glxgears stutters in Fedora, not in Ubuntu

[kdorf]

Originally Posted by PilotJLR

Although I haven't used glxgears, disregarding the OP's direct comparison is asinine.

I agree. If you're not going to help out and say something useful then don't say anything. Anything can be used as a benchmark. Some things may be less useful than others, obviously, but the OP's point was very valid: glxgears should have roughly the same performance in both distros.

In any case, I had posted another thread before I noticed that this thread had good info in it (and not just some jerk saying glxgears isn't a benchmark). I'll try passing nopat to my kernel when I get home and see if that remedies. I'll edit this post/reply back here when I find out.

Thanks guys.

[leigh123linux]

Originally Posted by PilotJLR

I had experienced the exact same issue as the OP... passing "nopat" to the kernel has resolved this problem!

Although I haven't used glxgears, disregarding the OP's direct comparison is asinine. I have used a variety of applications (X-Plane is one) that performed better under Ubuntu 8.X than in Fedora 9/10. Nopat is clearly the reason... adding nopat as a kernel argument has fixed the jitter and poor framerate I previously experienced.

Thank you for those who recommended this fix!

You don't need to use nopat with the 2.6.29 kernels from Koji , IMO 2.6.29 is much more stable than 2.6.28 or 2.6.27.

[mooninite]

Originally Posted by kdorf

I agree. If you're not going to help out and say something useful then don't say anything. Anything can be used as a benchmark. Some things may be less useful than others, obviously, but the OP's point was very valid: glxgears should have roughly the same performance in both distros.

It's not a benchmark of any sort. Stop trying to justify it. You have no experience with performance deltas or programming metrics so exit stage left, please. There's reasonable and sound information about why glxgears is not a benchmark. Trying to refute such info with a reply of "well you're a jerk" does not justify its use.

[kdorf]

Originally Posted by mooninite

It's not a benchmark of any sort. Stop trying to justify it. You have no experience with performance deltas or programming metrics so exit stage left, please. There's reasonable and sound information about why glxgears is not a benchmark. Trying to refute such info with a reply of "well you're a jerk" does not justify its use.

And you, apparently, have no experience with what I like to call the dictionary. In case you missed it, the whole idea was to establish a point of reference (look up benchmark in the dictionary). It may be a very basic point of reference and it might not completely test all of a card's features but it is something. Nobody here is trying to say that card A is better than card B because of a 50 fps difference in glxgears. That's what the "real benchmarks" you describe are for.

In my case, I noted my system had better significantly performance in Ubuntu than Fedora. The FPS provided by glxgears only reinforced this observation. Regardless of whether or not you're using a "real game" or a "real benchmark" the performance should be roughly the same across the board. Whether or not it's a "real benchmark" doesn't have all that much to do with it.

I am happy to report, though, that nopat seems to have resulted in performance gains here on Fedora 10. I can't verify whether or not leigh123's claim about kernel 2.6.29 is correct... I'm running 2.6.27. And in case you're interested, yes, my glxgears FPS went up by around 2000 as a result of nopat, about where it was on Ubuntu if I remember correctly.

[tzp]

There is some piece of truth on both sides, so you even might stop quarrelling.

It is true that glxgears does some simple things that are not typical for a 3D application, so it can be considered not representative.

But, if the slowdown is too big then other applications may be impacted too.
(Nevertheless, it may happen that the glxgears fps increases, but you still see no benefit for real 3D applications.)

I don't want to repeat my post again, so click here .