What was that about Cg *Not* favoring Nvidia Hardware?

I don't get it. I expose a very real current problem with the Radeon 9700, uncover some evidence that it will likely also apply to DOOM3, and people get upset?

How did you 'expose' something thats written in black and white for everyone to see?
 
SirPauly said:
I don't really play that many games

I figured as much but it's nice to know for sure.

Yea, but at least he plays good games. 8)

And I did indeed notice Z errors again in Morrowind today. I've just gotten back into the game with a save file I'd backed up of a 7th level elf ranger who was currently in Ald Rhun (sp?). The Redorans, prevalent in this area, don't show any model problems, but others wearing clothing do. Strange. At least the jittering of waterlines has been greatly reduced.

On the good side, my elf ranger (oh-so-creatively named Legolas) is now 12th level. :p
 
Althornin said:
How about you LEAVE IT OUT?
Believe it or not, I usually try.

Chalnoth, you seem to delight in coming up with "problems" for ATI.
So? The more problems exposed in current chip architectures the better. This brings greater liklihood that they'll be fixed.

What makes you think this problem wouldnt exist in the GFFX
Already stated.

And, who says its a "bug" - isnt it just a possible problem with Hier-Z implementations in general?
That's like saying nVidia's S3TC image quality issues are a problem with S3TC implementations in general. This is most probably just a failure of ATI's Hyper-Z to include optimized pathways for specific rendering methods. For example, why single out "stencil depth fail?" Shouldn't the hardware be able to do the same thing with a "stencil depth pass?" The fact that it doesn't on the Radeon 9700 speaks of a limitation wherein ATI failed to properly-aniticipate the direction games would be going.
 
You know what, screw politeness...

Chalnoth, you along with doomtrooper, hellbinder, and your likes have absolutely ruined these boards. Period.

Been waiting a looooong time, hoping for things to come back around but it seems things continue to spiral downward.


Hope admins dont take offense for knockn their baby.
 
Diespinnerz,

Get off your high horse dude, I've been a member of this board since the original crew of the two Dave B's, Kristof. Anyone that remembers those days also knows not much is different from those days with the heated 3DFX and Nvidia battles.

Funny I don't remember you.
 
Been around just as long, what difference does that make...

Doesnt surprise me to see you dicount my opinion because of a usernames post count. Absolutely infantile.

The only respect your 1500 count earns is for Dave and the rest of the crew for not outright banning you, and instead putting up with your non stop trolling and spamming .
 
I see..the moon turns blue too..I suggest you stick to what you were doing before..

Lurking.gif
 
For the record, I also do notice some Z errors on the 9700Pro with the latest drivers. The latest game to show this for me is 007 Nightfire.

Also, in case no one has reported the differences between a GF4Ti4600 (core/mem clocks 300/650) and a 9700Pro (324/620) in the Tenebrae mod :

Stencil OFF :
9700 Pro = 78.1fps
GF4Ti4600 = 67.5fps

Stencil ON :
9700Pro = 33.7fps
GF4Ti4600 = 27.1fps

Percentages between Stencil ON and OFF :
9700Pro = -56.85%
GF4Ti4600 = -59.85%

Stencil ON includes disabling scissoring (much more stencil fillrate requirement). Full mirror effects also enabled (not sure if the Tenebrae author uses stencil for his mirror implementation). Don't forget, of course, per-pixel lighting is used all the time.
 
Chalnoth said:
Yes, 8x AF and 2x AA. The only game that has real problems with this is UT2k3, so I run it at 800x600x32 instead of the usual 1024x768x32. Anyway, when playing UT2k3, the Radeon 9700 really shows its stuff, but I won't turn the settings down any further when I'm using the GF4. Can't stand to play anymore without FSAA/Anisotropic.

Question: are you using Rivatuner or any other 3rd party tweaker to adjust AF performance? Not trying to trap you into admitting anything wrong. . .simply curious.
 
Reverend said:
Stencil OFF :
9700 Pro = 78.1fps
GF4Ti4600 = 67.5fps

Stencil ON :
9700Pro = 33.7fps
GF4Ti4600 = 27.1fps

Percentages between Stencil ON and OFF :
9700Pro = -56.85%
GF4Ti4600 = -59.85%

Stencil ON includes disabling scissoring (much more stencil fillrate requirement). Full mirror effects also enabled (not sure if the Tenebrae author uses stencil for his mirror implementation). Don't forget, of course, per-pixel lighting is used all the time.

Very interesting. Did you do this with FSAA enabled? What level? Of course, the really interesting thing here is how incredibly close to one another the tests are between the Radeon 9700 and GeForce4. Usually this speaks of a CPU limitation. But, at the same time, the Radeon 9700's drivers usually aren't as good yet at CPU-limited situations, so even that seems a little bit out of place.

Additionally, it would seem odd if per-pixel lighting were the cause, though I suppose I should reinstall and check the one game that I have that doesn't support stencil but does use lots of DOT3 (Giants).
 
John Reynolds said:
Question: are you using Rivatuner or any other 3rd party tweaker to adjust AF performance? Not trying to trap you into admitting anything wrong. . .simply curious.

No, I haven't used RivaTuner for quite some time, for any reason (not since the anisotropic slider in the drivers was implemented properly). I have played around with the various performance tweaks available for anisotropic in RivaTuner, and didn't really notice enough of a difference in real gameplay to care.

And I don't mind people asking questions like this. It's very valid.
 
FSAA disabled for all tests. Used Tenebrae 1.02 and its built-in demo1 (it's a demo of the start level). And yes, Tenebrae will show a greater performance difference between faster CPUs than between a 9700Pro and a GF4Ti4600.

I wouldn't take Tenebrae to be indicative of DOOM3 as far as overall performance goes however.
 
I see ZBuffer errors on both the Ti4600 and the 9700 Pro with the most current drivers for both. The interesting thing is- the list of games you see this occurring are almost totally mutually exclusive between the two cards, hence why they can be categorized as ZBuffer errors.

Obviously, the unique implementations of ZBuffer optimizations is going to piss off some games and cause these issues, with the 9700 Pro's optimizations being the most aggressive. 16-bit Zbuffer on the 9700 Pro is almost laughable as you can count the number of games without problems on one hand, with the remaining majority having massive problems when using 16 bit Z.

As far as performance problems, I don't subscribe to Chal's theory concerning his reasoning. Overall, ATI's OpenGL drivers are about 1/4 as mature as their Direct3D drivers, and even games without stencil/checks are performing substantially under par on the 9700 Pro. It still retains a performance lead over the Ti4600, but not nearly the same lead it holds over Direct3D. The drivers also still have a plethora of issues with texture paging/loads that ATI is busily working on- and can cause "pauses" in game for users of certain motherboards and chipsets. Just two (2) days ago they released a new OGL driver that reduces this quite dramatically, but does nothing for overall top performance.

And yeah- the thread has been derailed off Cg and PS 1.4, but that's what happens around here.
 
Sharkfood said:
I see ZBuffer errors on both the Ti4600 and the 9700 Pro with the most current drivers for both. The interesting thing is- the list of games you see this occurring are almost totally mutually exclusive between the two cards, hence why they can be categorized as ZBuffer errors.

I'd like to know the games (and specific situations) that you've seen where the GeForce4 has z-buffer errors and the 9700 does not. I know I've seen a few z-buffer errors on the GeForce4. I know that the z-buffer errors in UT and UT2k3 appeared in the exact same form on the Radeon 9700. The only one that I didn't check was Serious Sam (I didn't feel like getting to one of the levels where I really noticed the problems).
 
Original post snipped:

Doom, intelligent people can spot bias when it's in a person's posts. . .you don't have to flame that person for all the world to see it. Moreover, I don't see you lambasting the ATi f@anboys around here. Just take a deep breath and relax a little. ;)

JRR
 
Chalnoth said:
Very interesting. Did you do this with FSAA enabled? What level? Of course, the really interesting thing here is how incredibly close to one another the tests are between the Radeon 9700 and GeForce4.

I doubt its using two sided stenciling either.

Rev - was this done with Trilinear or Bilinear?
 
Sharkfood said:
Obviously, the unique implementations of ZBuffer optimizations is going to piss off some games and cause these issues, with the 9700 Pro's optimizations being the most aggressive. 16-bit Zbuffer on the 9700 Pro is almost laughable as you can count the number of games without problems on one hand, with the remaining majority having massive problems when using 16 bit Z.

If the Zbuffer optimisations causes problems then it's a faulty implementation. A card with HyperZ or equivalent technology should generate the exact same picture as does a card without these optimisations, otherwise it's non-conformant and faulty. I'm fairly certain that neither ATi's or nVidia's optimisations causes any difference at all as compared to previous unoptimized Zbuffers.
Now I don't play a whole lot of games, but I don't see why 16bit Zbuffers would cause any problems, unless you have a huge world that is, in which case you shouldn't be using a 16bit Zbuffer. Many of my demos uses 16bit Zbuffer and I've never seen any rendering problems unless I use a large world.
 
DaveBaumann said:
I doubt its using two sided stenciling either.

Rev - was this done with Trilinear or Bilinear?

That's true, but even without two-sided stencil, if the test were fillrate-limited, you would think that the R9700 would be doing far better.
 
Back
Top