What improves what?

[Philippine Mango]

Ok so what would improved core speed/increase pipe line increase performance in, would it be fill rate? What would higher bandwidth from faster memory result in, higher quality textures, higher resolutions/AA/Anisotopic Filtering? Why would a 6600GT be faster than a 9800PRO if the 9800pro has 21GB/s of bandwidth while the 6600GT as 16GB/s of bandwidth? (Assumes the drop is because of the 128bit bus on the 6600GT). Having all of this clarified would be great and also some added detail on each of the two would also be great.

 

[Yuriman]

1. Yes, but core clock also effects the Vertex shaders
2. Higher resolutions and large textures require more bandwidth, but you need to increase core speed too. The faster you can load things into and out of ram, the better.
3. Perhaps the 9800 did not really need all of that bandwidth? Also, from generation to generation, Ati and NV add new features to their card.

Ati claims to have doubled the on-the-fly Z-Buffer compression from the 9800 series to the X700 and X800 series, from 24:1 to 48:1. Supposedly they added some new color compression, ect. Anyway from generation to generation, cores usually get more efficient per clock, even if they have the same number of pixel and vertex shaders.

 

[Philippine Mango]

Originally posted by: Yuriman
1. Yes, but core clock also effects the Vertex shaders
2. Higher resolutions and large textures require more bandwidth, but you need to increase core speed too. The faster you can load things into and out of ram, the better.
3. Perhaps the 9800 did not really need all of that bandwidth? Also, from generation to generation, Ati and NV add new features to their card.

Ati claims to have doubled the on-the-fly Z-Buffer compression from the 9800 series to the X700 and X800 series, from 24:1 to 48:1. Supposedly they added some new color compression, ect. Anyway from generation to generation, cores usually get more efficient per clock, even if they have the same number of pixel and vertex shaders.

thanks for the informative post, is fillrate controlled more by the core? I'd assume that the 6600GT would be a beast of a card had it had the 256bit bus, no?
Also what would you say a 9800pro would benefit more from, a core clock increase or memory increase? I've been trying to create an excel sheet (posted in another forum section) creating the relationship between increased core speed performance and increase memory speed performance and how memory needs to be increased by a certain amount for the core speed increase in performance to be tangible etc..

And I wanted to do this because I was noticing a pattern in HL2 Video test where if I increased the memory alone to 10mhz faster than an XT, I'd see a 1FP increase, increase the core alone to about the speed of an XT and I saw a 4FP jump, but increase it by 10mhz more and no improvement. I dunno, just wanted to see a relationship created among this information I've collected and it seems it's impossible to do in excel...

 

[Philippine Mango]

bum

 

[BFG10K]

The 6000GT is faster than a 9800 Pro in modern games because they are shader limited and its higher clock speed directly relates to how fast it can run pixel shaders.

 

[Regs]

If I had time I would sit down with you and explain the 100's of pipelines, driver controls, process ordering, cache logic, and core logic embedded in both the 9800p and 6600GT....but I don't.

 

[Munky]

So, in short the 9800p has plenty of mem bandwidth, but games only become bandwidth-limited at high resolutions with AA. So, at 1600x1200 with 4xAA the 9800p is faster in a lot of games, but neither of those cards can offer playable fps at those settings. At lower resolutions the games are more dependent on gpu processing speed, so a core clock increase on the 9800p would benefit it a lot. Dont forget that a 9800p runs at 380mhz, and the 6600gt runs at 500mhz, so that gives the 6600gt a higher fillrate and shader processing ability.

 

[Philippine Mango]

Originally posted by: Regs
If I had time I would sit down with you and explain the 100's of pipelines, driver controls, process ordering, cache logic, and core logic embedded in both the 9800p and 6600GT....but I don't.

I thought they had 16 pixel pipelines, where is this 100s?

 

[Gamingphreek]

Originally posted by: Philippine Mango

Originally posted by: Regs
If I had time I would sit down with you and explain the 100's of pipelines, driver controls, process ordering, cache logic, and core logic embedded in both the 9800p and 6600GT....but I don't.

I thought they had 16 pixel pipelines, where is this 100s?

There are more pipes in this world aside from Pixel Pipes.

-Kevin

 

[dug777]

hehe, note that a 9800 pro is FASTER than a 6600GT with AA (at arguably playable fps) in farcry at 1024. According to the AT benchies anyway.