Video also hosted on the youtube channel, please check us out there!
This article is intended as a written piece to compliment the video. Please watch the video for more details, images, and other information.
Recently I had been having some issues with some of my older GPUs. Specifically, the cards in question were my old R9 290s, which have always run hot, but now seemed to be hotter than ever. Something wasn’t right, with both cards sitting right at the thermal throttle limit of 94c with an undervolt and very good thermal conditions (one in open air, the other in a very good case for airflow, the Silverstone FT-05). Nothing had changed with the cards, and nothing seemed to be very different, the coolers were not damaged, there was no visible issue with the fans, the case that each was in was much better than the original cases, no apparent issues. The sapphire card was one that was interesting to me more so than the gigabyte. When I first got the card, I took the shroud off to paint it, and in the process of doing this replaced the VRM thermal pads with high quality ones, as well as swapped the thermal paste out for some good old Arctic MX-4. The gigabyte had problems with thermals before, but those had also been solved with redoing the thermal paste, however as the card had passed from one person to another I was unsure of the last time the paste had been done. I had helped with the first time, and then knew it was done a second time. I believe it was also done a third time by someone else (who bought it after the first owner, and then sold to me), but I can't be sure about that. So that left very little room for any kind of error. Were these 200 series cards finally at the point where they were too hot to keep at even stock clocks? I didn't think so, so we decided to swap the thermal paste out and investigate.
Swapping the thermal paste is an easy task on either of these two cards. Both were very straightforward to take apart, just like any other GPU. Beginning with the Sapphire, the thermal paste was bone dry. There was a pretty good application, maybe a little light, but god damn was the paste dry. After scraping that off and wiping the rest off with some isopropyl alcohol, blowing the dust of as best I could (no canned air) then we slapped it back together. We moved on to the gigabyte, which had almost no thermal paste at all, out of the gate this was going to be an issue. The thermal paste was even drier than the sapphire, and the spread was god awful. On top of this, after cleaning off the die, there was a scratch or a gouge out of the card, this is visible in the video. I’m not sure if this is what has lead to the temperatures always being worse on this card, or simply cooler design, but either way this is a major damage to the die that I've never seen before. This card has always been a bit above other 290s, even of the same type (compared to reviews) so it may be that it was simply defective from the factory and never returned. I’m not sure if this was caused by a screwdriver and an ass hat, or some kind of improper mount job, or even if the cooler somehow managed to react with the die, but we went ahead and replaced the paste, making sure to use plenty this time. With the thermal paste change out of the way it was time to see if this had solved the issue of high temps, or if these cards really were pushing 94c all the time.
Performance
After the change, fired up the same mining tests and lo and behold the temperatures were much improved. On the sapphire temperatures dropped from 94c down to 73c, the gigabyte dropped from 94c all the way down to 69c. Even though the sapphire has a better cooler and has always outperformed the gigabyte, the gigabyte is mounted in an open air environment, so that explains the differences there. Both dropped over 20c, which really goes to show that thermal paste can make a difference, and that this is something that should be done fairly regularly. I’ve never experienced a drop as severe as this though, and with both of the cards being bought around 4 years ago I never expected that the paste would dry up so quickly.
Final Thoughts
So what's the take away from this? Well, it would seem that thermal paste drying up can contribute a great deal to higher temperatures. No shit. However, the big wake up for me was that the Arctic MX-4, commonly recommended for this and a paste that I have never had any issues with in the past had managed to dry up and cause such a huge performance delta in just 4 years. I’ve had a number of CPU make it well over 3 years with virtually no change in temperature, so I was amazed to see the difference in the GPU temps. Some of this may be that these cards have both been either mining or gaming for a number of hours a day pretty much daily since purchase, often overclocked, but even then I would have expected the paste to hold up better. I’ve actually been forced to rethink my advice on thermal paste, and in light of this in the near future will be trying out some other thermal pastes to see if they make a difference. I don't want this to sound like I'm ragging on MX-4 though, until now I have never had any issues with the paste and outside of this instance I have no reason to believe that the paste is bad. It has always been a breeze to install, cheap, and performed well. It could be that I got a bad batch, or there could be something else going on. Either way, it’s worth noting. As a takeaway, I’d just like to highlight the importance of thermal paste. If you’re experiencing temperature spikes or higher than average temps on an old card, pop it open and take a look at the paste. It could be a cheap and easy way to shave some degrees off your card and get a few more years out of it.
