Whenever I get a new graphics card, I'll install recent drivers and make sure HWiNFO is running (with graphing turned on) to record highs and lows.
I generally run one pass each of Unigine Heaven, Valley, and Superposition, mostly to check that A.) the thermal solution is working, and B.) performance is within the ballpark of expectations. The three benchmarks do not push the graphics card exactly the same way so it's worth trying all three at least once. However, these three benchmarks do not include newer GPU technology.
Running the built-in benchmarks for
Rise of the Tomb Raider and
Shadow of the Tomb Raider makes sense since I own both games and those are very close to actual loads.
If it's an GeForce RTX card, I'll use the Nvidia DLSS demo to make sure the Tensor cores work and Unity Enemies to make sure the RT cores work. The 3DMark TimeSpy benchmark is not a bad thing to try out. There are a few other RT and DLSS benchmarks out there, I think the
Bright Memory Infinite one is free.
If the card is running satisfactorily, I might use the Unigine ones to configure fan curves for noise otherwise that's basically it for any GPU benchmarking.
More recently, there's one more test for GPUs: video upscaling software like
Waifu2x Extension GUI. This isn't really a benchmark. This is a real world usage case for me these days, just like Handbrake encoding.
