How about pads direct from an electronics supplier? Probably better pricing.
https://www.digikey.com/en/products/filter/pa … A4EMAmBnEBdAvkA

I don't think it's necessary to take every card apart to replace pads just because they're old. If the heatsinks are hot when the card is in use, the pads are doing their job. You could get a cheap IR gun and compare card/heatsink temps more accurately if you feel like it.

Also, when you pull cards apart and reassemble them, you are putting stress on the card and the solder. That's probably something to avoid.

Personally, I buy them in sheets or strips and cut them to size myself. So much easier than trying to have stock of many different sizes, pre-cut. The only pre-cut ones I have are ones that I use on things like voltage regulators.

When it comes to replacing pads or paste, that all comes down to the temps under load. If they're good, I don't bother, if I don't like what I see, then I change them. For paste, if I remove the cooler, I always clean and replace the paste. For pads, I almost always do the same if the card is "older", just becasue I may as well if I have it apart anyway, but I don't often do it on newer card, like those made in the last 10 years or so. It all depends on a per-situation basis.

swaaye wrote on 2024-09-27, 15:14:

How about pads direct from an electronics supplier? Probably better pricing. https://www.digikey.com/en/products/filter/pa … A […]
Show full quote

How about pads direct from an electronics supplier? Probably better pricing.
https://www.digikey.com/en/products/filter/pa … A4EMAmBnEBdAvkA

I don't think it's necessary to take every card apart to replace pads just because they're old. If the heatsinks are hot when the card is in use, the pads are doing their job. You could get a cheap IR gun and compare card/heatsink temps more accurately if you feel like it.

Also, when you pull cards apart and reassemble them, you are putting stress on the card and the solder. That's probably something to avoid.

The X850 had to come apart because it looked like it had lived it's life inside an oven. Caked with brown goo,tarnished the acrylic of the fan housing and packed with dust. Also it came to me as not working and somehow it fired back up after some trial and error. So i wasn't too worried about taking the cooler off since there wasn't much to lose at that point. I will consider leaving the 7900gto be for the moment.

I will take a look at getting an IR device and have a look at that site for thermal pads. Thanks for the tips.

StriderTR wrote on 2024-09-27, 18:53:

Personally, I buy them in sheets or strips and cut them to size myself. So much easier than trying to have stock of many different sizes, pre-cut. The only pre-cut ones I have are ones that I use on things like voltage regulators.

When it comes to replacing pads or paste, that all comes down to the temps under load. If they're good, I don't bother, if I don't like what I see, then I change them. For paste, if I remove the cooler, I always clean and replace the paste. For pads, I almost always do the same if the card is "older", just becasue I may as well if I have it apart anyway, but I don't often do it on newer card, like those made in the last 10 years or so. It all depends on a per-situation basis.

The sheets are more or less what i was referring to ( well more like 10x10 cm squares or similar) but these cost €15 a piece easily. I was more troubled with the thicknesses that i would have to get (0.5 up to 2.0 mm in .5 increments) meaning i need at least one pack of each. But i might just bite the bullet and order €75 worth of pads for a variety of cards (unless the link provided by @swaaye has quality but affordable ones)

The temps from what i've seen are usually okay on the GPU chip but most cards i've dealt with don't have memory temp sensors so it's hard to determine wether they are cooking or doing just fine. Mostly the reason i wanted to change these out is because i don't know how the current pads are doing. Like if you buy a second hand car .. has the p/o changed the oil and filter regularly or has it been in there 3 times the recommended interval (i just start fresh and change every fluid in the car so i know it's good)

The IR camera @swaaye had posted might be a good idea for this issue (though knowing what temps are actually good for any given chip is also difficult)

As with any repair or manufacturing trade, there is going to be waste or leftover. I don't see why not just buy sheets when you need it and going to do it, and store the leftover. You don't just stock up on everything unless you'll need it frequently. Buying specially sized pads are likely going to be more money than just buying the bulk.

As far as finding the right temperatures, just start collecting data with your own probes/camera, or get them off cards/cpu that do have sensors. Generally, all silicon is made the same, so at some point it's just not good. Personally I'd make sure the things don't goes over 60*C, but that's gonna vary how you look at it. They can tolerate more than that on the die but, I also like longer life. Just get a feel.

Sheets seem to be more expensive where you are. The set I buy over on Amazon here in the US is about $12 (about €10.75). It's 3 sheets, 100x100mm in size with 0.5mm, 1mm, and 1.5mm thickness. They used to be about $5-$6, but not since the world went crazy.

https://www.amazon.com/dp/B09DC772PR/?coliid=I1TZ43VT71UFJO

These work great, been using them for a while. Sadly, I don't know of any "cheaper" alternative outside of the US. 🙁

Isn't that 7900GTO from the days of failing solder BGAs on nvidia cards. Might not want to physically stress it unnecessarily, depends how hard it is to get off.

My 2 (inflated) cents on the matter: I never replace the thermal pads on these old cards. Now if this was a modern late GDDR5 or GDDR6 card, where every fraction of a Watt counts because the manufacturer already stretched the legs of the card as far as they'd go to get the most performance out it, then this might be different matter. But for old cards, this is not the case. Thermal pads offer a rather poor thermal transfer to the GPU heatsink. But that's not an issue, because the RAM chips themselves don't dissipate that much heat (maybe a Watt or two tops), so the thermal pads will slowly but surely be able to pump the heat away from the thermal pad onto the heatsink (or the other way around when the heatsink gets hotter than the RAM chips.) In fact, if you read the RAM datasheets, you will see that BGA RAM chips rarely suggest active cooling via thermal pads. Instead, they are meant to cool through their solder BGA ground pads to the board by design. Any cooling through the top (be it via thermal pads or just simply air passing over) is just a bonus. That's why some GPUs have thermal pads, while others use air from the cooler to pass over them. In any case, it all comes down to how cool the video card PCB runs. Thus, if you keep the GPU chip (and its VRM cool), the card's PCB will remain cooler too, and that will allow the RAM chips to also run cooler.
In some cases like the stock/reference Radeon HD4850 cooler, the RAM actually runs hotter because the GPU heatsink is undersized (and ATI/AMD made the "wise" decision to keep the fan speed low until the card hits 80 Celsius), so extra heat from the GPU is pumped into the RAM chips and dissipated through the PCB - a terrible terrible design overall. Though it should suffice to say the HD4850 is not the only card with this flaw. Actually, there's more cards from that era that are like this than ones that aren't. Usually the ones that aren't are the 3rd part cards with non-reference and bigger coolers.

With all of that said, if the old pads still make contact when you put the cooler back on, then you have nothing to worry about or replace. And if seems like they don't, add a tiny bit of thermal compound on one side. Yes, we all know a thick(-ish in this case) layer of thermal compound is not good for thermal transfer... but again, these older RAM chips don't really need to cool through their top, if the PCB is properly designed to sink heat away from their solder pads anyways. In any case, even a thick layer of compound will be OK to transfer the heat away (or to?) the RAM chips, because the overall TDP of each chips is low.

swaaye wrote on 2024-09-27, 15:14:

How about pads direct from an electronics supplier? Probably better pricing.
https://www.digikey.com/en/products/filter/pa … A4EMAmBnEBdAvkA

+1 on using Digikey or Mouser. Usually way cheaper than "gaming gear" brands. Same reason I usually don't consider buying Noctua fans - it's not that they are bad, but they are kinda overpriced for what they are. With some searching, I can find equivalently-performing quality brands of fans on Digikey or Mouser (Panaflo, Nidec, Delta, and etc.) Just have to do more digging through datasheets to look at noise figures, pressures, airflow, and etc... but overall it's worth it, because it allows me to get the right fan for the application. Some coolers with dense fins need higher pressure and not necessarily lots of airflow. Others with more sparsely-placed fins could use lower pressure and higher airflow.

swaaye wrote on 2024-09-27, 15:14:

Also, when you pull cards apart and reassemble them, you are putting stress on the card and the solder. That's probably something to avoid.

Now when did this myth come about?
If this was true, then simply walking in the room more frequently would destroy GPUs quicker as the floor vibrations from that will far exceed any stress you put on taking apart the cooler...
Seriously, solder is not that gentle, especially at room (cold) temperatures. And with most lead-free solders, you could be sitting there just 2 degrees below the melt point of the solder and it still won't be budging.

StriderTR wrote on 2024-09-27, 18:53:

If they're good, I don't bother, if I don't like what I see, then I change them.

Replacing RAM pads on a hot-running card won't make a dent in the temperatures.
If it's running too hot, it's usually because:
a) the GPU thermal paste is dry or starting to dry
b) the cooler is clogged with dust
c) the cooler was improperly matched to the TDP of the GPU (i.e. it is undersized) - the most common issue, in fact, as manufacturers always try to cut down on costs, and this is one easy place to do it.
d) fan is going bad
e) a combo of any of the above

StriderTR wrote on 2024-09-27, 18:53:

For paste, if I remove the cooler, I always clean and replace the paste.

That's wasteful in some cases.
If the paste is not dry and not contaminated with dust or dirt where the GPU die contacts the cooler (rarely the case with old video cards), then I'll re-use it and/or simply add a small amount to the existing paste. This is especially the case when I apply new thermal paste to a cooler that I've already cleaned and installed on a GPU, but removed to install onto another GPU (typically for testing.) So long as the paste wasn't overheated (and for a long period of time), then I scoop up the "old" paste from the previous GPU and move it onto the next (clean) one. I've even done tests specifically for this, to see if reusing the paste will degrade cooling performance. The conclusion that I've come to is that if not dealing with crazy-high TDP density (i.e. modern GPUs that have high TDP in a really small surface area), then difference is close to none (less than 1 degree C) between all re-use applications until the paste either gets contaminated with dust/dirt or starts to dry.

StriderTR wrote on 2024-09-27, 18:53:

It all depends on a per-situation basis.

Exactly, that's really the best way to treat ever video card restoration.

Always check everything and never assume the manufacturer did everything right to get you the best cooling solution. In many cases, they did give you the best cooling solution... for the money they put in that cooler. But that's not the same as picking the proper cooler for the TDP of the card. In other words, they only care to give you what will last the warranty and then maybe a bit more to be on the safe side. Past that, it's a waste to them (and really to most people that regularly replace their computer hardware every 3-5 years.)

ratfink wrote on 2024-09-29, 13:47:

Isn't that 7900GTO from the days of failing solder BGAs on nvidia cards.

All GF7 and GF8 series with G8x GPU chips are from that "bumpgate" era. Even the GF6 series are. GF 6150 chipsets are some of the most notorious ones for failing, both in laptops and on desktop motherboards. Though part of their problem is they are typically only fitted with a passive cooler. Add a small 40 mm fan on top, and they can actually work a lot more reliably (even reflown ones.)

momaka wrote on 2024-10-01, 01:50:

Now when did this myth come about?
If this was true, then simply walking in the room more frequently would destroy GPUs quicker as the floor vibrations from that will far exceed any stress you put on taking apart the cooler...
Seriously, solder is not that gentle, especially at room (cold) temperatures. And with most lead-free solders, you could be sitting there just 2 degrees below the melt point of the solder and it still won't be budging.

Prying off 1997-2001 glued heatsinks. Hardened paste on a large surface area. Aftermarket coolers that stress/bend the PCB. It's not entirely about the solder.

momaka wrote on 2024-10-01, 01:50:

All GF7 and GF8 series with G8x GPU chips are from that "bumpgate" era. Even the GF6 series are. GF 6150 chipsets are some of the most notorious ones for failing, both in laptops and on desktop motherboards. Though part of their problem is they are typically only fitted with a passive cooler. Add a small 40 mm fan on top, and they can actually work a lot more reliably (even reflown ones.)

Yeah, I had a 6600 , a 6800GT, and three 7950GX2s fail - I assumed due to solder. A Gainward agp 7900GS+ survived at least until it was sold a few years ago, but I went with AMD/ATI for the next upgrade (HD4850) and not back to nvidia until the GTX56oTi I am still running a few motherboard/cpu upgrades later.

StriderTR wrote on 2024-09-29, 13:43:

Sheets seem to be more expensive where you are. The set I buy over on Amazon here in the US is about $12 (about €10.75). It's 3 sheets, 100x100mm in size with 0.5mm, 1mm, and 1.5mm thickness. They used to be about $5-$6, but not since the world went crazy.

https://www.amazon.com/dp/B09DC772PR/?coliid=I1TZ43VT71UFJO

These work great, been using them for a while. Sadly, I don't know of any "cheaper" alternative outside of the US. 🙁

They don't seem to be available here. They do look like they have a random brand on it so it's possible there's several other sellers offering the same thing under another name (as is often the case on amazon with chinese sellers/manufacturers) i'll have a look.

ratfink wrote on 2024-09-29, 13:47:

Isn't that 7900GTO from the days of failing solder BGAs on nvidia cards. Might not want to physically stress it unnecessarily, depends how hard it is to get off.

I didn't know this era cards was plagued by this. I will test it out first and leave the card be if it's cool enough.

momaka wrote on 2024-10-01, 01:50:

My 2 (inflated) cents on the matter: I never replace the thermal pads on these old cards. Now if this was a modern late GDDR5 or […]
Show full quote

My 2 (inflated) cents on the matter: I never replace the thermal pads on these old cards. Now if this was a modern late GDDR5 or GDDR6 card, where every fraction of a Watt counts because the manufacturer already stretched the legs of the card as far as they'd go to get the most performance out it, then this might be different matter. But for old cards, this is not the case. Thermal pads offer a rather poor thermal transfer to the GPU heatsink. But that's not an issue, because the RAM chips themselves don't dissipate that much heat (maybe a Watt or two tops), so the thermal pads will slowly but surely be able to pump the heat away from the thermal pad onto the heatsink (or the other way around when the heatsink gets hotter than the RAM chips.) In fact, if you read the RAM datasheets, you will see that BGA RAM chips rarely suggest active cooling via thermal pads. Instead, they are meant to cool through their solder BGA ground pads to the board by design. Any cooling through the top (be it via thermal pads or just simply air passing over) is just a bonus. That's why some GPUs have thermal pads, while others use air from the cooler to pass over them. In any case, it all comes down to how cool the video card PCB runs. Thus, if you keep the GPU chip (and its VRM cool), the card's PCB will remain cooler too, and that will allow the RAM chips to also run cooler.
In some cases like the stock/reference Radeon HD4850 cooler, the RAM actually runs hotter because the GPU heatsink is undersized (and ATI/AMD made the "wise" decision to keep the fan speed low until the card hits 80 Celsius), so extra heat from the GPU is pumped into the RAM chips and dissipated through the PCB - a terrible terrible design overall. Though it should suffice to say the HD4850 is not the only card with this flaw. Actually, there's more cards from that era that are like this than ones that aren't. Usually the ones that aren't are the 3rd part cards with non-reference and bigger coolers.

With all of that said, if the old pads still make contact when you put the cooler back on, then you have nothing to worry about or replace. And if seems like they don't, add a tiny bit of thermal compound on one side. Yes, we all know a thick(-ish in this case) layer of thermal compound is not good for thermal transfer... but again, these older RAM chips don't really need to cool through their top, if the PCB is properly designed to sink heat away from their solder pads anyways. In any case, even a thick layer of compound will be OK to transfer the heat away (or to?) the RAM chips, because the overall TDP of each chips is low.

swaaye wrote on 2024-09-27, 15:14:

How about pads direct from an electronics supplier? Probably better pricing.
https://www.digikey.com/en/products/filter/pa … A4EMAmBnEBdAvkA

+1 on using Digikey or Mouser. Usually way cheaper than "gaming gear" brands. Same reason I usually don't consider buying Noctua fans - it's not that they are bad, but they are kinda overpriced for what they are. With some searching, I can find equivalently-performing quality brands of fans on Digikey or Mouser (Panaflo, Nidec, Delta, and etc.) Just have to do more digging through datasheets to look at noise figures, pressures, airflow, and etc... but overall it's worth it, because it allows me to get the right fan for the application. Some coolers with dense fins need higher pressure and not necessarily lots of airflow. Others with more sparsely-placed fins could use lower pressure and higher airflow.

swaaye wrote on 2024-09-27, 15:14:

Also, when you pull cards apart and reassemble them, you are putting stress on the card and the solder. That's probably something to avoid.

Now when did this myth come about?
If this was true, then simply walking in the room more frequently would destroy GPUs quicker as the floor vibrations from that will far exceed any stress you put on taking apart the cooler...
Seriously, solder is not that gentle, especially at room (cold) temperatures. And with most lead-free solders, you could be sitting there just 2 degrees below the melt point of the solder and it still won't be budging.

StriderTR wrote on 2024-09-27, 18:53:

If they're good, I don't bother, if I don't like what I see, then I change them.

Replacing RAM pads on a hot-running card won't make a dent in the temperatures.
If it's running too hot, it's usually because:
a) the GPU thermal paste is dry or starting to dry
b) the cooler is clogged with dust
c) the cooler was improperly matched to the TDP of the GPU (i.e. it is undersized) - the most common issue, in fact, as manufacturers always try to cut down on costs, and this is one easy place to do it.
d) fan is going bad
e) a combo of any of the above

StriderTR wrote on 2024-09-27, 18:53:

For paste, if I remove the cooler, I always clean and replace the paste.

That's wasteful in some cases.
If the paste is not dry and not contaminated with dust or dirt where the GPU die contacts the cooler (rarely the case with old video cards), then I'll re-use it and/or simply add a small amount to the existing paste. This is especially the case when I apply new thermal paste to a cooler that I've already cleaned and installed on a GPU, but removed to install onto another GPU (typically for testing.) So long as the paste wasn't overheated (and for a long period of time), then I scoop up the "old" paste from the previous GPU and move it onto the next (clean) one. I've even done tests specifically for this, to see if reusing the paste will degrade cooling performance. The conclusion that I've come to is that if not dealing with crazy-high TDP density (i.e. modern GPUs that have high TDP in a really small surface area), then difference is close to none (less than 1 degree C) between all re-use applications until the paste either gets contaminated with dust/dirt or starts to dry.

StriderTR wrote on 2024-09-27, 18:53:

It all depends on a per-situation basis.

Exactly, that's really the best way to treat ever video card restoration.

Always check everything and never assume the manufacturer did everything right to get you the best cooling solution. In many cases, they did give you the best cooling solution... for the money they put in that cooler. But that's not the same as picking the proper cooler for the TDP of the card. In other words, they only care to give you what will last the warranty and then maybe a bit more to be on the safe side. Past that, it's a waste to them (and really to most people that regularly replace their computer hardware every 3-5 years.)

ratfink wrote on 2024-09-29, 13:47:

Isn't that 7900GTO from the days of failing solder BGAs on nvidia cards.

All GF7 and GF8 series with G8x GPU chips are from that "bumpgate" era. Even the GF6 series are. GF 6150 chipsets are some of the most notorious ones for failing, both in laptops and on desktop motherboards. Though part of their problem is they are typically only fitted with a passive cooler. Add a small 40 mm fan on top, and they can actually work a lot more reliably (even reflown ones.)

Thank you for your input. I have just finished the X850XT-PE and re-used the thermal pads. The ram heatsink at the back of the card seems to get quite hot so at least that means the thermal transfer is still there. Wasn't able to get the fan working properly though ... oiled the shaft but probably i just created metal and oil gunk inside the sleeve now. It's still spinning okay at idle and full blast but it's not quieter by any means. It's spent alot of years literally caked in dust in the p/o's computer so that didn't help either. Won't expect it to last much longer ... and replacement seems difficult to find (Milennium B127530BU)

I'll have a good look on Digikey .. they seem to have a Belgian store aswell which is handy.

I didn't know the chips actually used the PCB to disperse heat more than the top of the chip. Very interesting!

It's indeed not surprising the manufacturers spend the least amount of money possible to cool the cards just enough to keep them from dying/throttling. Day's of overengineering are loooong gone (in any market really). Competition is great for consumers in pricing but also leads to company's trying to save every penny possible leading to "just good enough" products. Double edged sword .. but anyway this is nothing new and probably was the case in the early day's of computing. But global markets and the internet probably worsened this alot.

G-X wrote on 2024-10-01, 19:35:

Wasn't able to get the fan working properly though ... oiled the shaft but probably i just created metal and oil gunk inside the sleeve now. It's still spinning okay at idle and full blast but it's not quieter by any means. It's spent alot of years literally caked in dust in the p/o's computer so that didn't help either. Won't expect it to last much longer ... and replacement seems difficult to find (Milennium B127530BU)

If you can get to the shaft, then you should be able to take it apart. Normally there's a plastic c-washer type of clip on the shaft that keeps the rotor (blade assembly) from being pulled off. Once you remove the c-washer, you can remove the blade assembly and clean the separately. You will also then be able to see if the fan is a sleeve bearing type or a ball bearing type... or 1 ball, 1 sleeve bearing.

Sleeve bearings are easy to clean and restore back to like-new working condition. First, clean the sleeve with something small and slightly abrasive in a direction parallel to the shaft orientation. My personal choice is a small (smallest possible, really) flat head screw driver to scratch away and dust and dirt in there. The scratching parallel to the shaft orientation is important, because it creates micro-channels in the sleeve material that will later be used for holding an oil film, essentially turning the sleeve bearing to behave more like a rifle bearing. Once the cleaning with scratching is done, clean everything in the sleeve with IPA, then add fine machine oil (not-too-thick engine oil OK too.)

Ball bearings, on the other hand, aren't really meant to be serviceable. If the fan is loud because of this, the only long-term reliable option is to replace the ball bearing. If the ball bearing is loud but not overly worn out, you can *try* to take it apart and attempt to service. I've done this to a few bearings, but it's a PITA to get them opened. Usually, there's a very thin c-shaped wire on the outter diameter of the bearing that needs to be removed. After this, the protective shield over the ball bearing's balls should be possible to remove. Once the ball assembly in the bearing is exposed, use brake/carburetor cleaner to get all previous grease (and gunk) out of there. Let dry for a few hours. Then re-pack the ball assembly with white lithium grease. Re-assemble the bearing and test it. If it wasn't too worn out, it will turn nice and smooth and not make noise like it did before. If it's too worn out, it will spit the grease back out rather quickly and then become noisy again. So to verify operation, put it back in the fan and assemble the fan partially (no need to reinstall the c-washers yet), then connect the fan to a power source to let it run for a while.

And that's about all you can do to restore fans.

With sleeve bearing fans restored properly by the above mentioned technique, you should get at least a few years of good (and fairly heavy) use out it, if not more.
With ball bearing fans... roll the dice. :\

swaaye wrote on 2024-10-01, 15:41:

Prying off 1997-2001 glued heatsinks.

Well, in case of O/P's cards, neither have glued heatsinks, so he's safe from that at least.
Also will depend if it's thermal glue or thermal epoxy. The former is not that hard to pull and doesn't exert enough pressure to damage anything. Thermal epoxy is pretty much permanent, though. I suspect this is what's on one of my Radeon 9000 cards, because no amount of freezing or heating with a heat gun (went as far as 150C) will soften it up or weaken it enough to pop out. So I've left it alone for the time being. The card still works, despite some rather harsh attempts by me to remove the old heatsink. But given what I have tried so far, I'll probably leave it alone and maybe someday cut the fins on the old heatsink and slap a new one on top.

swaaye wrote on 2024-10-01, 15:41:

Hardened paste on a large surface area.

Not an issue. Heat with a hair dryer to 40-50ish degrees C and it should come off pretty easily. If it's very dry, it usually comes off by itself without any heating... unless is that yellow bubble-gun stuff. But even that tends to give out easily with heat.

swaaye wrote on 2024-10-01, 15:41:

Aftermarket coolers that stress/bend the PCB.

Do you know of any specific ones that do?
I've ran into a lot of 3rd party / aftermarket coolers and never saw one bend the PCB... or very insignificantly at best. Nothing like stock Intel s775 pushpin coolers or many stock AMD socket 754/939/AM2/AM3 coolers - now those can be real trouble, especially with many release/tighten cycles, because they tend to warp the board really badly.

ratfink wrote on 2024-10-01, 17:43:

Yeah, I had a 6600 , a 6800GT, and three 7950GX2s fail - I assumed due to crappy coolers.

Fixed that sentence for ya. 😉
The coolers used on the 6600 (GT) are often no good for anything past 20-25 Watts. The 6600 GT is a 49 Watt TDP card (that's almost 50 Watts). The 6600 LE (and 6200 based around the 6600 core) is a little better in that regard - about 30-ish Watts TDP... so the same cooler is just about adequate for these.

The 6800 GT cooler I'd rate for maybe 40 Watts. Yet, here we are pushing about 60W or more with the 6800 GT core. So of course it will run hot too.

But nothing comes close to the 7950GX2 in terms of how bad it is, especially the GPU in the 1st layer of that card. IIRC, the GX2 coolers perform even worse than the 7900 GS stock cooler... and those are pretty underwhelming for the 50W TDP of the 7900GS already.

The yellow bubble gum stuff is probably phase change.

Aftermarket coolers without a brace for the back of the card, or an ineffective brace, tend to bend cards.

momaka wrote on 2024-10-01, 20:11:

The coolers used on the 6600 (GT) are often no good for anything past 20-25 Watts. The 6600 GT is a 49 Watt TDP card (that's almost 50 Watts). The 6600 LE (and 6200 based around the 6600 core) is a little better in that regard - about 30-ish Watts TDP... so the same cooler is just about adequate for these.

The 6800 GT cooler I'd rate for maybe 40 Watts. Yet, here we are pushing about 60W or more with the 6800 GT core. So of course it will run hot too.

But nothing comes close to the 7950GX2 in terms of how bad it is, especially the GPU in the 1st layer of that card. IIRC, the GX2 coolers perform even worse than the 7900 GS stock cooler... and those are pretty underwhelming for the 50W TDP of the 7900GS already.

Oh that makes sense - some of my cards died pretty quickly in my son's (hard gaming) PC back in the day. The only one from that era (that I bought since) I have now is a Gainward 6600GT 256mb card which I think has the additional heatsink on the back of the card.

I think the GeForce 7 cards were affected by bumpgate solder failure. I don't think I've run into a 6000 series failure but it's hard to remember. Some of the FX 5xxx series cards failed because of RAM chip and possibly solder failure. It's the same story with some Radeon 9xxx cards.