Is glxgears an accurate measure of 3D performance?
No, it sucks in multiple ways. But it is readily available on all computers and a significant number of people find it both interesting and better than nothing. And, despite all the complaints, you'll generally find that faster 3D cards achieve higher glxgears frame rates. A Radeon X850 with the R300 driver really is faster at most 3D operations than an nVidia GeForce FX Go5200 with the nv driver. Because glxgears doesn't take into account many types of 3D operations, you'll find plenty of cases where a card with a fast glxgears frame rate doesn't perform a specific 3D operation as fast as a card with a slower glxgears frame rate. So, if you have very specific 3D acceleration needs, you'd be crazy to base your purchase on glxgears results.

Why aren't you using a more accurate 3D benchmark?
There isn't one that's readily available. When I started this site there was no significant data available on the relative performance of free software 3D drivers. But, because there is no readily available 3D benchmarking software included with X.Org, the choice I was faced with was to create a listing of glxgear results or do nothing. Most people I talked to preferred having glxgear data to no data at all.

When will you stop using glxgears as a benchmark?
When we can replace it with something more accurate. A suitable replacement needs to be a benchmarking utility that is included in the standard X.Org distribution, available on all common Linux and BSD distros, and as easy to use as glxgears. Until something like that comes along, we'll stick to glxgears.

Will you stop using glxgears if I send you an email complaining that glxgears sucks as a benchmark?
No.

What if I explain in great detail all the ways in which glxgears sucks and give examples of cases where it's inacurate?
Nope.

Is there anything I can do to convince you to stop collecting glxgears results?
Yes, write a better benchmarking utility and get it included in the standard X.Org codebase. If people spent half the time working on a decent benchmarking utility as they do complaining about glxgears, we'd have one by now.

Has anyone published accurate benchmark results of a wide range of 3D hardware and drivers and compared the results with glxgears frame rates of the same cards to authenticate the claim that glxgears sucks?
Not that I'm aware of. It shouldn't be hard to do and I'd love to see such a test done. If you do it, send me the URL and I'll add a link to the site.