Furmark is a popular benchmark tool for graphics cards. If you are not familiar with it, you can check it out here.

One of it’s great features is the ability to push the GPU to full load, which can really heat things up. Another is the timed benchmark, which provides a score the can be compared to others on their website. The concept is nothing new, FutureMark has been doing it for many years.

Obviously higher scores are better, but does the score really tell you anything useful? Do Furmark scores matter? Take a look at the comparison below for the surprising answer.

Is Furmark Safe?

First, there are warnings one might see before testing their graphics card with Furmark. It begs the question, “is Furmark safe?” Well, I have never had any sort of problem pushing my GTS 250, GTX 260, or GTX 470 cards from EVGA. In fact, even at sustained 100% load, the temperatures on my cards never get too hot.

Is Furmark Accurate?

What you should really be concerned with is the comparison of scores. I’m not just referring to the sometimes outrageous scores posted on their website. The scores you get yourself are misleading.

Solid Statement: Don’t trust Furmark scores when comparing anything but the same card.

There you have it. What does that mean? That means you shouldn’t compare one card’s score to the score of another card. In a perfect world, a higher series card would get a higher Furmark score, but that just is not the case.

GTX 260 Core 216 55nm Validation : 2837 Score
GTS 250 65nm Validation : 2902 Score

A GTX 260 Core 216 55nm packs a lot more punch than a GTS 250, but if I rely on the Furmark scores, the GTS 250 is a better card. The trick here is that these 2 tests are actually running on slightly different drivers. The GTS 250 is running 197.34 and the GTX 260 is 196.21. Not only does Nvidia tweak game performance when releasing new drivers, but they also specifically tweak for these benchmarks. Furmark’s lab-like results are totally consistent, but can mean nothing when compared to other cards. That is why you should only compare the same card with Furmark (like stock score vs overclocked score).

A Better Idea

For a more accurate depiction of performance across different cards, use a tool that better reflects actual gameplay, such as the FarCry2 benchmark. While it’s scores are less consistent (because the scene it generates varies slightly), you can run multiple loops and throw out the oddities. It will give you a more accurate comparison of one card versus another.