frogspasm_deux
That looks very fast!
What graphics card are you running it on?
Gainward GTS 450, 1GB, GLH(goes like hell:-)). I sold my Ati and buy Nvidia and I am very happy.
This Nvidia is nothing special. I would like to buy one more and try SLI, but I dont have money. Iam just student
Today I have done few other optimization. So mesher is around 30% faster and use very low ram and vram.
I tested 5mil, which start under 5s per frame(with 10substep), after few collision 10-20s - time include all(sim+meshing+sending data to houdini).
5mil. is not maximum on my hardware, at this moment without clouds, I can simulate only 8mil. Next thing is that Nvidia has official support only for OpenCL 1.0 so for many arrays I use vector4. OpenCL 1.1 support vector3, which saves memory, but at this moment there is only beta support. So with 1.1 I will be simming more particles, but I will prepare Clouds, so maximum particles will be limited with ram not vram
![](/static/djangobb_forum/img/smilies/smile.png)
edit:
ouuu, next thing is that I use old nvidia driver, I am lazy to update and I heard that new one is faster, maybe I will be surprised :shock: