View Single Post
Old 10-01-11, 00:14   #74
Grease Monkey
gidierre's Avatar
Join Date: Nov 2007
Posts: 1,527

Made some tests, thx to a relative lending me for a few hours a HP laptop from 2010 running on Win7 64bit and equipped with 2 gfx cards and ATI PowerXpress software which lets you switch from an Ati Mobility Radeon HD 5470 (Driver version: 8.712.0.0) to an Intel gfx media accelerator
so I installed v1.6 and using Radeon 5470 noticed that no depth artifacts appeared as can be seen in this screenshot (sorry for the poor quality due to time shortage):


except that...
moving around as distance increased they did show up after all, but only when 5 "squares" or so removed from the "critical" surface, in this case the wall and the poles in the room beyond as I tried to show in following screen:


just for the record, I checked again on a different pc with nVidia GeForce mobile 8600M GS and definitely no depth issues, not even backing off as far as I could from that wall i.e. up to the opposite wall: no poles (didn't bother to take screens there)

I went ahead instead on the Ati machine, but swapping gfx to the Intel adaptor through ATI PowerXpress and to my surprise, no artifacts, even if I made sure to back away farther than the spot where they started to show with the Radeon, as hopefully the following screen will demonstrate comparing positions vs. the pool's edge


Does this mean Intel is better than Ati here? Obviously not really.
Not to mention that Intel will detectably slow things down, see the fps figures in screens?
21 frames/sec vs. 29 frames for Ati so game lags using Intel, although there might be room for optimization here customizing settings especially to improve performance with Intel which I had no time to try.

Graphics-wise anyway, as a matter of fact Intel clearly takes advantage of the force 24-bit Z-buffer depth hack that the dll implements and the Ati seems averse to
if Intel's buffer can be raised from 16bit to 24bit, it's unclear how Ati handles the buffer on its own, whether 16bit (which I doubt it is, and it wouldn't explain why the precision fix won't work then) or >=24bit (so the hack wouldn't stick, but then there ought to be no issues in first place)
for sure Catalyst Control Center has no way of changing it as opposed to earlier Catalyst drivers' versions I saw

that the problems should come with distance is interesting, but surely no surprise: read e.g. here
12.040 Depth buffering seems to work, but polygons seem to bleed through polygons that are in front of them (...) As the ratio (zFar/zNear) increases, less precision is available near the back of the depth buffer and more precision is available close to the front of the depth buffer. So primitives are more likely to interact in Z if they are further from the viewer.
and here (the 65,535 number applying to 16bit buffer)
12.050 Why is my depth buffer precision so poor? (...) Everything at eye coordinate depths from -395.9 to -1000 has to map into either 65534 or 65535 in the z buffer. Almost two thirds of the distance between the zNear and zFar clipping planes will have one of two z-buffer values!
the sketchier the calculations, the fuzzier things get at a distance
to the point of some short but harsh "fits" of z-fighting

nothing special it's true, since I had already quoted some useful links months ago
like this and this.
We often forgive those who bore us--we cannot forgive those who find us boring.

Last edited by gidierre; 10-01-11 at 00:24.
gidierre is offline   Reply With Quote