General

Re: XNA Games 'jump', but professional games do not.

Nick Gravelyn — Sun, 22 Jun 2008 23:13:58 GMT

Ok. This topic has been discussed, benchmarked, and explained numerous times throughout this thread. Please read it (all however many pages) and you'll find all the answers you will ever need on the subject. Since this is thread is doomed to just repeat itself over and over, I'm going to just lock it.

Re: XNA Games 'jump', but professional games do not.

Firkin — Sun, 22 Jun 2008 23:08:24 GMT

Well, it's either something in your game process, or outside of it in Windows, that's interrupting the game periodically. Have you tried the latency check utility I linked above? That might point out if you have a Windows driver issue or something. It could also just be some service or background process waking up every four seconds, so you might try killing everything non-critical in process manager to see if that affects it, and try shutting down all services that you can etc.

If it's in the game process, then the only thing that comes to mind would be garbage collection, and you could experiment with System.GC.CollectionCount() to see how often the various generations of GC are being run and whether that's in any way associated with your problem.

Re: XNA Games 'jump', but professional games do not.

ilikegames11 — Sun, 22 Jun 2008 22:54:36 GMT

I have the same "jumping"/"jittering" issue with sprites in XNA - whether it be my own game or any of the XNA Creator samples. As mentioned previously in the thread, my framerate never seems to drop (perhpas the jitter happens so quickly that the framerate doesn't detect a change).

... So then I tried a little test. I created a game that does nothing except draw a red or blue screen depending on the IsRunningSlowly flag. So the Game1 class is untouched except for the draw method with these lines:

    
            if(isRunningSlowly)  
        
                Clear(Color.Red);  
        
            else 
        
                Clear(Color.CornFlowerBlue);

Almost every 4 seconds the screen would flash red, and then go back to blue. I have a rather sophisticated machine with hardware that is less than a year old; I wouldn't think that a game (that does nothing) should ever have the isRunningSlowly flag set to true. This, also as mentioned in the thread, does not seem to happen as much if isFixedTimeStep is false. I would be curious to know if anyone else has similar results with this test.

Re: XNA Games 'jump', but professional games do not.

Firkin — Sat, 07 Jun 2008 00:06:03 GMT

Here's a nifty little utility that checks your system for latency issues like kernel driver stalls, or anything else preventing delayed kernel events from happening. It came up in a discussion of sound card latency issues, but it might be interesting to see if some of the people here seeing jumping problems see any problems with this tool.

http://www.thesycon.de/deu/latency_check.shtml

Re: XNA Games 'jump', but professional games do not.

Lewis Walker — Sun, 01 Jun 2008 02:23:48 GMT

I have this issue. I'm using XNA 2.0 on a Intel core 2 Quad CPU, 4GB machine running 64 bit Vista, Nvidia 8600 GT GPU.

All my physics (including camera motion etc.) is calculated using the elapsedGameTime so that regardless of frame-rate, motion should appear smooth. I've written an entire game engine in OpenGL before now with this approach.

Approximately once a second the screen appears to glitch.. like objects suddenly stutter. It's just one frame at a time. It happens irrespective of strategies including isFixedTimeStep, SynchronizeWithVerticalRetrace etc.

I added some debug output during the Update method to log the value of elapsedGameTime (rounded to millisecond) and I found the following behaviour (with isFixedTimeStep = false and SynchronizeWithVerticalRetrace = true):

Update: 16
Draw: 16
Update: 16
Draw: 16
Update: 31 <== This is where the screen judders
Draw: 31
Update: 1
Draw: 1
Update: 16
Draw: 16

So you can see that on average the calls to update are happening 60 times a second, but that debug output tells me there's a dropped frame and then two calls to update and draw very close together.

So.. what gives?? And why should this cause such a judder? Because frankly the update and draw calls are at least synchronized so my physics is calculated in Update and is taking account of the elapsed time. So it should look smooth...

Re: XNA Games 'jump', but professional games do not.

Whiplash6 — Wed, 28 May 2008 13:17:59 GMT

MajSlayer:

I'm having the same problem and I'm just working with a simple template, all I'm doing to rotating a object and it not only lags(like when your out of memory and your hard drive accesses) but the graphics jump slightly, is there a fix for this or not? kinda silly, the code is like 40 lines long and yet, lag, only if I set my rotation speed to a small value will it run smooth:( I'm using a ATI RADEON 3870HD, AMD 5400+ X2 2GIGs PC-6400)

L8r

MajSlayer

EDIT: nevermind I found the answer on reply 4000, :) lol this worked

IsFixedTimeStep = false; // Put in Game1

float dt = (float)gameTime.ElapsedGameTime.TotalSeconds; // Put in update

That wont work for me, something else must be wrong on my end.

Re: XNA Games 'jump', but professional games do not.

MajSlayer — Wed, 28 May 2008 00:02:28 GMT

I'm having the same problem and I'm just working with a simple template, all I'm doing to rotating a object and it not only lags(like when your out of memory and your hard drive accesses) but the graphics jump slightly, is there a fix for this or not? kinda silly, the code is like 40 lines long and yet, lag, only if I set my rotation speed to a small value will it run smooth:( I'm using a ATI RADEON 3870HD, AMD 5400+ X2 2GIGs PC-6400)

L8r

MajSlayer

EDIT: nevermind I found the answer on reply 4000, :) lol this worked

IsFixedTimeStep = false; // Put in Game1

float dt = (float)gameTime.ElapsedGameTime.TotalSeconds; // Put in update

Re: XNA Games 'jump', but professional games do not.

scypior — Tue, 27 May 2008 21:35:06 GMT

I've only skimmed through this topic, and what I say isn't helpful, but I was in similar situation. I've bought a new PC to play The Witcher, and what I've got was occasional jumps(lags) in most of the areas. I was pissed of a little, but eventually I forgot about it:)

Re: XNA Games 'jump', but professional games do not.

Whiplash6 — Tue, 27 May 2008 16:19:30 GMT

My point is it doesnt matter how a game handles garbage collection, it will still jump. Every 2D sample from MS does this. EVERY ONE. And nearly all 2D games do this regardless how amazing the garbage collection is. Net Rumble does it but not as bad, and I still didnt notice it in the RPG game. A basic game where you draw a sprite and move it (NO objects removed or drawn other than the one) will still jump a LOT worse than a more advanced game.

The more advanced the game becomes, the less it jumps. You might not like the way the garbage collection is on the alien aggressors, but thats not the issue here for me. EVEN World in Conflict and Warcraft 3 have HORRIBLE jumping in some areas.

Re: XNA Games 'jump', but professional games do not.

PaulCunningham — Tue, 27 May 2008 08:07:32 GMT

If you're seeing the behaviour on the 360, could it have anything to do with the 50 / 60hz refresh difference between US and UK (Europe?) TVs? e.g. You're defaulting to a fixed step of 60 but you TV is syncing at 50?

Re: XNA Games 'jump', but professional games do not.

Charlie Talley — Tue, 27 May 2008 00:08:31 GMT

I did this for XNA 1.0, and it turned out that a Draw() call was being dropped once in a while. I got around this problem by dropping an Update() whenever IsRunningSlowly was set to true, therefore synching Update-Draw so that there was never an extra update called without a subsequent draw. In 2.0, however, I sometimes see dropped Draw calls, but the IsRunningSlowly flag seems to be getting set arbitrarily with respect to the dropped Draw calls. If you read my post a handful of replies back, I basically ran a lot of timing tests for 2.0 to come up with three scenarios, the first two being instances of Update or Draw calls being dropped, while the third was when the Framework was jumping in to do something, freezing the game. The suspicions that there's some process on a PC that's stealing CPU time are pretty much moot, as the stuttering happens on the 360 as well, and last time I checked there were no anti-virus software or email clients on my Xbox.

One possible avenue I've been exploring today is that the Framework jumps in and stops the game momentarily because of a graphics pipeline stall. If Draw is feeding the GPU faster than the GPU can keep up, this would (I think) cause the Framework to kick in and do some housework. This would explaing why the stuttering is more noticeable in simple 2D games than in complex 3D games... the simpler the game, the faster the CPU can feed the GPU, causing it to stall even for simple drawing operations like sprites....Any GPU experts in here that can verify if I'm totally talking nonsense or if this is actually possible?

Re: XNA Games 'jump', but professional games do not.

ShawMishrak — Mon, 26 May 2008 23:09:36 GMT

Firkin:
I don't suppose something like fraps would be able to capture the behaviors that people are seeing? Or even point a DV cam at the screen, etc.

Fraps locks the frame rate to the target video frame rate. This could have any number of effects on the present issue.

What about an event log? Place a call to a trace routine in your Update and Draw methods which records:

The exact sequence of Update and Draw calls.
The GameTime given to each Update and Draw call, including all time fields and the value of IsRunningSlowly.
An independent timestamp for each call.

That may help to narrow down the problem. The time deltas passed along in the GameTime structure may be buggy in certain conditions, or you could just be a dropping a frame somewhere.

Re: XNA Games 'jump', but professional games do not.

Firkin — Mon, 26 May 2008 22:54:47 GMT

jwatte:
I think there is some confusion here, because the first case is probably more common, and the symptoms look more or less the same as in the second case.

Yes, exactly.

It would be helpful to actually see some of these problems. I have the suspicion that the first case issues are effectively single draw() or update() dropouts where the second case issues might be multi-frame dropouts where some driver monopolizes the system for some substantial number of milliseconds.

I don't suppose something like fraps would be able to capture the behaviors that people are seeing? Or even point a DV cam at the screen, etc.

Re: XNA Games 'jump', but professional games do not.

jwatte — Mon, 26 May 2008 19:16:15 GMT

There appears to be two separate cases. One is when you set "IsFixedTimestep" to true, in which case you will get jumps unless your time step is an EXACT multiple/divisor of the monitor refresh rate. And, because the monitor crystal is separate from the CPU crystal, there's no chance of them being exactly the same. The monitor can't be made to "follow the XNA timestep size" -- it's the other way around. The XNA framework needs to follow the monitor. If you use IsFixedTimestep, then you must make sure your timestep is significantly different from the vsync refresh rate, or you will get jumps that are basically caused by time sampling aliasing. Thus, the better way to get fixed timesteps is to turn vsync on, and turn off the "IsFixedTimestep" variable.

There appears to be a second case, where some people see jumping/jerking even when turning on vsync, and even when not turning on IsFixedTimestep. That jerking could be caused by hardware or software or some interaction thereof -- although the exact causes seem to be so far unknown.

I think there is some confusion here, because the first case is probably more common, and the symptoms look more or less the same as in the second case.

Re: XNA Games 'jump', but professional games do not.

Nick Gravelyn — Mon, 26 May 2008 16:35:57 GMT

Whiplash6:
I used alien aggressors as an example because its basic, and anything basic will jump.

That's fine, but my point is there are lots of other things in Alien Aggressors (such as the probably numerous garbage collections that occur) that will cause jumps as well. It is not a good measurement of the problem discussed in this thread.

Now it could be my computer, but with lots of people experiencing the same problem with different hardware configurations makes it seem unlikely.

All I know is that in my normal games that I work on, I never see this problem. That could be that it never happens or that I mask it because I, like Lord Ikon, always use elapsed time for any movement or it could be for some other reason. But between my old tower and my newer laptop, I never see this issue besides in these little test apps that flash when the framerate drops. Maybe my eyes just aren't as picky or something, but I personally don't really see this as anywhere near as big of an issue as some do.