Silly bullcrap with my R300

[Xmas]

The problem is that more TMU's would only increase performance for fixed-function applications.

If you call quality texture filtering (aniso, trilinear) 'fixed-function applications'...

 

[Ante P]

The problem is that more TMU's would only increase performance for fixed-function applications. I would find it more than a little disappointing to see hardware manufacturers turn around and start focusing on performance of "low tech" at the expense of "high tech" performance.

A second TMU would surely increase Aniso+Trilinear performance by a large margin in UT2003 for an example, or so I think I can't of course back it up with any real world benches.

 

[Humus]

The problem is that more TMU's would only increase performance for fixed-function applications.

Care to elaborate on this one? I see no connection.

 

[PeterT]

I would find it more than a little disappointing to see hardware manufacturers turn around and start focusing on performance of "low tech" at the expense of "high tech" performance.

Like nVidia did when they decided to keep the "old" integer register combiners in nv30?

 

[noko]

Sounds like the big increase in frame rates at a lower resolution with AA on was due to the limited memory on the card, abeit 128mb. At higher resolutions using 6x or even 4x does take up alot of memory and the newer games do have alot of texture data that needs to be on the card vice main memory in order for it to work fast. If that makes any sense.

 

[WaltC]

That's probably because UT2K3 is an extremely cpu-limited game...

huh, what would that have to do with it?
it's not like 1600x1200 puts a whole lot of loaf on the CPU compared to 1024x768

What I had in mind was the fact that you can run at 1024x768 with 6x FSAA as fast as you can at 1280x1024x2x FSAA (or faster) because in both cases the cpu is the limiting factor. Were it not, I would expect that 1280x1024x2FSAA would run faster than 1024x7x6x FSAA. IE, the cpu is slowing you down at the higher resolution even with a much lower hit on FSAA--or another way of looking at is you are using a lot more of the card's potential at 1024x7x6 than at 1280x1024x2, IMO. Wasn't talking about stress on the cpu--but stress on the 9700P which, since it is cpu-limited at 1280x1024x2 in this game, is relatively relaxed compared to 1024x768x6 where the game is just as cpu limited, but the card can flex much more of its muscle because of the 6xFSAA mode you are running.

 

[Ante P]

That's probably because UT2K3 is an extremely cpu-limited game...

huh, what would that have to do with it?
it's not like 1600x1200 puts a whole lot of loaf on the CPU compared to 1024x768

What I had in mind was the fact that you can run at 1024x768 with 6x FSAA as fast as you can at 1280x1024x2x FSAA (or faster) because in both cases the cpu is the limiting factor. Were it not, I would expect that 1280x1024x2FSAA would run faster than 1024x7x6x FSAA. IE, the cpu is slowing you down at the higher resolution even with a much lower hit on FSAA--or another way of looking at is you are using a lot more of the card's potential at 1024x7x6 than at 1280x1024x2, IMO. Wasn't talking about stress on the cpu--but stress on the 9700P which, since it is cpu-limited at 1280x1024x2 in this game, is relatively relaxed compared to 1024x768x6 where the game is just as cpu limited, but the card can flex much more of its muscle because of the 6xFSAA mode you are running.

yeah but the whole issue here was that 1024x768 with 6x AA is FASTER than 1600x1200 with NO AA

 

[OpenGL guy]

yeah but the whole issue here was that 1024x768 with 6x AA is FASTER than 1600x1200 with NO AA

That seems quite reasonable. 1024x768 is a lot less pixels to shade. This becomes a big deal because MSAA is not fillrate intensive.

Thus, if the application is fillrate limited, lowering the resolution and adding AA may actually increase performance.

 

[Ante P]

yeah but the whole issue here was that 1024x768 with 6x AA is FASTER than 1600x1200 with NO AA

That seems quite reasonable. 1024x768 is a lot less pixels to shade. This becomes a big deal because MSAA is not fillrate intensive.

Thus, if the application is fillrate limited, lowering the resolution and adding AA may actually increase performance.

hehe yes, exactly what we said above, it was just that WaltC made a strange comment on it

 

[OpenGL guy]

yeah but the whole issue here was that 1024x768 with 6x AA is FASTER than 1600x1200 with NO AA

That seems quite reasonable. 1024x768 is a lot less pixels to shade. This becomes a big deal because MSAA is not fillrate intensive.

Thus, if the application is fillrate limited, lowering the resolution and adding AA may actually increase performance.

hehe yes, exactly what we said above, it was just that WaltC made a strange comment on it

Hehe, sorry. I lost track of the thread and just replied to the last post

 

[Ascended Saiyan]

Since we are on the topic of "silly bullcrap' pertaining to the R300.I just found the weirdest thing,it turns out that the using the Catalyst 3.1's in conjunction with Directx 9 every Direct 3d program asks for internet access even though I don't have the,"Check for WHQL digital signatures" failure in doing so results in Direct 3d being Disabled but a good reboot solves that but it gets annoying after awhile.


I wonder if this is a by product of Direct X or the ATI drivers. :?


My apologies in advanced if this has been discussed before or brought up for that matter.

 

[Doomtrooper]

http://www.rage3d.com/board/showthread.php?s=&threadid=33665501