[SOLVED] Jitter and FPS drops in 0.7

Hey guys,

I've been playing Xonotic 0.6 a lot, it's such a great shooter, but after the update 0.7 it's unplayable for me. In 0.6 it ran completely fluid, 60FPS straight all the time, no FPS drops, most fluid gaming experience on the computer for me.

But in 0.7 the game jitters (mouse movement is all jittery) and the FPS drop to 30-45 FPS in many occasions. It gets worse with the number of people/bots/projectiles on the screen. Especially the open area in Glowplant can be really, really bad.

Specs:

Intel Core 2Duo T6600 @ 2x2.20GHz
4096Mb RAM
NVIDIA GeForce GT 220M (1024 MB)
Windows 7

It's a 3 year old Asus Laptop, which has never let me down. I can play Unreal Tournament 3 @60FPS, so it should be good enough to play Xonotic. Which it was, because like I said, Xonotic 0.6 ran smooth as butter.

Any ideas what could have happened and what I could do? Graphics driver is up to date.

Hmm. I've had the opposite experience. 0.6 ran like shit for me and 0.7 is much smoother. It still has it's occasional twitches but not nearly as bad as 0.6.

You have a very similar machine to my own, not sure what might be causing it :/

Have you still got 0.6? Set each of them up as a clean install in a different directory and try to compare like for like just in case there is something else wrong. Try running the Big Benchmark and looking for the inconsistences. It could be CPU side or GPU side and this should allow you to see this better. RAM shouldn't be an issue as Xonotic should be fine with much less RAM but it might be wort6h monitoring RAM usage just to make sure it doesn't page while you're playing.

It gets worse with the number of people/bots/projectiles on the screen. Especially the open area in Glowplant can be really, really bad.

That's a graphically hard area and one where a lot of texture memory is used but that shouldn't be an issue here. You could try running 'memstats' on the console in each of 0.6 and 0.7 with exactly the same view on screen and see what happens. There shouldn't be a difference which would mean texture memory could be discounted (what I expect).

Graphics driver is up to date.

Tried any other graphics driver? Is it a different version from what you were using in 0.6? Whether the driver is bang up to date doesn't always help performance, it's normally only a help with the latest chipsets, basically fixing all the bugs and poor optimisations in the early driver releases for that card. For a much older card you can in many cases see performance drops if you use the latest driver as opposed to one from say a year after the device was released.

Thanks for the replies!

Have you still got 0.6? Set each of them up as a clean install in a different directory and try to compare like for like just in case there is something else wrong.

Well, I've re-downloaded 0.6 from the site, tried the Glowplant Singleplayer level with the many bots and it ran as it used to in 0.6, no problems whatsoever.

That's a graphically hard area and one where a lot of texture memory is used but that shouldn't be an issue here. You could try running 'memstats' on the console in each of 0.6 and 0.7 with exactly the same view on screen and see what happens. There shouldn't be a difference which would mean texture memory could be discounted (what I expect).

I did that without bots and there was no noticeable difference. I should have said this before, my 0.7 runs smooth as long as there are no other players or bots. So, I suspect it has something to do with player animations, probably with the new animation blending that was mentioned in the 0.7 release notes:

"Animation blending
Our last release ushered in a new wave of possibilities with client-side rendering of players which allowed us to implement full animation blending to playermodel rendering, among other unique features. Because of this we can now play two separate animations on a player at the same time. This means we can show a specific walking direction on the legs and play an attacking animation on the torso simultaneously. This implementation also lets us bend the playermodel skeleton based on the direction they are looking, which provides a seamless appearence of where other players are aiming."

-> maybe it has something to do with this? The performance gets worse, the more players there are on my screen.

Tried any other graphics driver? Is it a different version from what you were using in 0.6? Whether the driver is bang up to date doesn't always help performance, it's normally only a help with the latest chipsets, basically fixing all the bugs and poor optimisations in the early driver releases for that card. For a much older card you can in many cases see performance drops if you use the latest driver as opposed to one from say a year after the device was released.

Well, I updated my graphics driver because of the trouble I had in 0.7. I thought it might help, but it didn't. There was no difference. Before the update I ran 0.7 with the same graphics driver, which I have been using for 0.6 all the time.

I would jump to so much of a conclusion on the player model animations as it shouldn't really add much load. Player animations themselves would not add to the graphics load at all and CPU load is minimal. I have tried Xonotic 0.7 on systems a far slower CPU than yours without slowdown so I would not be tempted to think of the animations as the problem. That is unless something works massively inefficiently in some certain circumstance.

When more players are on screen it means there is far higher graphical load as well because of their complex geometry and high resolution textures. For example if you run out of texture memory this becomes the most obvious area of poor performance. As I suggested before, try running The Big Benchmark in both 0.6 and 0.7 and see if there is a difference in performance:
http://forums.xonotic.org/showthread.php?tid=2562

That benchmark is quite player heavy so expect to see slowdowns to appear if they are a problem. Also monitor RAM usage if you can for 0.6 and 0.7 while you play. This would allow you to exclude RAM usage from the equation at least.