ZGameEditor forum

Here is a little demo that shows some of the things I have been working on this weekend,. I noticed the new beta has vbo-support so I will rebuild this in the morning and see if the speed improves,. I get a solid 60fps here,. the fps display is only on one of the title screns,. . there are a lot of polys,. but just a few models,. needs shaders! I am really happy with the background I am sure it will be re-purposed for a game,. in fact I will likely just edit this into a simple shmup., also getting happy with what I am able to do with the sounds,. this is some minimal glitch-house type buisness,. again I will listen to it in the morning and see if then I don't think it sucks at that time as often happens.

I will go post about the modeling tool I used here,. it is a sic subdivison setup,. . TopMod. l8r.

REMOVED the file for space,. get latest ver. below..,

I get around 30fps here so I think this is a typical scenario where vbo will improve framerate: multiple instances of a high-polycount mesh. You are on a productive spree jph . I like this too. In this case a subdivision-component would have been useful so that you could start with an imported low-poly model and repeatedly subdivided. I'm sure this will come to zge at some point.

oddly enouph I get worse preformance with the VBO ver. ?!?

I added a control (Up/Down arrows) to set the number of knot-balls dancing,. at 5 I now get around 30fps,. and with 3 I get a solid 60fps,. . any ideas why that may be ? I got 60fps on the last ver.

Anyway I am going to get rid of the heavy-poly balls and see if I can build a little shmup on that sweet backdrop,. but what ver. should I use??

Also I found some sort'a bug in the material component,. if you define a bitmap on texture 2,. some of the settings on OTHER materials get broken,. it looks like it has to do with useing bitmaps on texture2 and perhaps 3 as well,. I will try to build a sample,. .

Strange. It's not that graphic driver setting again? Vertical sync?

I still get 30 fps in this version. 40 with no balls.

How many triangles are the mesh?

No, don't tink so., if I run the non-vbo ver. I get 60 FPS,.
when I run the vbo ver. I get 30 FPS w/ all five balls,. and only does it bounces back to 60 FPS when I reduce the balls to only 3.
The knot-ball thingies are 32k triangles,. the back ground mesh is like 57k and it is drawn 3x,. how many polys can these gpu reasonably draw? I have a middle of the road setup so I just go by my performance,. and then try to scale things back,. guess I should look into what games I know are doing and try to make reasonable design decisions eh,.

Try to see what is different to my vbo-test project. Because you still get good performance from that, right? Also if you want you can send me the zgeproj so I can check if there is anything strange going on in the debugger.

I have to say I do not understand this. When I make some tests it seems like VBOs are faster with meshes having around 50k or more triangles. Otherwise they are equal speed or slower. So not sure what to do about vbos in zge now. Should I only enable them for large meshes? Or is it only our graphics cards that suck, jph?

computers are evil

well sure,. I suppose it would make sence to add a switch per mesh, if that is possible, or auto over a reasonable size where there are gaines,. if the two can be mixed np.
I'm going to just keep building stuff,. and work it out case by case for now,.. some lower poly objects for some action game buisness are now in my head.

Hey guys,

If you want I can test-drive a VBO and non-VBO version on several systems / gpu's again. Just let me know ~

K

here is the non VBO ver. updated with the up/down arrow to set the number of objects,.

The VBO version runs around 3-4 times faster on all systems I tried.

K

well now i wish i was building games for consoles,. .

You got the latest graphics drivers jph? I need to make some kind of compromise here. There is obviously a benefit of using VBOs, but not on older graphics cards. Perhaps I'll activate VBOs for meshes larger than 1k triangles, and keep the old style for smaller meshes. I don't want to make it a user-setting because I consider it a technical detail that the engine should decide for itself. There are options to investigate with VBOs such as packing the meshdata in different ways, so maybe I can optimize it further later on.

yeah just do what you think is best,. I will live with it. This whole machine is only a couple years old,. I guess that is out of date the way these things go though,.
but this vidcard playes most commercial games well,. HL2, TF2, etc,.

http://www.asus.com/products.aspx?l1=2& ... odelmenu=1

I just find it strange that on different hardware we find different results,. in my mind things should not be so differently implemented,. but then again in my mind we should be living in a world of peace without oppression and greed,. so there you have it. (you may say,.. and I'm not the only one.)

with a console the hardware is all the same,. so no guess work,. thats the frustration.

Hi guys,

Keep in mind that the oldest machines I tested Ville's VBO experiment on showed the most significant differences in favor of using VBO. The Ahtlon system for example is only a 1Ghz from 2001 or so.

Developing for consoles has it's own share of annoyances .. for starters all the GUI guidelines you must follow to get through the approval process

K