Re-Evaluating Reviews: Component Testing

A quick glance at the benchmarks we’ll be using when we test hardware this year

NEXT UP ON the agenda is component benchmarks. We typically don’t change a lot when we tweak these—mostly just new game titles, and the like—but it’s often a good idea to take a step back, re-evaluate how relevant these are to genuine everyday use cases, and see whether there are any new benchies worth taking a look at.

Each set of benchmarks has its own unique permanent test bed. We can’t list them here, because there’s way too many. We always try to keep specifications similar between them where we can. For instance, our mainstream CPU tests always feature 16GB of DDR4 at 3,200, a GTX 1080, and a Samsung 850 Evo 500GB SSD (one SSD with a unique install dedicated to each platform), regardless. Jump up to X299 or TR, and we double the memory to allow for quad-channel, but still retain the rest of the spec. This helps ensure that we can easily compare figures across platforms.

So, what changes have happened this year? Well, they all stem around optimizing, and looking for bottlenecks. In our processor tests, we’ll be using 1080p game titles, to try to hunt for those elusive high-end frame rate bottlenecks. For motherboard testing, we need to make sure the GPU and CPU are running in tandem efficiently, so 1440p is our go-to for that. For GPUs, we actually test our cards at all three resolutions, then report the figures for the resolution that makes the most sense.

We’re also adding a new price-toperformance metric, for both CPUs and GPUs. This will fluctuate depending on the time of writing, but is calculated as the 3DMark: Fire Strike or Cinebench R15 Multi score, divided by the price the component is currently available for on Newegg. So, the GTX 1080 scoring 17,704 would get 21.3, while the GTX 1060 3GB with 10,251 would get 27.0; higher being better value for money.