In the Cryptocurrency Era, Approach Used GPUs With Caution

In the Cryptocurrency Era, Approach Used GPUs With Caution

We’ve written a number of articles on the current lousy state of the GPU market and what you can practically do about it, ranging from optimization guides for existing cards to whether AMD’s Ryzen 5 2400G can serve as a plausible low-end gaming solution to tide you over if you need a new card. But there’s another option available — the used GPU market. Does it provide a reasonable alternative?

The short answer is: Maybe. But you’re going to want to be extremely careful before you pull the trigger.

The problem with buying a used GPU off an unknown seller is that you’ve got no idea what the card’s operating conditions were. While this has always been true, cryptocurrency mining puts unique stresses on GPUs compared with conventional gaming.

First, cryptocurrency miners typically run 24/7, while most gaming GPUs don’t. Gamers also tend to understand and prioritize good cooling, while cryptocurrency miners tend to maximize performance first and worry about cooling later. Four GPUs that are perfectly capable of cooling themselves individually can run extremely hot when put loaded into a chassis at the same time. GPU miners may also have overclocked a GPU or tweaked its voltage for stability. While this is also a risk when buying a used gaming card, again, we’re talking about the difference between 24/7 operation and occasional use.

eBay offers certain protections to guard against being sold defective equipment. But just because hardware works the day you receive it doesn’t mean it won’t fail in a month or two, or display erratic crashes when loaded in a very specific fashion. We recommend buyers looking for a used card prioritize owners who specifically state that the GPU has not been overclocked or used for cryptocurrency mining.

Still, this advice relies on the original owner being honest. What can we do to avoid relying on the owner’s honesty?

Simple: We can test the card. But in order for the tests to be effective as diagnostic criteria, you’ll need to test your current GPU first.

It’s common to see people recommend benchmarks like Furmark for GPU stress testing, but I’m going to disagree with that assessment. AMD and Nvidia both scan for thermal viruses like Furmark and typically prevent them from pushing GPUs as hard as possible. Instead, I’m going to recommend you run your tests in games you already own, either by looping built-in benchmarks or with play-through tests.

The following slide show is from our RX 580 versus GTX 1060 test. Of the games listed, all of them except for Doom have built-in benchmark modes. One game, Metro Last Light / Metro Last Light Redux has a benchmark you can loop for n number of tries.

To start this process, you’ll want to download the same driver you’ll use for your new card (this does assume that your current GPU if still compatible with modern driver releases — if it isn’t, omit the initial driver removal and reinstallation steps on your current GPU). Uninstall your old driver and then run a utility like DDU (Display Driver Uninstaller) from inside Windows Safe Mode to absolutely remove all trace of a previous driver install. Next, install the new driver from a standard boot of Windows. With the new driver in place, fire up your game or benchmark of choice and start playing. Loop tests or game for 15-30 minutes, and make sure you test more than one title. It’s absolutely possible for a dodgy card to be stable in five titles and badly unstable in the sixth, so you’ll want to test a selection of games.

What we’re doing here is establishing a baseline for system stability. When you receive your used card, uninstall your GPU driver using the instructions above and reinstall it alongside the new card. Run the same benchmarks and gaming checks again. If your “new” card is stable and your PSU is appropriately specced for your GPU, you can be reasonably confident that the card will perform properly thereafter. This process doesn’t eliminate the chance of a longer-term failure, but it does offer a test sequence you can perform fairly quickly to ensure a GPU isn’t broken out of the box.

Also, always make sure to check current used prices against current new prices. With the speed at which GPU pricing can change these days, that’s the only way to make sure you won’t end up paying more for a used card than a new one. Also, tip of the hat to Tom’s Hardware, which has some additional tips on evaluating whether a GPU is safe for purchase.

Continue reading

Intel Launches New Xe Max Mobile GPUs for Entry-Level Content Creators
Intel Launches New Xe Max Mobile GPUs for Entry-Level Content Creators

Intel has launched a new consumer, mobile GPU — but it's got a very specific use-case, at least for now.

MIT Creates Battery-Free Underwater GPS
MIT Creates Battery-Free Underwater GPS

GPS radio signals dissipate quickly when they hit water, causing a headache for scientific research at sea. The only alternative is to use acoustic systems that chew through batteries. A team from MIT has devised a battery-free tracking technology that could end this annoyance.

Intel Details XPU Strategy, Launches New Server GPU, OneAPI Gold
Intel Details XPU Strategy, Launches New Server GPU, OneAPI Gold

Intel made a pair of announcements today regarding its OneAPI initiative and the launch of its first server GPU based on Xe graphics.

Nvidia Unveils Ampere A100 80GB GPU With 2TB/s of Memory Bandwidth
Nvidia Unveils Ampere A100 80GB GPU With 2TB/s of Memory Bandwidth

Nvidia announced an 80GB Ampere A100 GPU this week, for AI software developers who really need some room to stretch their legs.