re: Master Race -- Excellent Article On The Myths Of Graphics Cards

God damn this is why Tom's Hardware is such a joke. Just read the bit about whether overclocking always improves performance... What a shocker.

This is the same site that reviews a motherboard's "overclocking performance" by pressing the auto-overclock button.

quote:

---

Performance Is An Envelope, Not One Number

---

Performance is a thumbtack, and a lamp, and a Pepsi bottle.



Nice way to overcomplicate it. Yes, raw frame rate should not be the sole determining factor, but you also don't need to evaluate each individual video card using a three-dimensional graph.

If you use this "envelope" system, you can't just look at, for example, a generic GTX 780; you have to evaluate all of them from every AIB partner because the noise levels will vary across coolers.

And it doesn't stop there. The "Quality" component of that graphic is useless. That should be a constant. That's why comparative benchmark data shows raw frame rate at constant quality settings across various games. And we all know that different games require more resources than others, and further, these games may require different types of resources, which is why you'll often see two similarly powerful video cards perform differently in certain games, and weaker cards sometimes outperform stronger cards in a select few games. And then it varies even more when you bring different resolutions into the mix. If you don't have your resolution/quality requirements and a general list of desired games in mind before you start comparing cards, you're doing it wrong.