R600c/g benchmarks

[pingufunkybeat]

Yeah it's become a standard because the mouse polling rate was 125fps. Which sort of become the defacto standard.

No, it became standard because it maximised the rounding error.

http://www.funender.com/quake/articles/fps.html

Another magic number is 333 fps and there are others. 125 is pretty close to optimal, though, and most people could achieve it, so it became standard.

At 125 FPS, the rounding errors at each frame (done to conserve bandwidth during multiplayer games) add up in a way that lets you jump higher and accelerate faster.

This is a very unfortunate bug, but it is not really cheating, because EVERY frame rate has different physics. You always have advantage over some players and disadvantage over others. At 125Hz, everyone has the same physics, and it is the "best" one.

[b15hop]

No, it became standard because it maximised the rounding error.

http://www.funender.com/quake/articles/fps.html

Another magic number is 333 fps and there are others. 125 is pretty close to optimal, though, and most people could achieve it, so it became standard.

At 125 FPS, the rounding errors at each frame (done to conserve bandwidth during multiplayer games) add up in a way that lets you jump higher and accelerate faster.

This is a very unfortunate bug, but it is not really cheating, because EVERY frame rate has different physics. You always have advantage over some players and disadvantage over others. At 125Hz, everyone has the same physics, and it is the "best" one.

Yes you're right. It was a mathematical rounding issue and 125fps was the solution at the time. I guess as you said, floating points, especially higher bit floats, have less room for error. Eg int bitshifting causes a huge loss if done several times.

[yotambien]

The thing is that I can tell between 90 fps and 125 fps in oa/q3. I don't understand why because the monitor refresh rate is 60, but still, 90 sucks while at 125 everything is more fluid. Note that I'm not talking about the difference in physics (which obviously is present), but about how the game feels.

[pingufunkybeat]

^^ Mouse input, and shorter latency.

If it oscillates, you notice it even more.

I'm not terribly bothered, as long as I get flat 60fps, synced, and 125-fps physics, which is needed for many levels.

[b15hop]

Frame rate can be a problem at a closer level. For instance in fallout 3, the game feels chuggy at 30fps because there is a lot of micro stutter. Micro stutter causes the game to render inconsistent frame rates. Lets just say 15 of the frames are done 30ms but one or two frames take 150ms to render. And this occurs every several fast frames. Things like that can make even a 100fps game feel chuggy. So as always, it's the lowest frame rate, that matter the most. Not the millions of frames rendered in less than 10ms.

[yotambien]

Yes, if it oscillates it's just unbearable. That was the reason why I resinstalled a Windows partition, no matter how I lowered the graphics details both fglrx and radeon used to give me anything from 80-125 fps.

As for the ioquake3 fps independent physics, I'm not sold. At this point it may be psychological, but it doesn't feel the same. I prefer the legacy mode and capping the frame rate.

[yotambien]

One thing, the "bug" that allows strafe jumping in q3 is the same "bug" that makes the physics dependent on the frame rate?

[pingufunkybeat]

I think it's a different one.

The strafe jumping relates to the bug that if you change direction abruptly, the engine does not limit your speed to the normal running speed (320 ups) immediately, but after a short delay. Not sure why, but it is probably tricky to detect and cap.

If you move straight, you run at 320 ups. If you strafe left, you run at 320 ups. If you suddenly combine them, they add up to more than 320 ups, and the engine takes a short amount of time to detect and correct this.

I think that this also has to do with rounding at some point, but I'm not sure.

I know that Carmack fixed it once, and all the players cried murder, so he put it back. FWIW, the movement is the best thing in Quake games, and it's what makes them so much more interesting than the walk-walk-walk-shoot-shoot-shoot borefests that most other games are for me.

[b15hop]

I like how F3quake put it. If your position is calculated every frame, then more fps means more physics calculations per second. Thus the game feels smoother. The problem here is that older game engines use integers for many variables and that is great for saving bandwidth but bad for physics.

yotambien: fps independent of game physics is only good if it doesn't vary too much. Otherwise low fps still makes the game unplayable. Ie 125fps physics and 10fps rendering. :P

[b15hop]

125fps is a magic number in the quake engines, interesting. You can jump higher, run faster and play better just due to a rounding issue with integer variables. When John Carmack fixed the "bug" it made the game suck lol. That's funny... =D