yeah... this is kind of ok, but not where I want to stop. I now have a whole booklet/binder of mlccs to throw at this problem, but I don't have access to a schematic, board view, or any other reference resource to go on. If anybody has such information, or information on Quadro 980 XGLs (which seem to be visually practically identical), please let me know! If it might be a software/driver related issue, please feel free to point me in a direction of inquiry! I feel like I'm so close to a good result but I don't think the components like being used as game pieces in high-temperature combinatorics guessing game. 😣

What capacity value did you use for the bigger 0805 ones?
They should be 4.7uF. I had to replace some of those on a Ti4200 recently.
The schematic tehsiggy posted seems to support that.

Another day, another repair:
Are X700 Radeons prone to die? [solved]

ChrisK wrote on 2026-02-18, 10:32:

Another day, another repair:
Are X700 Radeons prone to die? [solved]

Have had that same card (and a second one) for years, even in my main rig. It got pretty abused but never failed me, glad you got it working again. Somewhere I have a schematic from Gigabyte laying around for RV410 + Rialto, if you ever have the need for it.

tehsiggi wrote on 2026-02-18, 12:12:

ChrisK wrote on 2026-02-18, 10:32:

Another day, another repair:
Are X700 Radeons prone to die? [solved]

Somewhere I have a schematic from Gigabyte laying around for RV410 + Rialto, if you ever have the need for it.

That would be great if you could have a look for that!
I puzzled over this with a mix of RV370 and RV410 schematics, but all were PCI-E without the Rialto bridge.
The card is very tightly packed with the video core and the bridge chip placed almost back to back leaving only very little to no possibilities to probe any signals between them.
In the end it was a 1:1 chance beeing the bridge or the video core beeing the problem. Seems I was lucky to try the bridge first 😀

ChrisK wrote on 2026-02-18, 12:37:

That would be great if you could have a look for that! I puzzled over this with a mix of RV370 and RV410 schematics, but all wer […]
Show full quote

tehsiggi wrote on 2026-02-18, 12:12:

ChrisK wrote on 2026-02-18, 10:32:

Another day, another repair:
Are X700 Radeons prone to die? [solved]

Somewhere I have a schematic from Gigabyte laying around for RV410 + Rialto, if you ever have the need for it.

That would be great if you could have a look for that!
I puzzled over this with a mix of RV370 and RV410 schematics, but all were PCI-E without the Rialto bridge.
The card is very tightly packed with the video core and the bridge chip placed almost back to back leaving only very little to no possibilities to probe any signals between them.
In the end it was a 1:1 chance beeing the bridge or the video core beeing the problem. Seems I was lucky to try the bridge first 😀

There you go, found it 😀