Why We Will be Testing GPUs “Inside The Second”

I am just finishing up some testing for a video card review tonight – we’ll be shifting some focus away from the latest, greatest, most expensive cards around, to see which is a viable solution for a sub-$200 video card from AMD or Nvidia. If you haven’t noticed, we haven’t really spent a lot of time reviewing video cards in recent years. The main reason for this is, if I don’t think I can do an extremely useful job at reviewing hardware, I won’t do it. And for the longest time, I didn’t think I could test video cards properly, in a way that is extremely useful for you, the reader. As easy as it is to breeze through some bench marks and show you the resulting Frames per Second data, I don’t think that is useful to the end user. It is also why we don’t test power supplies here – we just don’t have the equipment to do it properly.

Like many hardware junkies, I came across Scott Wasson’s brilliant “Inside the second” article /slash epiphany. He seemed to be having similar frustrations, particularly with SLI configurations. He realized that Frames per Second doesn’t tell you the whole story – “60 frames per second” isn’t always smooth, especially when you take SLI micro stuttering and vsync into consideration.

Scott, being the brilliant man that he is, developed a way of looking “inside the second” and getting the most detailed look at how video cards perform possible. I urge you to read that article if you haven’t yet.

Needless to say, we have employed his method of video card testing, and I urge other hardware reviewers to do the same. In this example from our upcoming review, you will clearly see why it’s so important:

Here we have a Witcher 2 test run with both video cards. They look to perform about the same, at about 40 fps with the “High Spec” preset in 1080p. If anything, this chart tells you that both cards should probably turn the settings down a bit to play at this resolution, because 40 fps is really choppy, right?

Look at what happens when we look “inside the second”

This is the exact same benchmark run, but it paints an entirely different story. What we are looking for in a chart like this is a line that is as tightly-packed as possible. That means that there is very little lag between each frame. In this case, the GTX 560 manages to offer a relatively smooth performance outside nine or so really laggy frames out of 3000. Other than that, it is running smoothly at 20-30ms per frame, or around 40 fps as the first chart showed us.

The Radeon HD 6870 on the other hand, although spits out a similar “frames per second” result as the GeForce, has a lot more lag between each frame – one frame will be an expected 16-20 ms, while the next will be 30-40 ms and higher, with quite a few lag spikes throughout the test run. Subjectively, the Radeon felt much more ‘laggy’ even though they were both running at “40 fps”. In fact, the Witcher 2 launcher recommends a lower preset when the Radeon is installed.

Just a note – not all benchmarks look like this – it is just one extreme example. Watch for our full review coming soon, where we will also look at overclocking, power consumption, media encoding, and OpenCL performance of these two cards. They aren’t the newest video cards out there, but they won’t break the bank, and for many people, they may be enough.