Shawn Hargreaves Blog
As of version 4.0, XNA Game Studio no longer supports point sprites.
One big reason, plus two small ones.
The big reason is that DirectX 10 and 11 do not support point sprites, so if we kept them in our API, we would be unable to someday move that API to future DirectX versions. This would violate our "break it good" goal of taking all the pain now to avoid future compatibility breaks.
It is theoretically possible to emulate point sprites using geometry shaders, but after some investigation we concluded this would be:
A smaller reason is that point sprites behave differently on Windows and Xbox, and are not even consistent from one Windows GPU to another (the max size varies, as does the result of TEXCOORD interpolators).
The final reason is that point sprites are slower than triangles on some common graphics chipsets. Hardware manufacturers are sure to keep point sprites around for DirectX 9 compatibility, but they are unlikely to spend much time optimizing what is now a legacy feature, so this performance delta will only grow over time.
If you are worried about the idea of life without point sprites, the first thing you should know is that this worried me at first, too. But I'm not worried any more. By the end of this article, hopefully you won't be either.
When we looked at why people use point sprites, we found two main purposes:
If you've been reading my blog for long, you probably already heard me mention how the first step in performance work is to measure. Using GS 3.1, I changed the particle sample to draw triangles instead of point sprites. In fact I made three versions:
All three versions used insignificantly tiny amounts of CPU time. The difference was in GPU performance:
A 30% gain from not using point sprites is pretty sweet, but an 80% penalty is painful indeed. I see more people using rotating particles than otherwise, but still... ouch!
One of the best things about my job is the chance to work with amazingly smart people. When I mailed around these performance figures, I was lucky enough that one of the graphics gurus who created the Xbox GPU driver happened to see them, and was sufficiently intrigued to spend some time looking at PIX captures of my test app. Jason discovered that my indexed triangle implementation was bottlenecked by rasterizer performance, specifically by an obscure hardware penalty which occurs if you use both center and centroid mode interpolators with a shader that is fast enough not to be bottlenecked by anything else (an extremely rare situation).
Once identified, it was trivial to remove this bottleneck, which left my indexed triangles bottlenecked by texture fetches, just like the point sprite original. Revised timing data:
Armed with these figures, removing point sprites didn't seem so painful any more.
I can't promise exactly when, but we are working on an updated version of the Particle 3D sample which shows how to use indexed triangle lists, and includes Jason's magic optimization. We will be sure to get this out by the time 4.0 ships, if not before.
Are the final figures for the WP7, XBox360 or the PC platform? If the latter, for what specs?
These figures are from Xbox 360.
PC performance obviously varies wildly from one chipset to another, but if you average across many different GPUs, the median result is similar to what we see on Xbox.
Music to my ears Shawn. You guys ROCK!
I can't wait to get my hands on the new Particle 3D sample. I'm going to need it for game #2.
Hmm, What about cache effects?
4x the number of vertices could be quite a bit of memory... How does the performance compare when other stuff is running in the background thrashing the CPU cache(even on the same core)?
(I guess smaller vertex formats would help either one... Plus balancing the cache size against the batch overhead)
"Once identified, it was trivial to remove this bottleneck."
Was this the use of 'nointerpolation' on the Color and Rotation vertex attributes? Since all vertices have the same Color and Rotation there's no need to interpolate these over the polygon(s)?
Forgive my noobness but I use point sprites to render a space dust field. When do you say Particles that rotate versus Particles that do not rotate. What exactly does that mean?
Particles that rotate are the ones that are constantly rotating to face the camera right?
and particles that do not rotate are the ones that have their orientation fixed?
Or is it the other way around?
Is this shader part of XNA 4.0 for WP7? If not, are you guys going to include it?
Is PIX for Xbox 360 something we can download now with GS 4?
I'm disappointed to hear that there is no geometry shader solution. I am using a much more complex vertex shader than the Particle sample (but a very simple pixel shader), and running it three or four times is a real problem. I will have to try and come up with a two-pass solution, rendering particle state to a texture, which sounds horribly complex.
Eduardo; Shawn means particle sprites that are rotated around their centrepoint, in 2D (screen) space. Both types always face the camera.
> 4x the number of vertices could be quite a bit of memory... How does the performance compare when other stuff is running in the background thrashing the CPU cache(even on the same core)?
The CPU usage (and CPU memory bandwidth) for this style of particle system is vanishingly small either way. Even when drawing enough particles to entirely saturate the GPU, the CPU load is insignificant.
> Was this the use of 'nointerpolation' on the Color and Rotation vertex attributes?
The optimization is to use either all centroid or all center mode interpolators, rather than one center and one centroid.
> Is this shader part of XNA 4.0 for WP7? If not, are you guys going to include it?
No, and no. This kind of GPU intensive particle system isn't really a good fit for mobile GPU hardware, so it wouldn't perform well no matter how it was implemented.
The best way to do particles on the phone is to animate on the CPU rather than GPU. In 4.0, the resulting data can be efficiently drawn using SpriteBatch even for 3D scenes (I'll write more about that later, but it's a way down the priority list :-)
> Is PIX for Xbox 360 something we can download now with GS 4?
PIX does not work on retail Xbox 360 consoles. It requires a devkit.
If you are a registered Xbox 360 developer who has devkit access, you can use PIX with any version of Game Studio.
>> 4x the number of vertices could be quite a bit of memory... How does the performance compare when other stuff is running in the background thrashing the CPU cache(even on the same core)?
>The CPU usage (and CPU memory bandwidth) for this style of particle system is vanishingly small either way. Even when drawing enough particles to entirely saturate the GPU, the CPU load is insignificant.
I wasnt concerned about the CPU usage, rather the memory bandwidth from L2 => Main Memory.
Bandwidth is a scarce resource on 360, I would think if it [lack of point sprites] is going to be a problem, it would show up when there is lots of other stuff going on, not just a synthetic benchmark. (ie it would be interesting to know your results with the other cores walking through memory, so they miss the L2 constantly)
nics sample shawn, i use the billboard two triangle or sometimes the 4 triangle like and x to give a 3d fell and look, allways for my paraticle system
any chance of seeing the hidef profile in a beta before summer time