LinusTechTips shared a new video recently showing the advantages of buying second-hand GPUs right now, even if the cards were used for mining. The video's goal was to disprove the growing ideology that mining cards are bad purchases altogether and show how mined GPUs can be in as good, or even better condition compared to used gaming GPUs.
For testing, LinusTechTips bought a plethora of used graphics cards last year, that were desirable for mining purposes at one point in time. Models range from AIB partner model RTX 3060s, all the way to RTX 3080 10GB cards, as well as AMD's Radeon RX 5700 XT GPUs. All models were compared against lightly used counterparts LTT had in its possession.
With these GPUs, LinusTechTips used Kombuster to test GPU clock speeds and silicon quality, then used several triple AAA games running at 1080P, 1440P, and 4K to verify GPU performance. Any GPU that was operating 3% under the reference cards was considered a faulty card.
Overall, nearly all of the 19 cards passed, including all of the used RTX 3070s, RTX 3080s, most of the RTX 3060s, and most of the RX 5700 XT GPUs. The only issues the outlet had had was with two cards, a Gigabyte Eagle RTX 3060 OC and a Sapphire Pulse RX 5700 XT.
The 3060 was a complete failure, featuring incredibly low clock speeds due to temperatures surpassing 100C. Linus went so far as to test the VRAM reliability of the card, and it failed the test completely. The RX 5700 XT was not nearly as bad, only failing the Kombustor test by having 3% or worse clock speeds compared to LTT's reference RX 5700 XT model. But in gaming, the GPU performed perfectly, with effectively identical performance compared to all of the other RX 5700 XTs that were tested.
LTT doesn't explain the reasoning behind the clock speed issues in Kombustor, but it may simply be that the Sapphire Pulse model has a reduced power target, which was only visible in Kombuster and not the gaming tests. This wouldn't be surprising considering the Sapphire model ran the coolest out of all the RX 5700 XTs.
Besides these two cards, all other GPUs performed flawlessly and were effectively performing as well as when they were new. Again, it's worth noting that all these GPUs were probably used by miners at some point in their life cycle.
So, at least according to LTT's testing, it appears that buying a second-hand mining GPU isn't as dangerous and dodgy a proposition as it might appear to be. Still, make sure wherever you're getting the card from has a return policy in case it arrives with performance issues.
There is definitely some merit to LTT's testing, but keep in mind the sample size is pretty small. And we've also heard some counterarguments directly from Nvidia's partners regarding mining GPU degradation. Graphics card manufacturer Palit claims that 24/7 mined GPUs could have reduced performance by around 10% per year. We don't have the details to back up Palit's claims, but it's definitely something to consider.
Either way, it is always best to make sure you know as many details as possible about the graphics card you're looking into buying before jumping into a purchase. It's impossible to verify how well-treated a used GPU was in its previous home, and whether or not it will retain the same performance as when it was brand new - whether or not it was used as a mining card.
For our take on whether or not you should buy a used GPU, check out our guide here.