Hi all
Say we have a card spec 8600/8800/256/512ram etc
What would the target asset buils size for polys and textures?
Lightmapped and one/two pixel light of normals etc
4k asset groups sharing one texture sheet etc
Polys - ?
Textures - ?
Cheers
Andy
To run at 40fps and 800x600
I don’t mean this to be sarcastic. I am actually interested:
Do people with those cards still use 4:3 monitors?
I don’t have anything that good yet, and I have not yet run into a lot of money, but I haven’t used a 4:3 monitor for over 3 years, save a short stint with an iBook. Then again, perhaps you are using a more square aspect ratio to maximize the effective rendering area, given a certain amount of pixels, with intention to use pillarboxing? Or, maybe the 4:3 monitors really are still the norm for PC’s, and that is your target platform…
Anyway, sorry to add a question instead of an answer.
Not at all
800x600 is the res we work too as a standard for web based realtime, we have to have a stadard as there’s sooo many different macs/pc’s/laptops people will use, so if it looks crap and customer moans we cover our ass by setting on that size and everything works ok etc etc
I’m running an FX 5600 @ 3840 x 2400 that has 1.5gig ram, over two Lacie 321 screens
The Unity player we are making will be aimed at the new 8800 mac Pro card, untill we get some in… we can’t test.
So was wondering what Unity can handle
Thanks
Andy
No one? 
Hows 25k polys and 20meg textures sound?
If the target is a quite good (GeForce 8600/8800) 256MB card, then having something like 200-300MB of textures seems good (that number includes both textures and render textures, and back/front buffers).
But I think the texture size hardly matters, as long as you make it fit into VRAM. What matters is the fillrate (“how many pixels have to be drawn, also counting the ones that are drawn multiple times”), shader complexity (diffuse is obviously faster than parallax bumpmapped specular) and video memory bandwidth (Not having mipmaps will hurt a lot. Using trilinear and/or anisotropic filtering is also slower than bilinear filtering).
Polygon wise, GeForce 8600/8800 are pretty good at vertex transformations. I don’t know the exact numbers, but I guess a million polygons rendered per frame is good. Maybe it could handle two or three million polygons/frame as well, not sure. Note that this is total number of polygons rendered, and not just number of polygons in your objects (some objects might be rendered multiple times, while others might not be rendered at all).
Cheers, most informative.