How long do you think my resurrected GPU will last
The thermal compound might have been the problem all along. It dries out then doesnt conduct heat as well, I've also had a card from Asus apart for heat issues.... The factory kinda missed where it goes. Half the die had no paste and there was a good amount smeared on the pcb. :/ A quick cleaning and some Artic Silver MX-2 fixed it right up.
Comment has been collapsed.
I don't think that was it because I checked the temps when it was dying and they were reporting normal. I replaced the thermal compound with Noctua NT-H1 when I bought it 4 years ago. When I wiped it off today, it wasn't crusty at all, it wiped off like a thick paste. Not sure how long it lasts, but I still have the same tube and I used it today when I put it back together. I bought the thermal compound in 2012.
Comment has been collapsed.
Its crazy what you can do to put a short term fix on things.
I remember around 2004 and a lot of people I knew had phones that would randomly turn off / lose battery power. The fix turned out to be simple enough - a small folded bit of cardboard between the battery and the casing to push the battery onto the connectors (which themselves couldnt be moved in most cases).
I'm guessing you'll get maybe six months on your baked GPU before you maybe need to do it again (purely a shot in the dark).
Comment has been collapsed.
That's what I was thinking. If it doesn't die right when I load a game, I was thinking it would last around 3-6 months.
Comment has been collapsed.
Usually baked GPUs don't last very long as the cause of the problem isn't really solved, you just realigned its connectors. You could start looking out for another video card in the used market for cheap or you could wait for Nvidia to release their 1100 series, which should arrive in a couple of months.
Comment has been collapsed.
If it lasts more than 10 minutes in a game, I was thinking it would probably last around 3-6 months at most. I am really surprised so many people are voting for over 1 year, maybe they are just trolling.
I am already doing both. I have been following the rumors of the new 11 or 20 series Nvidia cards for almost a year. Hopefully this rumor will be true and they will be released in the next couple months. I probably won't be able to afford one though. I really wanted to buy a used GTX 970 and I should have when they were selling every day for $120 shipped for a good card or around $100 shipped for a reference design. I have been waiting for them to come back down in price, but they haven't dropped that far yet. I'm sure there will be a ton of used 970s when the 1170 or 1160 is released.
I would be happy with an 1160, but I don't think the price for performance will be as good as a 970 for $120 shipped. Another problem is that I still use a CRT monitor and the 9 series was the last generation to have analog outputs. I really don't want to use an active adapter. A 980ti would be awesome, but they are going for more than I want to risk on a used GPU. I am also looking at getting a 144hz, 1440P, IPS monitor, but they are also still too expensive even though the prices dropped recently.
Comment has been collapsed.
If you plan for gaming at 1440p/144hz then my suggestion is to get at least a 1160 or even better a 1170 that, according to rumors, should be somewhat powerful like a 1080ti. It's expensive, but mining is slowly decreasing over time and probably prices will go down accordingly.
Otherwise if you'll stick with the CRT monitor for a long time just get a used 980. Memory usage is increasing in modern games and that fuckery with the 3.5 gb of the 970 will show up quite early.
Comment has been collapsed.
I baked my brother's old laptop's gpu at least 5 times, it lasted some months between the bakings, but laptop was rarely used. It died in the end.
It can work for a while, but be ready to buy a new one, or (as I did), ask some friends if they have an extra one and get it second-hand from them (if they didn't use it for mining, of course).
Comment has been collapsed.
In my experience, you'll prolly get a couple months tops. I'd be looking into a new card asap.
Comment has been collapsed.
I have been looking for a new card for about a year and will hopefully upgrade soon. If this card dies again, I don't mind going back to the old HD 4850 for a while until I can get something new. I have a backlog of around 800 games to play and I enjoy indie games or old classics that are not difficult to run, so I have plenty of games to hold me over.
Comment has been collapsed.
my brother did that with an old 560ti i've used for years, till it died T_T
i would recommend you to underclock it, and force the fan to 100% speed when it hits 60c (you can make a speed/temp curve for it).
i used msi afterburner and it kept the video card working fine for a very long time. the only times we had to "fix" it again was when it hit 70c in summer (seems that was the temp limit for the flaw that broke it).
Comment has been collapsed.
I'm not too worried about saving it, I enjoy playing indie games and classics that are easy to run and I have tons of them to play in my backlog if this card dies again. They can hold me over until I am able to upgrade if needed. I am more interested in seeing the effects of heating a dead GPU.
I am curious to see what it can withstand, but I don't want to try and kill it. I might try that after I get a new GPU though. I was only running about an 8% overclock and the factory voltage because it has a blower cooler and runs hot. I might just turn it back down to the stock 580 clock speed and run it that way. I don't think I can really drop the voltage much anyway, it idles at 0.963 and only goes up to 1.025 at full load.
Comment has been collapsed.
It's more of an experiment for me than actually trying to save the GPU. I am curious to see how much it really brings the card back to life and don't mind if it dies. The real test will be once I start playing a game that pushes the card to 100% usage again. The games I started playing and was planning on playing are not very demanding though because I made plans when I was using the HD 4850 GPU, so it might be a while before I push it too hard.
Comment has been collapsed.
I can't either which is why I have been waiting for a year. I will probably end up getting a used 970 after the new cards are released. I have a ton of classic and indie games that are not difficult to run and I really enjoy them. If this card dies again, I can go back to my old HD 4850 and play those until I am able to upgrade. That is what I have been doing for the past 4 months since this card died. I probably have enough easy to run games in my backlog to play for at least a year or 2 if needed, maybe longer.
Comment has been collapsed.
reflow lasts for about 6 months and you can do it again if you want. better check for a new gpu though.
Comment has been collapsed.
I resurrected my ps3 a couple of times this way and each time it lasted a couple of days / weeks before giving up.
Comment has been collapsed.
same with my Xbox360, towel trick in my case... :3
sold it to a 2nd hand shop then who scammed people and bought stolen smartphones , haha... XD
Comment has been collapsed.
That's kind of screwed up. The shop probably sold it to someone with no warranty and that person got screwed. I wouldn't mind fixing a card like this and then selling it, but I would specifically list it as "for parts" and explain what happened and that it may not work for long.
I always wondered why I see dead GPUs selling online for so much money and now I think I know why. People are buying them and doing this or replacing a single dead capacitor and then selling the card as working knowing it will probably die again in a few months. I bought my last 2 GPUs used and they were in great condition, but I am a worried now knowing there is probably a lot of this going on.
There's a lot of dead GTX 970s listed as "parts only" that sold for around $130 recently even though just last year you could buy working 970s for $120. That's crazy, these people must be doing shady fixes and then throwing them back up for sale as working.
Comment has been collapsed.
the buyer can bring the xbox back... in germany you have even some time to bring back 2nd hand electronics... so only this criminal shop got fucked...
Comment has been collapsed.
My venerable R9 290X just died. (The same fate also befell my aging R9 290 a few weeks ago). I tried reapplying thermal paste but to no avail. I want to try the heat gun method. Did you apply the heat to the top of the die or also to the bottom side?
Comment has been collapsed.
I didn't do anything on the back, I just placed it on the floor and only heated the top of the die through the hole in the foil.
You will need to be a bit more careful because you don't have a heat spreader. You have components surrounding the die and you need to make sure you don't move them at all if you melt the solder holding them on. I believe they are little capacitors. After you get done heating the die, I would leave it right where you heated it and not touch it for about 10 minutes or till it feels cool enough to handle. This way, if any solder was melted on any other components, they won't get moved. Also, use the low air speed setting on the heat gun if you have the choice.
It would be best if you could cover the capacitors with foil so they don't get a lot of heat, but that will be hard because they are so close to the die and the air blows under the foil. The second time I applied heat, I left the thermal compound on and since I cut the hole in the foil smaller than the heat spreader, I stuck the foil down onto the thermal compound and it stayed stuck down when I used the heat gun and sealed the air out from going around the die.
To get the hole, I gently wrapped the card and pushed the foil down over the die. I marked where I wanted to cut with a marker and then took the foil off and cut the hole with a razor.
Comment has been collapsed.
Also, I don't know too much about electronics, but you may want to look the board over and see if there are any other signs of failure before you heat the die. I don't know what dead VRM components looks like, but you should be able to spot a bulging or leaking capacitor. If you see anything like that, heating will not help and you may be able to fix it with a few dollars worth of capacitors if that is why it failed.
Comment has been collapsed.
Thank you for the detailed account and advice. The capacitators look fine, so I'm pinning my hopes on heating the die. My heat gun should arrive sometime tomorrow so I'll report back after trying it out!
Comment has been collapsed.
I did it and it worked!
My first card is a Sapphire Radeon R9 290 Tri-X, and my second one is an MSI R9 290X, and by applying generous heat to the die for about 10 minutes each, I could get both cards running again. I really wasn't expecting too much lest I be disappointed so I am really happy with the result. It remains to be seen how long they will last of course, but both cards passed several 3D Mark benchmarks as well as two rounds of stress tests (20 loops each) as well as some tests with demanding titles like BATTLETECH and the Witcher 3 at 1440p. I have undervolted both cards slightly and that seemed to have lowered temps by a few degrees with no discernible performance loss. Anyway, I've decided that my Crossfire days are over and will just keep one card in the PC so that I have a spare one should the current one die again. Fingers crossed they will live for a while yet, though.
Comment has been collapsed.
That's great, I'm glad it worked! My first attempt only lasted a day, but I think that was because I didn't use enough heat. I figured I could always do it again with more heat and I had no idea how much heat was being transferred into the die or what effect it would have, so I wanted to take it slow. If it dies again, I'll try heating it again with a bit more heat. I am curious to see how long it lasts and how many times it can be brought back by heating it.
Good luck. Hopefully they will last you a while :)
Comment has been collapsed.
UPDATE after 9 days: The MSI Lightning that I had been using exclusively as my only card since the operation failed again. I repeated the process but heated it for a bit longer this time (approx. 12 min). Let's see how long it lasts this time.
UPDATE 2 (31/8): Died again after about a week. Repeated the operation and it's working again for now. Hmm....
UPDATE 3 (10/9): Had to perform two more resurrections. At this rate, it fails about once a week. Wonder how long I can keep this up. May just end up biting the bullet and getting a new graphics card after all. Hmm....
Comment has been collapsed.
Great read so far, making me think I can do something with this Radeon 7850 HD card I have that pooped out on me...
Comment has been collapsed.
It's an old card, so it's probably not worth selling for parts. It looks like they sell on Ebay for around $5-20 plus shipping.
If you have a heat gun and the card is dead, it can't hurt to try :)
Comment has been collapsed.
Well .... once (another GPU) baked in the oven but broke in two weeks XD
After that, the power supply of the PC broke, never to move again.
There is concern that the power supply was involved in GPU problems.
Attention to accidents in pleasant experiments.
A note at that time.
(Use an appropriate translation.)
Comment has been collapsed.
Oooh, interesting thread. Please keep us updated :)
Comment has been collapsed.
I will. I update my current game playtimes here pretty much every day and if anything major happens I will post an update. I don't usually bump the thread though, so you may have to remember to check back.
Comment has been collapsed.
In total I baked 2 GPU's in my oven with positive outcomes. The first one I baked was a laptop GTX675 MXM type video card. Before putting it in my oven the GPU stopped working all of a sudden when playing WoT and gave me BSOD's on restart afterwards. I decided that I had nothing to lose and therefore baked the GPU on 190 degrees Celsius for 10 minutes about 1.5 - 2 years ago. Until this day the GPU is still going and blowing out those frames per second without issues (I did apply new thermal paste before using it again). The other GPU which I baked the same way was a Geforce G210. Before, this video card showed artifacts upon startup and playing games. Right now, the card has steadily been working for about 2 months playing sims 3 on it xD. As of now, I hope that my Nvidia GTX980 SoC will last until I can afford another GPU! :)
Btw, the last thing I baked in the oven on 190 degrees Celsius for 10 minutes (TWICE this time) was my laptop motherboard (YES, a WHOLE MOTHERBOARD), also with success and again... because I had nothing to lose :). I bought the laptop in 2011 which came with a GT555m soldered on the motherboard itself. It worked fine until 2014/15 when it started to give me artifacts and eventually stopped working altogether. Only a few months back I baked my laptop's motherboard in my oven and as of now I can play games such as Just Cause 1, Brothers in Arms, Codename Panzers, Sims 3, Return to Castle Wolfenstein @ 1080p using the GT555m on my laptop without any issues or overheating (of course I did apply new thermal paste before gaming on it). Also, take into consideration that this GPU is an oldie, it will not be able to play 'modern' games.
This was my experience baking videocards xD!
Comment has been collapsed.
That's great! It's good to hear other people's success stories. There has been a lot of talk on the internet about how it doesn't work and people shouldn't do it, especially the Louis Rossmann videos, but really, if you have no plans to get the card professionally fixed and it isn't worth much to sell as parts, it can't hurt to try it, there is really nothing to lose. Especially if it is the GPU die that is dead from age because that is not really fixable.
I can understand them saying not to bake the entire PCB in an oven in your house because of the fumes and chemicals that remain, but there are ways to do it that are safe and I think it is a great experiment that is fun and can let people game again that can't purchase a replacement right away, even if it doesn't last for a long time.
I would really like to know what is actually happening inside the die when you heat it. I want to know exactly what fails inside the die and why heating it brings it back. I remember reading years ago about how voltage slowly kills a CPU and I think it had something to do with the electricity passing through the conductive material and slowly eroding it away, causing the path to become skinnier and skinnier until that path can no longer carry the current required to go through it. If heating the die actually melts the conductive material and basically "reflows" the paths inside the die, it could be more of a long term fix than many people claim. I really have no idea though and I am just experimenting because I think it is fun :)
Comment has been collapsed.
I would really like to know what is actually happening inside the die when you heat it.
I was also wondering about this, and contrary to the most widely proclaimed explanation that you are re-melting the solder, the actual science behind this seems to be something else if this iFixit guide is to be believed:
Re-flowing might work for some circuit-boards, but new more scientific data has revealed that most "quick, temporary" repairs like this are actually caused by expanding and contracting "Bumps" on the bottom of the surface mount chips such as the main GPU or the memory modules (which can happen at lower temperatures as well).
Comment has been collapsed.
I didn't think we were reflowing any solder that was used to attach the die to the PCB, I thought we might actually be reflowing an internal part of the die, like part of the transistors or the interconnects. I don't know what materials are used for this though and if they could melt at that low of a temperature. Other than silicon, if it is all made of copper, there is no way we can reflow that because the temperature would need to be extremely high. Maybe the internal structure of the die is still in good condition and that is not what fails over time from use. I really want to know what wears on CPU/GPU over time.
Comment has been collapsed.
This technique doesn't do anything to the inside of the die. It only works if the chip itself is still functional, but there is a connection problem to the board due to bad solder joints, displacement and misalignment caused by thermal stress, etc. Notice how "baking" and "reflowing" are only employed on GPUs and laptop chips, as these are directly soldered on to the PCB (via BGA). Nobody recommends reflowing a faulty desktop CPU as these are attached through much more robust sockets (LGA for Intel and PGA for AMD). Putting the die inside an oven will do nothing in this case. It's about the connection between the chip and the PCB. As you've probably heard, the proper way of repairing a bad BGA is by reballing the entire chip which requires specialised equipment and skill. The reflowing technique is a much simpler process, and baking the chip in an oven or with a heat gun is an even cruder approximation of the latter method. As I mentioned above, we're probably not even relying on melting the solder here but rather expanding and contracting the bumps on the bottom of the chip surface.
Comment has been collapsed.
That makes a lot more sense because I don't think there is anything inside the die that could be affected by a couple hundred degrees C. I remember Louis Rossmann talking about the bumps on the bottom of the chip and reflowing vs reballing, but I didn't really understand what he was talking about. I should watch the video again.
So, if the die isn't dead and it is just the solder under the BGA, it is possible to fix the GPU, but you need special equipment.
So, if the solder is going bad, what makes it go bad? I wonder if it is related to the voltage going through the chip, prolonged exposure to high heat or number of heat cycles. I wonder if really high temps increase the speed of corrosion or if a large number of heat cycles causes the solder to crack. If the solder was still in good condition, but just cracked from stress, reflowing should be a good fix and then you should be able to push the card just as hard as you previously were. If the problem is corrosion, then reflowing will probably only work for so long till it corrodes again. I don't know anything about how solder corrodes.
Now I am curious to know if I actually got it to reflow. Do you really think it would fix the card and make it work without any issues just from expanding and contracting the solder bumps a little? I find that hard to believe. If the problem is corrosion, I don't see how that would do anything and it the problem is cracked solder balls, will that little bit of movement really make the card stable again? If the solder melts at around 190C and I used a heat gun that was around 325-350C for a couple minutes, I would think that would be enough, but I don't know.
Now I am also curious whether it is better to heat from the die or from the back of the card. If I heat the heat spreader, all the heat has to transfer through the die, then through the BGA. I would imagine the die is much smaller than the area of the BGA where the solder is, so how well is the die going to spread the heat out to all the solder joints and would it actually be better to heat from the back to get to them all evenly. I would imagine there are copper wires going from the die to each of the solder balls, I wonder if the wires would be able to transfer enough heat to melt the solder.
So many questions :)
Comment has been collapsed.
I don't have the answers to all of your questions but this article should help clear things up a bit: https://en.wikipedia.org/wiki/Rework_(electronics)
Reflowing or baking are only considered temporary fixes because to properly fix faulty BGAs you really need to replace the entire solder and reapply the individual solder balls holding the pins as accurately as possible. I had my dad's broken Macbook Pro CPU reballed by a professional repair technician once, but it was worth the cost and effort because the entire device could be saved this way. For most GPUs that fail after several years, reballing is probably not worth it. You'd be better off just buying a new GPU instead of ponying up the cash for the job (or of course baking it! :)
Comment has been collapsed.
Here is an image of a GTX 580 PCB. Are all those little holes on the PCB the actual connections to the BGA? Do you think they go straight through to the BGA? Does that mean all those components that are soldered onto the back of the PCB under the chip are connected to the BGA? If that is the case, I would think they might fall off if I actually got it to properly reflow since I heated it from the top and they were on the bottom being held on just by the solder.
Comment has been collapsed.
I'm not sure if the holes are the actual connections (not sure about the layout, I assumed GPU dies had evenly spaced pins all the way, like CPUs) but the actual chip will have hundreds of pins for sure. As I quoted earlier, if the latest evidence is to be believed, melting the solder by baking may not be as crucial for success as just heating the component enough for the bumps to realign.
Comment has been collapsed.
I cooked my old 8500GT to install it over an old PC and it worked for almost a month and now the screen went pink back again. I'll do the burn once more but i'll wait someone to use the oven for something else to be worth.
It was cool to see it running in 2018 since it's a +10 years card and died like 8 years ago.
Comment has been collapsed.
Wow, do you game on it? I thought the HD 4850 I have been running is slow. Even though it was released only a year after the 8500 GT it has a passmark score of 1026 while the 8500 GT is only 137. I guess that is probably more do to one being near the top end for it's time while the other was more of a mid range. I have a hard time imagining gaming on something like that, but I guess it is possible considering I was running a Geforce 3 before the 4850 and that only has a passmark score of 5.
Good luck. I hope it comes back to life again :)
Comment has been collapsed.
My main PC have an AMD 260X - Which is pretty slow too but can deliver 50~60FPS on PUBG and play some nice recent (2017) with "decent quality". I also own a I7 3770 + 8GB DDR3 and a Generic MB hehehe. I mean, i'm running the stone-age here and i shall sell it all soon to get what is coming, but for now i just cant afford.
The 8500GT 512MB was a warrior when i bought it, i played GTA "Unoptimized" 4 when it launched with very silly graphics, but it saved me till i manage to buy a 5750 (1GB).
The cycle for my actual card already passed but there's nothing that motivated me to exchange it. Maybe when RDR 2 comes out i can make some effort. 😝
Comment has been collapsed.
I don't know, I have uninstalled it and I'm running the GTX 580 again. When I installed it, I just went to AMD's driver site and they had one, but now I don't see it. They only show the 4850 X2 now. It should be the same driver though, all the 4000 series cards list the same driver. Also, I am using Windows 7.
Comment has been collapsed.
Update bump: Card died and fixed for the 3rd time.
Comment has been collapsed.
Good luck! I have a dead 560 GTX that im tempted to try and revive. How long did you apply heat the 3rd time reheating?
Comment has been collapsed.
I held the tip of the heat gun pretty close this time, probably just under 1 inch from the top of the GPU die (heat spreader) and I heated it for 4 minutes. The heat gun is rated at 350C on low which is pretty hot, so I wasn't sure if I should do constant heat or remove it once in a while to let the heat spread out some and not create a spot that is way too hot, so I moved the heat gun away for 10 seconds after each minute to hopefully let the temperature even out a bit. So, I guess I heated it for 3m 30s over a span of 4 minutes.
Comment has been collapsed.
I've had to perform the fix four times by now on my R9 290X. That's about once a week on average. Wondering if I will need to bite the bullet and just get a new card soon...
Comment has been collapsed.
I would definitely be looking for a new card if you have nothing to replace it with. I am going to get a replacement, I am just watching used prices to try a get the best deal I can with the new cards being released next week. I also have an older card to replace it with right now if it dies completely, so I'm not really worried, I'm just having fun with experimenting.
If it only lasts you 1 week each time, I would imagine that means it's not getting hot enough. If the problem really is bad solder balls, getting it to actually reflow should fix it for a while, which is what I am trying to do, which is why I add more heat each time. If it keeps dying, I will keep adding more heat each time until something falls off the back of the PCB from too much heat. Then I will truly know I got enough heat in it to fully reflow the solder :)
Comment has been collapsed.
I do have a spare 290 to replace it with (also gone bad, and had to be baked), and I increased the heat and duration with each operation. Last time I could smell the fumes of the molten solder, so that was definitely hot enough. Anyways, I'm waiting for the new shiny Geforce RTX's to hit the market and send shock waves down the line for the older generation cards (which I am looking to buy) before I decide if I've finally had it with my old cards :)
Comment has been collapsed.
Maybe the things soldered to the visible part of the PCB are having their solder melted, but the solder balls under the die are sandwiched between the main PCB and the board that the die sits on which could be insulating those solder balls and possibly not melting.
I was thinking that if you actually got the solder to melt completely that the parts might fall off the bottom of the PCB. I wonder if gravity would pull them off or if the liquid solder would hold them on. From how easily components move in videos I have watched online of other people repairing boards, I figured stuff would just fall off the bottom if the solder melted. I was worried that if I heated the die from the top too much and enough heat got through to the back of the PCB to melt solder, the little capacitors and other surface mount devices might just fall off without touching them.
Comment has been collapsed.
Update bump for anyone still interested: Died again today and fixed for the 4th time. Used more heat than last time. Gamed for 183 hours on the 3rd fix. Total of 298 hours played since the original fix.
Comment has been collapsed.
Not a huge amount of hours by the sounds of it, but still pretty impressive for such a fix.
Comment has been collapsed.
I think it has to do with properly getting the solder to melt and reflow. I don't have any sophisticated equipment to set a specific temperature and I am pretty much just guessing on how long to apply the heat. The first time I tried to fix it, it died again after only 2h 45m of gaming, so clearly I didn't get the solder under the die to reflow. The 2nd time I used more heat and I was able to game for 113 hours. Then I used more heat again on the 3rd fix and I was able to game for 183 hours. Now I used more heat again and will see what happens.
Since it is dying reasonably quick, I don't think I ever got all the solder to reflow and more heat is needed, so I will continue to apply more heat each time until it stays working or completely dies. I am pretty just having fun experimenting with a GPU that I considered completely dead. It doesn't take much time to pull the card, remove the heatsink, heat the die, and reassemble.
Comment has been collapsed.
Thanks for keeping us updated. I really like seeing what's possible to revive a "dead" card.
I also voted for 1 year when you first posted the poll so it has to keep going. :p
Comment has been collapsed.
Update bump. Died again today. It only lasted 46 hours of gaming this time which is much less than the previous 2 times. Total gaming time since original fix is currently at 345 hours. Fixed for the 5th time with more heat again.
Comment has been collapsed.
Update bump: The GTX 580 is still working, but testing has ended as of 12/10/18 because I finally bought a new GPU. In the short time since the last fix, I only gamed for another 10 hours which brings the total gaming time since the original fix to 354 hours over 4 months and 12 days.
If you are curious, I picked up a new (sealed) GTX 980 Strix.
Comment has been collapsed.
Good luck with the new card; hope this one will last!
Comment has been collapsed.
Me too, especially with no warranty. The GTX 580 was more of a risk because I bought it used, but it still lasted 8 years. The 980 I just bought is 4 years old, but it was never used, so I am expecting it will probably last longer than I want to use it. If I pick up a 1440p monitor in the next couple years, the 980 probably won't give me the frame rates I want and I will want to upgrade again, so I am expecting to be looking for another GPU in about 3 years.
Comment has been collapsed.
Note - There are no giveaways here, just a fun poll.
Note - I didn't bake the entire GPU, I used a heat gun only on the GPU die.
TLDR at bottom.
I bought a used GTX 580 when it was 3 years old for about $115 shipped (great deal for a $550 card when new). I ran it somewhat hard and hot for 4 years since it was an overclocked blower style card. I didn't thrash it though, I kept it around 82C when 100% in games and it was running the stock voltage.
Then after a few years the card started to cause certain demanding games to randomly crash. I would dial back the overclock a little in those games to prevent crashing and then turn it back up for the other games. After about 6 months, I started playing a new game and it started to lock up and crash. I dialed back the overclock like normal, but it only lasted a few minutes before crashing again. I set everything back to stock and it still crashed, this time causing my computer to restart. When the computer rebooted, there was artifacting and color banding even on the desktop. I tried turning the voltage and clocks way down below stock to see if I could even get it stable in windows, but when I restarted the computer, it was all messed up even in BIOS.
This all happened over the course of about 20 minutes and I was sure the card was dead. It was a 7 1/2 year old card and I was happy with what I got out of it. I had been looking to buy a new GPU ever since I noticed the card was starting to get unstable about a year ago, but I didn't act quick enough and the prices went through the roof with the crypto mining craze. Knowing I wasn't going to buy a card anytime soon, I stuck my old Radeon HD 4850 (now 10 years old) back in since it's about twice as fast as the integrated graphics on my 3570K and I have been running that for the past 4 months.
I wasn't sure if I wanted to sell the GTX 580 for parts or just throw it out. I thought about baking it in the oven, but it didn't sound like a great idea to get everything that hot and I didn't want to put in the time and effort getting it setup right, getting the right temp and time, and having to deal with the fumes. Then today I thought, why not just only heat the GPU die with a heat gun? I pulled the card apart, wrapped it with foil, and cut a hole over the heat spreader on the die. No idea how long to heat it for, I just know my heat gun on low only goes down to 350C and that is too hot, so I just guessed and tried to get the heat into the die without melting anything else. I left the large heat spreader on the RAM and figured that would spread a lot of the heat out that got under the foil to hopefully not melt anything else while the die was heating.
After a couple minutes (not constant), I turned it off, let it cool, and put the card back together with no thermal compound to see if it was going to work or if I needed to try again and get it hotter. To my surprise, Windows booted and everything looked fine. I pulled it back apart, gave it a quick cleaning, applied some new thermal compound, installed it again, and now I am writing this.
I have not run a game or even loaded a youtube video yet, it has only ran at idle for about 30 minutes so far. I thought it would be fun to post a poll here and ask you guys how long you think the card will run before it dies again. I will run the card the same as I used to, as hard as I can without going over about 82C. I only have old games installed right now though because I was running the HD 4850. These games will not be able to push the card to 100%. I am playing Doom 3 and I just installed the first Witcher game. I will run basically at idle for a few more hours today and then I will start gaming on it tomorrow. I usually play about 2-4 hours a day.
7/29/18: I just started a game for the first time and I managed to play for 30 minutes without any artifacting or other problems. This is not the most demanding test though considering I am playing Doom 3. Even with the settings maxed on this 8 year old GPU, it only uses 10-30% of the GPU. Unfortunately, the game has to be capped at 60fps because the engine ties the physics and animations to the fps and anything higher will make everything in the game speed up.
Passed 2 hours now.
7/30/18, Dead again: I managed to play about 2h 45m of light gaming without any problems and then I was browsing the internet and the entire screen artifacted for a couple seconds before my computer restarted. After it restarted, I loaded a game just to see what would happen. It ran for a couple minutes and then locked up for a second which caused my GPU to be stuck in 2D mode. This is one of the things that happened a few times right before it died the first time.
I turned it off, pulled the GPU apart in about 5 minutes and took the heat gun to it again. I didn't want to waste more thermal paste on a dead GPU, so I left it on and just kept it clean. I didn't notice any negative affect on the thermal paste from the 350C heat. I heated it for a bit longer this time, but I'm not sure how much because I didn't time it again.
So I guess the winning choice for the poll is 1 to 10 hours gaming even though it was very light gaming and probably would have died quicker if I played a game that pushed it to 100% like i said I was going to do. I guess 1 day to 1 week also wins since it did last over 1 day.
I am not really trying to save the card, I am just curious to see what heating the GPU does, it is more of an experiment.
8/11/18: Passed 26 hours of gaming since the 2nd fix. I completed Doom 3 and I'm playing The Witcher now, which is also not a very demanding game. With settings maxed, it usually stays around 30-60% with rare spikes of 80-100%.
8/11/18, The real test starts: I don't think anyone is reading this anymore, but I will continue to update it because I think it is interesting. Up to this point I have played about 29 hours of low and moderately demanding games. I just installed Dirt Rally and the GPU cannot push as many frames as I want with the settings maxed. I have been running the factory clocks up to this point, but now I am running a moderate overclock (11%) to extract as much as I can from the card. While playing Dirt Rally the card stays close to 100% usage most of the time.
9/6/18, Over 100 hours played: I passed 100 hours played and these were the game times: Doom 3: 5h 45m, The Witcher: 26h, Dirt Rally: 69h. I have started playing other games and I don't know if I will continue adding up all the times. I may just keep track of how much time has passed since the fix.
9/11/18, Died and fixed again: The card died for the 3rd time, so I pulled it apart and heated it again. This time I timed how long I applied the heat, so I can make sure I apply more heat the next time it dies. I used the same thermal compound again for the 3rd time since there is no point wasting money replacing it if it is doing its job well enough, the card is already dead anyway and I am not having heat issues. In total I managed to game for 116 hours so far. That's 2h 45m on 1st fix and about 113h on the 2nd fix. I have a feeling I am not using enough heat to properly reflow the solder, maybe the 3rd fix will last even longer :)
11/7/18, Fixed 4th time: The card died again today. I was able to game for 183 hours on the 3rd fix which brings my total gaming time since the original fix to 298 hours. I heated the GPU for 5 minutes on this fix instead of the 4 I did on the last fix. I am using a heat gun on low (rated 350C), holding it about an inch above the die (heat spreader), and removing it for 10 seconds after each minute to hopefully let the heat spread a bit and not create too much of a hot spot. So, when I say 5 minutes of heating, it actually only has heat applied for 4m20s over 5 minutes. I have not removed or reapplied new thermal compound since the first time it died.
12/1/18, Fixed 5th time: I was only able to game for 46 hours on the 4th fix, which is much less than the last 2 times. Total gaming time since original fix is now 345 hours. I heated the GPU for 6 minutes this time. I am still using the same thermal compound which seems to be getting crusty now. It doesn't seem like it would work well, but my temps still look good. I wonder if the temperature censor is even accurate anymore after being exposed to that much heat.
12/10/18, Bought new GPU: The GTX 580 is still working since the last fix, but I have finally managed to get a new GPU, so the testing of the GTX 580 has stopped. In the short time since the last fix, I only gamed for another 10 hours which brings the total gaming time since the original fix to 354 hours over 4 months and 12 days.
If you are curious, I picked up a new (sealed) GTX 980 Strix.
TLDR: My GTX 580 was over 7 years old when it slowly died. I heated the GPU die with a heat gun and got it working again. I thought it would be fun to see how long it will last and document it here. The card died again after only 2h 45m of light gaming. I heated the card a second time with more heat this time and got it working again. I overclocked the card and started playing Dirt Rally which keeps the card close to 100% usage most of the time. The card died again after 116 hours of gaming and I fixed it for a 3rd time with more heat. I was able to game for 183 hours on the 3rd fix and then only 46 hours on the 4th fix. I fixed it for the 5th time, again with more heat, and the GPU is currently still working, but testing has stopped as of 12/10/18 because I bought a new GPU. The total gaming time since the original fix is 354 hours over 4 months and 12 days.
Comment has been collapsed.