Thanks for the info, tehsiggi! If I manage to fix the problematic Asus card, I'll order just one chineese orange ("golden") cooler for it. Tomorrow I'll get all the new replacement capacitors I need.

Fingers crossed for a successful repair! Looking forward to the result!

This is the current state of the card - all Hitano junk removed:

Tomorrow I'll report back if there is any change for good after soldering the new caps.

It's been a while, but I had tinkering with other retro PC parts. First time used polymer capacitors (some china made ULR - whatever this means, and these are not cheap), but they're within +/-20% limits for their capacity. Samxon electrolytic caps are pretty good - spot on with their labeled capacity.

Here it is how the card looks after the full recap:

Unfortunately the "sudden death" with black screen and system hang did not disappear, but... Maybe there's another problem, 'cause currently I have no 100% proper PSU for powering the used s. A board (Athlon XP CPU) for tests. The FSP PSU (shevalier gave a clear definition why it is not appropriate for other than Athlon 64 systems) I'm using just don't work well without enough load on its +12 VDC line, which s. A or older system can't provide. The other (period correct) option I have is a "300 W" (probably ~ 180-200 W peak power) Codegen junk, which gives even more unstable voltages (100% needs full recap, but I doubt it's worth the price). The interesting moment here is, that my other two R9550 videocards (GeCube and Gigabyte) somehow works fine with the test system powered by improper FSP PSU, but as I mentioned before - their PCB design is totally different and they have much more soldered electronic components compared to this Asus card.

I'll retest the card, when get some proper (period correct) ATX PSU for s.A Athlon XP system without extra +12 VDC CPU connectors.

Thanks to shevalier and tehsiggi for all the help!

analog_programmer wrote on 2025-07-15, 16:35:

s.A Athlon XP system without extra +12 VDC CPU connectors.

I would recommend repeating the same procedure with the motherboard.
The peculiarity of the Athlon XP is that it is a processor with modern functions (Bus Disconnect is an early implementation of power saving, later developed into Cool and Quaet) and significant power consumption.
This processor can already create a very unpleasant load with large differences. Of course, it is not Ryzen (10 W at rest, 150 W under load), but it is not Pentium 3 either, which changes the load by 20 percent.

shevalier wrote on 2025-07-16, 04:53:

I would recommend repeating the same procedure with the motherboard.
The peculiarity of the Athlon XP is that it is a processor with modern functions (Bus Disconnect is an early implementation of power saving, later developed into Cool and Quaet) and significant power consumption.
This processor can already create a very unpleasant load with large differences. Of course, it is not Ryzen (10 W at rest, 150 W under load), but it is not Pentium 3 either, which changes the load by 20 percent.

Yep, I'll retest the entire system, when I get a proper working PSU for s.A motherboards. This FSP ATX-350PNF PSU seems not bad, but now I'm sure it really doesn't suit something older (or newer) than s.AM2 motherboard + Athlon 64 (X2) CPU. The "300 W" Codegen 200x ("200x" model rather suggests а 200 W peak power, which is also questionable) PSU junk, that I have and suits, is just a total junk. I have a slightly less junky JNC LC-B250ATX PSU, which is a period correct match for these s.A motherboards, but it is in semi-disassembled state and I still can't find a proper circuit diagram for it. I think it can be restored and even improved to become a descent test-bench PSU, but again... I can't do much on my own without a circuit diagram for it 🙁 I'm considering opening a thread for helping me restoring it.

Ah, I'm too late to the party again, as always.

But a couple of things worth mentioning for any future readers.

First and foremost: any hardware that functions OK when "cold" but starts to have issues when "hot' is absolutely NOT an electrolytic capacitor-related issue.
If anything, the ESR/impedance of electrolytic capacitors improves with heat - sometimes very significantly. And that should not be a surprise. Most chemical reactions speed up with temperature increases. In the case of the electrolyte in the capacitors, it becomes more "active", to put it more simply. And you can see this on datasheets of low ESR capacitors too. Usually they list two values for the impedance - one at -10C and one at 20C. The impedance often becomes 2-4x worse at -10C than at 20C. So at higher temperatures, you can expect the ESR to improve even further (though possibly with not such great amount, as the impedance-to-temperature relation is hardly a linear one.)

That said, it's good that you changed those Hitano caps, as they indeed appear to be failing.
However, reading through the thread, I was almost sure that your new caps would not solve your issue, despite hoping for it myself.

The issue with these passively-cooled Radeon video cards is that they *can* run too hot under load, causing the GPU chip to fail in some cases. The reason I say "can" is because these GPUs are fairly tough and normally do last OK even with stock passive cooler. However, when installed in a case with not-so-great cooling or when used in rooms with non air-conditioned environments (common where you and I live in Bulgaria), that can cause the GPU to overheat and cool down repeatedly over time... eventually causing damage.

IMO, these cards should really have had a small slow-turning fan installed from the factory (e.g. a 50 or 60 mm 12V fan constantly operated at 5V). This would have eliminated the issue of some of these cards running hot in a poorly-ventilated cases or at elevated room temperatures. But then again, this era of video cards wasn't know for well though-of cooling solutions... and the RV360 being a "budget" mid-range card, it shouldn't be a surprise that it is what it is.

Since yours still appears to work "OK" when cold, the only "solution" for it at this point is to just try to keep it as cool as possible (well, within reason anyways - no need to go for crazy stuff, like water cooling or some chunky huge cooler.) IMO, you can opt to keep the stock heatsink. Just install a 60 mm fan on top and run it on something like 7V to keep it more quiet while still providing sufficient airflow... though that will depend on the fan you use. Some more powerful fans will provide sufficient airflow even at 5v.

I tend to have more 80 mm fans, so I usually just "side-mount" a standard-power 80 mm fan next to these cards and run it at 7V. This is sufficient to keep the heatsink "just warm" (around 45-50C) when the card is under full load, compared to "too hot to put your finger on it" (55-62C, IIRC) without the fan.
FWIW, I also have the same ASUS A9550GE card as yours, except mine came stock with Panasonic (FJ or FJS??) caps, which I'm not surprised about, as ASUS did try to stick with Japanese only capacitors back in those days. Unfortunately, Hitano are probably the only Japanese not worth a damn... if they are even Japanese at all (they claim to be... but who knows really 😀 ). So unfortunately, you drew the short straw from ASUS in regard to the caps.

And I also have a Gigabyte Radeon 9550 like yours too. For me, it's that one that has issues sometimes. Well, it works fine with everything I threw at it, except one game (I forget which one) where it showed very very slight artifacts that really can't be noticed at all most of the time (slight yellowing of some grays.) I checked all of the RAM termination resistors, and all is OK there, as are all of the RAM pins. It likely is failing the same way as your ASUS - just probably ran too hot too many times in its life. Meanwhile, my ASUS 9550GE is running absolutely fine, as is a Sapphire 9600 Pro I acquired recently. I also got a Gigabyte Radeon 9600 Pro a little later, and that one has one bad RAM termination resistor, resulting in constant slight artifacts in Windows (again, just yellowing in the grays at certain patterns on the screen.) In games, though, it's comical - jumping triangles and vertices along with smeared textures everywhere... but it doesn't crash. I ran a full race in Collin McRae Rally 04 with it as a challenge... and despite knowing the whole map/rally route, I still had a very hard time navigating/driving. 🤣

momaka wrote on 2025-07-18, 07:00:

Ah, I'm too late to the party again, as always.

But a couple of things worth mentioning for any future readers.

No worries, it's always a good thing to have some more information. Thanks for your input!

momaka wrote on 2025-07-18, 07:00:

First and foremost: any hardware that functions OK when "cold" but starts to have issues when "hot' is absolutely NOT an electro […]
Show full quote

First and foremost: any hardware that functions OK when "cold" but starts to have issues when "hot' is absolutely NOT an electrolytic capacitor-related issue.
If anything, the ESR/impedance of electrolytic capacitors improves with heat - sometimes very significantly. And that should not be a surprise. Most chemical reactions speed up with temperature increases. In the case of the electrolyte in the capacitors, it becomes more "active", to put it more simply. And you can see this on datasheets of low ESR capacitors too. Usually they list two values for the impedance - one at -10C and one at 20C. The impedance often becomes 2-4x worse at -10C than at 20C. So at higher temperatures, you can expect the ESR to improve even further (though possibly with not such great amount, as the impedance-to-temperature relation is hardly a linear one.)

That said, it's good that you changed those Hitano caps, as they indeed appear to be failing.
However, reading through the thread, I was almost sure that your new caps would not solve your issue, despite hoping for it myself.

I'm not into electronics theory, so this ESR value change (in unexpected way) according to operating temperature is something new for me. Anyway, it was necessary to replace all of these Hitano c(r)aps.

momaka wrote on 2025-07-18, 07:00:

The issue with these passively-cooled Radeon video cards is that they *can* run too hot under load, causing the GPU chip to fail in some cases. The reason I say "can" is because these GPUs are fairly tough and normally do last OK even with stock passive cooler. However, when installed in a case with not-so-great cooling or when used in rooms with non air-conditioned environments (common where you and I live in Bulgaria), that can cause the GPU to overheat and cool down repeatedly over time... eventually causing damage.

IMO, these cards should really have had a small slow-turning fan installed from the factory (e.g. a 50 or 60 mm 12V fan constantly operated at 5V). This would have eliminated the issue of some of these cards running hot in a poorly-ventilated cases or at elevated room temperatures. But then again, this era of video cards wasn't know for well though-of cooling solutions... and the RV360 being a "budget" mid-range card, it shouldn't be a surprise that it is what it is.

My next suggestion is for broken BGA GPU connection due to continuous overheating.

momaka wrote on 2025-07-18, 07:00:

Since yours still appears to work "OK" when cold, the only "solution" for it at this point is to just try to keep it as cool as possible (well, within reason anyways - no need to go for crazy stuff, like water cooling or some chunky huge cooler.) IMO, you can opt to keep the stock heatsink. Just install a 60 mm fan on top and run it on something like 7V to keep it more quiet while still providing sufficient airflow... though that will depend on the fan you use. Some more powerful fans will provide sufficient airflow even at 5v.

I tend to have more 80 mm fans, so I usually just "side-mount" a standard-power 80 mm fan next to these cards and run it at 7V. This is sufficient to keep the heatsink "just warm" (around 45-50C) when the card is under full load, compared to "too hot to put your finger on it" (55-62C, IIRC) without the fan.

Just adding a fan won't help. The problems begin when the card and GPU's heatsink become warm, not too hot.

momaka wrote on 2025-07-18, 07:00:

FWIW, I also have the same ASUS A9550GE card as yours, except mine came stock with Panasonic (FJ or FJS??) caps, which I'm not surprised about, as ASUS did try to stick with Japanese only capacitors back in those days. Unfortunately, Hitano are probably the only Japanese not worth a damn... if they are even Japanese at all (they claim to be... but who knows really 😀 ). So unfortunately, you drew the short straw from ASUS in regard to the caps.

As far as i know Hitano c(r)apacitors are chineese (not even taiwanese) despite their japanese sounding name.

momaka wrote on 2025-07-18, 07:00:

And I also have a Gigabyte Radeon 9550 like yours too. For me, it's that one that has issues sometimes. Well, it works fine with everything I threw at it, except one game (I forget which one) where it showed very very slight artifacts that really can't be noticed at all most of the time (slight yellowing of some grays.) I checked all of the RAM termination resistors, and all is OK there, as are all of the RAM pins. It likely is failing the same way as your ASUS - just probably ran too hot too many times in its life. Meanwhile, my ASUS 9550GE is running absolutely fine, as is a Sapphire 9600 Pro I acquired recently. I also got a Gigabyte Radeon 9600 Pro a little later, and that one has one bad RAM termination resistor, resulting in constant slight artifacts in Windows (again, just yellowing in the grays at certain patterns on the screen.) In games, though, it's comical - jumping triangles and vertices along with smeared textures everywhere... but it doesn't crash. I ran a full race in Collin McRae Rally 04 with it as a challenge... and despite knowing the whole map/rally route, I still had a very hard time navigating/driving. 🤣

Unfortunately Gigabyte card doesn't have any fan connector terminals. For now it runs fine.

This Asus A9550GE/TD/256M/A card has no value for me. Reballing is not for me, but I can try BGA GPU connections "resoldering" with hot air gun and if GPU doesn't survive, RAM chips and other elements will be Ok for spare parts.

As for my old and problematic PSUs, I really have to restore one ATX unit for older systems (with PIII or Athlon (XP) motherboards) testing. I managed to find two very similar circuit diagrams for LC-B250ATX PSU, which I think is ATX ver.1.x standard. This weekend I'll start to reassemble it. I also have one ATX ver. 2.1 standard AT-2005B PSU, which is in even worse state (I tink I keep all of it's main components) and I can't remember which MOSFET or double diode where goes on the PCB, but for now I'm not in need to restore it.

momaka wrote on 2025-07-18, 07:00:

First and foremost: any hardware that functions OK when "cold" but starts to have issues when "hot' is absolutely NOT an electrolytic capacitor-related issue.
If anything, the ESR/impedance of electrolytic capacitors improves with heat - sometimes very significantly.

This is while the capacitors at least somewhat correspond to the datasheet.
If the capacitor ESR has increased 10 times (from 40 mOhm to 0.4 Ohm), then no amount of heating will help it.
And the rest of the circuit will operate outside the operating mode.
For the same MOSFETs, the transconductance decreases with increasing temperature.

As for checking the resistors, it's a good idea unless they are 0402. Then you either need to be young or have a microscope.
But being young is better 😀

shevalier wrote on 2025-07-18, 11:27:

This is while the capacitors at least somewhat correspond to the datasheet. If the capacitor ESR has increased 10 times (from 40 […]
Show full quote

This is while the capacitors at least somewhat correspond to the datasheet.
If the capacitor ESR has increased 10 times (from 40 mOhm to 0.4 Ohm), then no amount of heating will help it.
And the rest of the circuit will operate outside the operating mode.
For the same MOSFETs, the transconductance decreases with increasing temperature.

This is obvious, but worth mentioning for people who'll gonna ask questions like "how do I measure voltage on this "thing"?".

shevalier wrote on 2025-07-18, 11:27:

As for checking the resistors, it's a good idea unless they are 0402. Then you either need to be young or have a microscope.
But being young is better 😀

I've lost my super sharp sight at close "reading" distance when I passed 40. Still I'm not in need to use a dioptric glasses for reading, but I have to use a magnifying glass for very small things/text, which is pretty normal for mid-age (mature) and older people 🚫 👓 🚫 Ussually being young is physically better, but from wisdom and experience perspective bring young/immature is worse, unless you are an "ordinary/average person, who doesn't constantly learn new things every day 😉. Enough of lyrical digressions 😁

What resistors I have to check with microscope (I have a USB-chineese-thing called "digital microscope" that's better than magnifying glass)?

P.S. Anyway, my next step is to restore my less junky ATX v.1.x 200 W PSU, just to be sure that the culprit is not in the backbone of the entire system - the PSU. And if this doesn't help, I can start fiddling with the GPU voltage for more stability... The second to last step will be to use my hot air gun for GPU "reflow". The final step is clear - the card goes in the box labeled "for spare parts" 😀

analog_programmer wrote on 2025-07-18, 11:53:

What resistors I have to check with microscope (I have a USB-chineese-thing called "digital microscope" that's better than magnifying glass)?

in red square.
in green - no sense
There's nothing to check on this card.
Asus didn't bother with impedance matching, so there are simply no resistors.

A more advanced design on the same RV360 looks like this.
https://theretroweb.com/expansioncards/s/sapp … 00-xt-256mb-ddr
https://theretroweb.com/expansioncards/s/sapp … 56mb-ddr-v-d-vo
And there are a million 4-in-1 resistor arrays there.

shevalier wrote on 2025-07-18, 12:39:

in red square. in green - no sense There's nothing to check on this card. Asus didn't bother with impedance matching, so there a […]
Show full quote

analog_programmer wrote on 2025-07-18, 11:53:

What resistors I have to check with microscope (I have a USB-chineese-thing called "digital microscope" that's better than magnifying glass)?

in red square.
in green - no sense
There's nothing to check on this card.
Asus didn't bother with impedance matching, so there are simply no resistors.

A more advanced design on the same RV360 looks like this.
https://theretroweb.com/expansioncards/s/sapp … 00-xt-256mb-ddr
https://theretroweb.com/expansioncards/s/sapp … 56mb-ddr-v-d-vo
And there are a million 4-in-1 resistor arrays there.

I meant I can fiddle with the resistors in the GPU core voltage circuit: Re: A glitching Asus A9550GE/TD/256M/A

P.S. I like these "good" bulging bad capacitors even in more adequate PCB designs. No joke, Hitano c(r)aps are just "a silent death".

analog_programmer wrote on 2025-07-18, 07:54:

Anyway, it was necessary to replace all of these Hitano c(r)aps.

Oh absolutely, no argument about that. 😁
I mean, even if their ESR had gone up 10x, some circuits (e.g. linear regulators) could still tolerate that and operate fine. But when the capacitance goes so low (less than half), that's *not* good.

analog_programmer wrote on 2025-07-18, 07:54:

My next suggestion is for broken BGA GPU connection due to continuous overheating.

Well, these cards use leaded solder, so it's unlikely for the BGA solder between the GPU substrate and GPU PCB to have failed. Rather, it's usually where all the heat and stress is: between the actual GPU die and the GPU substrate (the square PCB on which the die is mounted on.)
That said, a "reflow" (reheat) process can still bring these back to normal operation.
Unfortunately, it just cannot be considered a permanent fix and the problem will likely come back again. My experience is that the lower the temperature the GPU is kept at after (while operating), the better the chance of the reheat "fix" to last... though also worth noting that I find that the more complex and larger the GPU die, the quicker the problem will come back. Luckily RV360 die is pretty small, so it should have a better chance of lasting longer.

analog_programmer wrote on 2025-07-18, 07:54:

Just adding a fan won't help. The problems begin when the card and GPU's heatsink become warm, not too hot.

Have you tried it, though?
FWIW, having a fan blowing on these cards can lower their idle run temps by at least 5C and more under load. So you might actually be able to see an improvement... or at least a delay in when the card starts to crash.

analog_programmer wrote on 2025-07-18, 07:54:

As far as i know Hitano c(r)apacitors are chineese (not even taiwanese) despite their japanese sounding name.

You know what, you might be right. It's been many years since I did my reading on those. In any case, all I remember is the same thing you said in your earlier post: they are crap and should be replaced anyways. 😁

analog_programmer wrote on 2025-07-18, 07:54:

Unfortunately Gigabyte card doesn't have any fan connector terminals. For now it runs fine.

It may run fine, but that's the problem with these: they are fine until they are just not. 😁

I need to dig out mine next time I'm in the countryside house and take a look at it for you. IIRC, there are a number of places where you can take 5V and 12V from and I think I even labeled it on mine. That said, on many cards (even if they have fan connectors), the easiest way for me to give their fan 7V is to solder new wires to spots that have 5V and 12V. (And before anyone says that's not a proper/good way of wiring things, for small loads such as a fan or three, it's fine. Almost all of the computer hardware I have from the early 2000's has a fan wired like this one place or another. Haven't had a problem with it in over 20 years. And I don't expect it to. The person I actually learned this from learned it from doing industrial wiring... so it's not necessarily an unsafe "hack".)

analog_programmer wrote on 2025-07-18, 07:54:

Reballing is not for me, but I can try BGA GPU connections "resoldering" with hot air gun...

Yup, that's the plan here anyways.
A reball would be a waste of time as it won't help anything anyways, since such issues are not the BGA between the substrate and card's PCB that fails.

analog_programmer wrote on 2025-07-18, 07:54:

As for my old and problematic PSUs, I really have to restore one ATX unit for older systems (with PIII or Athlon (XP) motherboards) testing. I managed to find two very similar circuit diagrams for LC-B250ATX PSU, which I think is ATX ver.1.x standard. This weekend I'll start to reassemble it. I also have one ATX ver. 2.1 standard AT-2005B PSU, which is in even worse state (I tink I keep all of it's main components) and I can't remember which MOSFET or double diode where goes on the PCB, but for now I'm not in need to restore it.

LC-B250ATX is a basically a very low-end and cost-cut Deer/Solytech unit. But their design is overall solid and workable with. So if that's what you have for your PSU (or another similar LC-BxxxATX unit), they make for decent ATX 1.0 PSUs with all of the proper parts installed. Namely, L&C puts much smaller and/or less caps on the output of these PSUs, along with often omitting the "PI" coils/inductors between the caps. Adding these back and making sure the 3.3V and 5V rail get at least 2x 2200 uF low ESR caps and the 12V rail at least 2x 1000 uF caps is really all that's needed for these PSUs to be decent again... well that and a full recap of the small caps too, along with the 5VSB circuit, as all of those will bring nothing but problems if disregarded.

FWIW, I have two LC-B300ATX units. One I semi-restored (added the proper caps on the output and also the missing PI coils), but didn't fully recap. The other is still in parts in one of my parts boxes. Going to restore it eventually too, despite having more than enough 5V-heavy PSUs I would ever need. At $0.50 (1 Lev) on the flea market here, it's hard to pass them up sometimes. Just the two BJT transistors inside it are worth more. 😀

shevalier wrote on 2025-07-18, 11:27:

This is while the capacitors at least somewhat correspond to the datasheet.

True.

shevalier wrote on 2025-07-18, 11:27:

If the capacitor ESR has increased 10 times (from 40 mOhm to 0.4 Ohm), then no amount of heating will help it.

Well, you might be surprised of how much the ESR can vary on failed caps then. I've seen them improve their ESR by 5-10x on a regular basis just from heating them with the soldering iron or with a hot air gun (usually heating them anywhere from 40-55C T-case).
So heating them most definitely can help them get "better" (temporarily anyways.)

What the heating won't and can't help is to restore lost capacity due to "consumed" foil plates from an unstable electrolyte... and ultimately it's this and not so much the ESR increase, that brings electronics circuits from functional to non-functional state... though that will depend greatly between different circuits. For example, if a problematic cap is filtering a rail that also has ceramic caps as part of the filtering, then an increase in the ESR of the electrolytic cap won't be that problematic... not until the filter's whole capacitance becomes so low that it starts messing with the compensation on the voltage feedback loop of the controller circuit. But in other circuits like the small "start-up"/bootstrap capacitors of standby power supplies, the ESR of these small electrolytic caps (which are responsible for filtering the primary auxiliary winding and also storing a charge at the circuit's "boot up" loop) may play a much greater role in when the circuit stops functioning.
So it really depends on the actual circuit.

That said, most electronics circuits are overall designed to be quite tolerant of the ESR of electrolytic caps.
Since the 2-4x increase in ESR/impedance with temperature decrease (from 20C to -10C) is really more of an average value than a hard number, and since the capacitors are expected to degrade over time, many (properly-designed) circuits will usually have enough overhead to easily tolerate a 10x increase in the ESR values. Thus, heating up caps that have gone bad (but not completely bad like the Hitano's above with almost non-existent capacity 😁 ) can sometimes "restore" a device, at least for a quick test.

shevalier wrote on 2025-07-18, 11:27:

As for checking the resistors, it's a good idea unless they are 0402. Then you either need to be young or have a microscope.
But being young is better 😀

Yeah, I can start to feel the weight of this statement a lot more now as time goes by.
FWIW, I can still see and measure these fine without a need for magnifications... but it's getting harder. 10 years ago, I wouldn't even blink once when having to work with these. Now I have to "train" my eyes for a few minutes before I can finally get them to focus... and it's still not as clear as I'd like it. But I'll take it.

analog_programmer wrote on 2025-07-18, 12:59:

I meant I can fiddle with the resistors in the GPU core voltage circuit: Re: A glitching Asus A9550GE/TD/256M/A

Solder a 20-22k resistor on top of the 222 (2.2k) resistor.
The supply voltage will increase by about 50mV.

momaka wrote on 2025-07-18, 14:29:

Oh absolutely, no argument about that. 😁
I mean, even if their ESR had gone up 10x, some circuits (e.g. linear regulators) could still tolerate that and operate fine. But when the capacitance goes so low (less than half), that's *not* good.

Yep, at least the original dead electrolyte capacitors are now out of failing equation 😀

momaka wrote on 2025-07-18, 14:29:

Well, these cards use leaded solder, so it's unlikely for the BGA solder between the GPU substrate and GPU PCB to have failed. Rather, it's usually where all the heat and stress is: between the actual GPU die and the GPU substrate (the square PCB on which the die is mounted on.)
That said, a "reflow" (reheat) process can still bring these back to normal operation.
Unfortunately, it just cannot be considered a permanent fix and the problem will likely come back again. My experience is that the lower the temperature the GPU is kept at after (while operating), the better the chance of the reheat "fix" to last... though also worth noting that I find that the more complex and larger the GPU die, the quicker the problem will come back. Luckily RV360 die is pretty small, so it should have a better chance of lasting longer.

Hmmm. I can't recall when exactly unleaded solder became a mandatory "human health care" bullsh*t. I remember it was in the very early '00s. But seems like "a shot" with hot air gun (+ installing a fan) may be considered as a possible treatment 😀

momaka wrote on 2025-07-18, 14:29:

Have you tried it, though?
FWIW, having a fan blowing on these cards can lower their idle run temps by at least 5C and more under load. So you might actually be able to see an improvement... or at least a delay in when the card starts to crash.

Not yet, 'cause card is not running fine with stock passive cooler without any serious GPU load (3D software under win is still mission impossible).

momaka wrote on 2025-07-18, 14:29:

analog_programmer wrote on 2025-07-18, 07:54:

Unfortunately Gigabyte card doesn't have any fan connector terminals. For now it runs fine.

It may run fine, but that's the problem with these: they are fine until they are just not. 😁

Yes, sir, that's why I keep it only as a backup AGP 8x card 😀

momaka wrote on 2025-07-18, 14:29:

I need to dig out mine next time I'm in the countryside house and take a look at it for you. IIRC, there are a number of places where you can take 5V and 12V from and I think I even labeled it on mine. That said, on many cards (even if they have fan connectors), the easiest way for me to give their fan 7V is to solder new wires to spots that have 5V and 12V. (And before anyone says that's not a proper/good way of wiring things, for small loads such as a fan or three, it's fine. Almost all of the computer hardware I have from the early 2000's has a fan wired like this one place or another. Haven't had a problem with it in over 20 years. And I don't expect it to. The person I actually learned this from learned it from doing industrial wiring... so it's not necessarily an unsafe "hack".)

This will be great, thank you!

momaka wrote on 2025-07-18, 14:29:

Yup, that's the plan here anyways.
A reball would be a waste of time as it won't help anything anyways, since such issues are not the BGA between the substrate and card's PCB that fails.

As hot air treatment may cause a permanent GPU death, I'll use this as second to last repair step. The last step is card goes into "for spare parts" box 😁

momaka wrote on 2025-07-18, 14:29:

LC-B250ATX is a basically a very low-end and cost-cut Deer/Solytech unit. But their design is overall solid and workable with. So if that's what you have for your PSU (or another similar LC-BxxxATX unit), they make for decent ATX 1.0 PSUs with all of the proper parts installed. Namely, L&C puts much smaller and/or less caps on the output of these PSUs, along with often omitting the "PI" coils/inductors between the caps. Adding these back and making sure the 3.3V and 5V rail get at least 2x 2200 uF low ESR caps and the 12V rail at least 2x 1000 uF caps is really all that's needed for these PSUs to be decent again... well that and a full recap of the small caps too, along with the 5VSB circuit, as all of those will bring nothing but problems if disregarded.

FWIW, I have two LC-B300ATX units. One I semi-restored (added the proper caps on the output and also the missing PI coils), but didn't fully recap. The other is still in parts in one of my parts boxes. Going to restore it eventually too, despite having more than enough 5V-heavy PSUs I would ever need. At $0.50 (1 Lev) on the flea market here, it's hard to pass them up sometimes. Just the two BJT transistors inside it are worth more. 😀

As this is already off-topic, but still valuable for me, I'll switch to our native language.

shevalier wrote on 2025-07-18, 14:42:

Solder a 20-22k resistor on top of the 222 (2.2k) resistor.
The supply voltage will increase by about 50mV.

It's easier and cleaner to replace the resistor. I don't like these piggyback/sandwich solutions 😉 So, 18 to 20 ~1.9 to 2 kOhm resistor in place of 220 2.2 kOhm one should be fine, right?

EDIT: Corrected miscalculated values.

analog_programmer wrote on 2025-07-18, 15:55:

shevalier wrote on 2025-07-18, 14:42:

Solder a 20-22k resistor on top of the 222 (2.2k) resistor.
The supply voltage will increase by about 50mV.

It's easier and cleaner to replace the resistor. I don't like these piggyback/sandwich solutions 😉 So, 18 to 20 kOhm resistor in place of 220 kOhm one should be fine, right?

Since the resistor is 2.2k and you would lower it by putting a resistor in parallel, meaning if you do a complete replacement of the resistor, you'll need something lower than 2.2k as a replacement.

analog_programmer wrote on 2025-07-18, 15:55:

shevalier wrote on 2025-07-18, 14:42:

Solder a 20-22k resistor on top of the 222 (2.2k) resistor.
The supply voltage will increase by about 50mV.

It's easier and cleaner to replace the resistor. I don't like these piggyback/sandwich solutions 😉 So, 18 to 20 kOhm resistor in place of 220 kOhm one should be fine, right?

Nope, 222 = 22 * 10^2= 22 * 100= 2200 Ohm

Replace resistor must be 2kOhm
marking as 202 or 30b (29b or 31b fine too)

shevalier wrote on 2025-07-18, 16:10:

Nope, 222 = 22 * 10^2= 22 * 100= 2200 Ohm

Replace resistor must be 2kOhm
marking as 202 or 30b (29b or 31b fine too)

My mistake. I calculated the new value based on 220 kOhm resistor instead of 2200 Ohm 😁 Thanks for the correction!