R420 IQ Comparison

yes you probably want the registered version for more freedom. You can do movies with the demo version but it's limited in options

file format is:
12bit per pixel, avi
compression = fraps , don't know what that means but i would say from the size of the file there is no additional compression going on

 

Maybe its just me but don't you guys think some of you are asking a lil much of Dave. I mean first we wanted pictures, then since those didn't show what some of you wanted to see, you want videos. Then those won't show what you want to see, so you'll want him to program a demo. I read through the whole thread and I don't think I saw one person say thank you(If you did, forgive me). All I saw were people asking for more this or more that. I'm sure Dave has a lot of other things to do right now beside taking screen caps so we can argue for another 30 pages. Anyway, I can't tell the difference. I'm not an IQ expert but I do have good eyes and I just can't tell the difference aside from the obvious things like the change in sky, gun position, and lack of grass on the lower end card. Anyway, I guess I'll step off my soap box now.

Thx Dave

 

Areas where you can see the R420 appears to be doing less blending:

http://www.reflectonreality.com/nv17r420/

Note, I just took the original pics and added red circles, then saved them as PNG again.

 

Thank you Dave!

 

if you wanna show the quality of trilinear filtering or the lack thereof then you really need to show it in motion.

 

does that cause any extra aliasing though because if it doesn't then that's actually a benefit.

 

I'm getting a little confused. How does R420 decide to use full Tri vs the optimized one? Is it when the GPU is taxed for performance as well as qualities of the mip maps?


The major difference in the second picture that i see is the light inside the entrance seems to be slightly better/cleaner with the NV17. Is that something unrelated to filtering?


Also is there any difference in blending output do to different number precision i.e. int, fp16, fp24, fp32?

Thanks

I hope this post doesn't get locked.

 

The driver analyses the texture when it's uploaded and sets an optimisation flag for that texture. It appears that this optimisation flag is not set for textures that are provided with mipmaps (like colour test pattern mipmaps).

 

The indignance at B3D is reaching epidemic proportions! Seriously, I tend to thank Dave for most of his work (at least in my PMs and initial article thread responses), so if I don't, it's either because I hope it's assumed by now or simply that I forgot. I'm sure most regulars think this way, too, just as Dave must appreciate the contributions of Coulourless and MDolenc and X-mas and others who have added to his test suite.

But I also think Dave wants to get to the bottom of this as much as some of us, and he just happens to be the guy with the X800P ATM.

But if you insist on abject grovelling, I'll just echo what crushinator said. I'm doing the Wave right now. Woo.

Dave, are Bink movies an option? The HL2 Bink movies looked extremely good at native res. I'm not sure how easy they are to use in-game, though, or even if they're available for use to just anyone. (I'd *guess* the Bink ppl wouldn't mind letting B3D use it, though, in exchange for the publicity they'd get with this developer-heavy crowd.)

 

Firstly thanx to Dave for the images, very informative.

I guess this proves one thing, people will see what they want to see. Anyone that is rational and had good eyesight can easily observe that ATi is doing true trilinear, at least in these screen grabs. Sadly some out there just can't wrap their brains around the fact that ATi is using an adaptive routine to save bandwidth, or cheating depending on who you ask.

Why not have the card render every pixel whether it needs to or not, without any smart, adaptive routines just to be fair.

 

lay off the drugs man. there is NO difference in those red circles.

 

My subjective opinion is that the R420 actually looks better. From what I can see it is slightly sharper and the mipmap transitions look just as good. It could just be my monitor though.

 

First off, thanks Dave.

Second off, my meager eyes see no difference.

 

To be honest, I saw some very minor differences. There is a small line that does look different in the middle of the left oval. I'm not sure if it is due to lighting or filtering. However, the right oval looks like a lighting issue. I couldn't see any difference on my (very crappy) LCD at work - I had to come home and look again with my 19" CRT to clearly see either difference.

 

maybe you just need glasses? cuz i have looked on my monitor, and 3 of my friends monitors and there is NO difference in those red circles

 

You could be a little nicer in your postings.

 

too much green cordial being dolled out at the day care centre I think...

 

I can see a very very small difference the R420 shot - it's slightly shaper in thoughs red circle areas.

I have to put my face very close (and even strain my poor eyes to see the miniscule difference).

Thank you Dave for your time.

 

Heh, as in the image makes you go blind?

I can't look at that for more than a few seconds without my eyes hurting bad, so I ain't gonna be able to vote.

BTW, since this is a pattern/synthetic test and not from an actual game, how do we know for sure this is not one of those "special cases" where the ATi optimization is disabled to get the "correct output?"