Pretty good 5800 review

[SurfMonkey]

Originally posted by Hellbinder
As I have pointed out over and over Anisotropic filtering was never intended nor appropriate for all angles. Primarily 90' and some others. This is a mathematical principle that is true even outside the graphics world.

And that's why you have an AAF (Adaptive Anisotropic Filtering) algorithm. Something that works along the lines of:

1. Classify each sample block to either edged, flat, or textured.
2. Classify each pixel - block class, spatial position, and distance from closest edge.
3. And then filter each pixel using some predefined discrete convolution according to classification.

That way you can account for nearly all the edges with your algorithm if you choose to. If you aren't going to account for anything other than parallel edges then there is no point in using AAF or even AF. If ATi have chosen to cut some corners in order to decrease the bandwidth requirements or increase algorithm speed then that's their design choice. But it definately doesn't make it more correct than an AAF algorithm that does account for more permutations.

[Lucien1964]

Originally posted by jAkUp
for the last ****ing time..
the nv30 DOES NOT beat the 9700 in any tests... high resolution, low aa or no aa... the 9700 takes the lead by a huge margin as you start to increase the aa though. quality modes compared to quality modes... the geforcefx trys to keep up in some tests... (im talking about application slider, not aa or af)

[hovz]

marcocom i cant test radeon aa modes, i have a ti 4200

[surfhurleydude]

GeForce FX and Ati Radeon 2xAA modes look almost exactly the same from that screen. The lighting differences present more jaggies in certain areas of each shot that aren't in the other... Good news for those of us that game at 1280x1024 with 2xAA eh? :D

[jAkUp]

Originally posted by CoWBoY
Does anyone know if the shadows are still missing with the 43.51's in Splinter Cell or did it fix it along with the aurora effects of the lamps like you see in the 9500Pro's pics?

I understand that the bullet tracers are still not present under the 43.51's in Wings of Fury Demo on 3DMark03'.

all effects are there now... including bullet tracers.. i used to have a shadow bug with splinter cell, but the new drivers fixed it.

[SurfMonkey]

Originally posted by CoWBoY
Hear is the 42.72 Det's with a 5600Ultra compared to a 9500Pro with Cat 3.1's:

I can see things are missing... ofcourse, it may be just my eyes in particular.

The most annoying thing about that, is that it's going to make the game harder. The visibility slider moves as well, which means that not only is it, *maybe*, a rendering defect but the actual game engine is picking up the fact that the lights are brighter and so there are less shadows.

[jAkUp]

im pretty certain that those issues you see are fixed... the drivers have improved a whole lot since that review

[bkswaney]

Originally posted by jAkUp
im pretty certain that those issues you see are fixed... the drivers have improved a whole lot since that review

Yep. :)
The 51's are pretty good. I cannot wait for the Det FX's. :angel:

[eagle17]

here is another pretty good review for the 5800 ultra
link (http://www.hartware.net/review.html?id=303)

you can look at the graphs or use a translator

[garikfox]

Man i want a 5800 DUSTBUSTER, lol

[Behemoth]

why 5800 ultra AA/AF performance differs so much under different games?

SS SE from no AA/AF to 4X/8X
1024*768 85.8 -> 73.5 (14.34% hit)
1280*1024 76.8 -> 57.0 (25.78% hit)
1600*1200 64.3 -> 38.3 (40.44% hit)

AquaNox from no AA/AF to 4X/8X
1024*768 134.3 -> 46.3 (65.52% hit)
1280*1024 89.0 -> 30.4 (65.84% hit)
1600*1200 63.6 -> 22.0 (65.41% hit)

mystery...

[hovz]

it seems to slow down much more heavily on texture intensive games cuz it has weak bandwidth

[Behemoth]

Originally posted by hovz
it seems to slow down much more heavily on texture intensive games cuz it has weak bandwidth
ah... damn...

[Behemoth]

i dont know how that reviewer took that FX 4x/8x SS screenshot for geforce FX, i took one from 800x600 mode myself, it certainly looks different.
due to size limitation, i cant upload the png version, so keep in mind that this should look a bit better in png format too.

[Behemoth]

Originally posted by hovz
it seems to slow down much more heavily on texture intensive games cuz it has weak bandwidth
hmm, even if its bandwidth problem, the performance hit should increase as the resolution increase, i think its more like FX AA/AF doesnt work properly with that aquamark, or aquamark doesnt work properly with nvidia AA/AF.

[Behemoth]

this is my benchmark results

P4 1.7/FX 5800 non-ultra
driver version 43.51 WHQL
quality mode
Aquamark v.2.1, settings:
FSAA : NONE (will enable externally)
TEXTURE : 40M (max)
PIXELSHADER : YES

1024X768X32
no AA/AF 38.3
4XAA/8XAF 33.1 (13.58% hit)

1280X960X32
no AA/AF 37.2
4XAA/8XAF 28.5 (23.39% hit)

1600X1200X32
no AA/AF 35.6
4XAA/8XAF 20.6 (42.13% hit)

i dont know what version he was using, but the AA/AF performance hits do not look right.
my results are more inline with his serious sam results in terms of AA/AF hits.

[StealthHawk]

In what ways do your screenshots look different?

Of course different games will have different performance hits for AF. Depending on what kind of scenes are in the game, how intensive the engine is, etc.

[Behemoth]

Originally posted by StealthHawk
In what ways do your screenshots look different?

Of course different games will have different performance hits for AF. Depending on what kind of scenes are in the game, how intensive the engine is, etc.
actually i see lots of differences but check the area in red circles. and since i re-save this jpg again, you probably better check the original screenshot, coz jpg uses lossy compression.

[hovz]

Originally posted by Behemoth
this is my benchmark results

P4 1.7/FX 5800 non-ultra
driver version 43.51 WHQL
quality mode
Aquamark v.2.1, settings:
FSAA : NONE (will enable externally)
TEXTURE : 40M (max)
PIXELSHADER : YES

1024X768X32
no AA/AF 38.3
4XAA/8XAF 33.1 (13.58% hit)

1280X960X32
no AA/AF 37.2
4XAA/8XAF 28.5 (23.39% hit)

1600X1200X32
no AA/AF 35.6
4XAA/8XAF 20.6 (42.13% hit)

i dont know what version he was using, but the AA/AF performance hits do not look right.
my results are more inline with his serious sam results in terms of AA/AF hits.

which driver settings are you using

[Behemoth]

Originally posted by hovz
which driver settings are you using
i am using quality mode, it was called application in det. 43.45, if this is what you wanted to know.