PDA

View Full Version : I wish reviews would seperate AF/FSAA performance comparisons..


ChrisRay
11-12-03, 08:06 PM
Main Reason I would like this. Is I really think reviews are being lazy. From my tests with the Geforce FX 5900. The AF is still much slower than my 9500 Pro (% Performance Hit wise) While the FX AF is very usable still. I think these things should be noted in reviews


While most would probably use both on high end cards, It would be nice to where we can see strengths and weaknesses.

I Must question Nvidias "Performance' Mode. As it doesnt offer any real speed boost to Nvidias "Quality" Mode. While Nvidias High Performance Mode offers a considerable performance increase Over "Peformance And Quality"


Tomshardware was doing Seperate AF/FSAA tests and the combined FSAA/AF Tests.

Once again, I feel reviewers have been somewhat lazy lately. As to be kinda honest, I was a little unprepared for the performance of AF, As it was under my expectations, While the AA performance was above my expectations.

Instead of throwing them all together, Because I think it would give users a more educated . While reviewers are striving to give more detailed reviews, They seem to have lost site of other important details. While taking 2 steps forward, They take 1 step back.

I want Seperate AA/AF charts :(

Rytr
11-12-03, 08:57 PM
My review of the FX5900U did not show a significant hit on frame rate while AA did, especially at 4x and above (6x showed as much as 50%in some cases). This was with the 45.23's. I have not checked the 52.16's.

My 9500Pro is simular with AA providing the main hit on frame rates with all CAT drivers. However, I agree with you that whatever shows a strain on frame rates should be covered in reviews.

Just as a side note, I am currently running CoD at 8xAA and 2xAF with the 5900U and getting an averages running between 44 and 48 FPS.

ChrisRay
11-12-03, 09:01 PM
Originally posted by Rytr
My review of the FX5900U did not show a significant hit on frame rate while AA did, especially at 4x and above (6x showed as much as 50%in some cases). This was with the 45.23's. I have not checked the 52.16's.

My 9500Pro is simular with AA providing the main hit on frame rates with all CAT drivers. However, I agree with you that whatever shows a strain on frame rates should be covered in reviews.


AA is always going to produce a bigger hit than AF, But thing I was saying is. The performance hit for AA was less than I expected. The performance hit for AF was more than I expected.

It evens out in the very end. But not really heh.

Rytr
11-12-03, 09:29 PM
I understand what you are saying. That is why I added the bit about CoD which surprised me. It runs very smooth at these settings and looks very good to boot!

That's my problem, I can't get it to even out! :D

vandersl
11-12-03, 11:15 PM
I'm not sure what you're saying Chris.

A fair number of reviews bench AF separately from AA. It's late, so I won't dig up links, but usually the bigger sites do this.

Personally, I don't think there's any point to doing ANY benchmarks without AF. And I don't think there's much point to doing AA without AF. The only exeption is when AF or AF+AA brings the FPS below a minimum acceptable target.

For example:
- if AF+AA is above 30fps, bench game with AF+AA enabled
- else if AF is above 30fps, bench with AF enabled
- else bench with no AF or AA

The flip side also holds. If a game is only getting 30fps with no AF and AA, then enable AF+AA, and see if the FPS drops. If not, its 'free'.

I don't think we're ever going to get the full matrix, i.e.
3 resolutions (10X7, 12X10, 16X12)
3 AF settings (2X, 4X, 8X/16X)
2 AA settings (2X, 4X)
= 18 benchmarks per game

Hanners
11-13-03, 05:25 AM
I agree with you in a way, but the general consensus is that people will run with AA and AF enabled, not one or the other. Also, nVidia and ATi themselves recommend testing with both options turned on, and to be honest high-end cards should be able to cope with both in the majority of games.

Finally, there is the sheer amount of testing required - It's basically three times more work to do separate AA and AF scores as well as an AA and AF score - Very time consuming stuff. That's not to say that it shouldn't be done, but it can be impractical.

ChrisW
11-13-03, 05:34 AM
I think reviewers should find the highest image quality settings where the game runs at >30 fps. That's all I really care about. I can care less if one card can get 200 fps vs. 220 fps. I want to know what the maximum image quality I can get from each card and still get at least 30 fps in my games.

Quitch
11-13-03, 05:48 AM
AF and AA are designed to be used together, therefore the card should be able to handle running both. Sure, you can bench with only one, but it's not much use.

I think reviews SHOULDN'T contain seperate AF/AA benches, they should be done as one.