Graphics Card Benchmarks
FOR THE MAJORITY of people, a good GPU is all about gaming, but that’s not the only area where they’re effective. In computational tasks with large datasets, and in video-rendering programs, having a good graphics card is key. To that end we’ve dramatically readjusted how we’re going to be testing these cards.
For synthetic benchmarks, we’re running the tried and true 3DMark tests, including Fire Strike for DX11 and Port Royal for RTX (or ray tracing), all at 1080p. What we’re looking for is a fixed figure to compare graphics cards across multiple generations without having to worry about resolution-based bottlenecks.
Moving on to the games, and this time we’ve picked four AAA titles. We’ll test each card with these games at a resolution that’s appropriate for the card itself. These titles are TotalWar: ThreeKingdoms, MetroExodus, Red DeadRedemptionI I, and Assassin’s CreedOdyssey, all of which have in-built benchmarks. However, with Assassin’s
Creed we will be swapping this out for the latest title, Valhalla, when that launches.
We’ve also added CompuBench 2.0’s N-Body Simulation 1024K, an OpenCL benchmark primarily used to showcase N-Body physics simulations. It spits out a number based on how many iterations/ sec the GPU can produce. These kinds of simulations mimic a system of particles and bodies under the effects of gravity in a vacuum, and are used in astrophysics with large datasets. We’ve also added FAHBench. This is another OpenCL test that mimics workloads assigned during Folding@Home tasks to identify how well a GPU can simulate molecular dynamics. On top of that, we’ve included the same Premiere Pro and After Effects benchmarks we’re using in our CPU testing.
For temperature and power testing, we’re going to pull figures during a run of 3DMark: Fire Strike @ 1080p. We’ll run a custom, repeating, combined test for 10 minutes, noting down max temp and watts drawn from the wall.