IQ Filtering Test: Which is your Preference
- Thread starter Dave Baumann
- Start date
bloodbob said:I gotta love how at the start of the POLL brilinear was winningI'll put up some interesting modification to these images later.
it might be due to that less aliasing in from the nv40.
Remember it uses a different LOD calculation with 8 segments and not 4, and the 45° angles (and also towards the sides of the screen on a flat surface) get a fairly notable +ve bias compared to the refrast and any other card. With a texture as prone to aliasing as this one this extra bit of blur does actually help it look a bit nicer.
I love how diffrent these two polls are .
you go from 46% of the voters not able to tell the diffrence in one poll.
To 38% picking board b as the iq to perfer.
You would think of the pic r420s poll was accurate you would see both polls in favor of can't tell the diffrence .
But hey whatever floats your boat.
you go from 46% of the voters not able to tell the diffrence in one poll.
To 38% picking board b as the iq to perfer.
You would think of the pic r420s poll was accurate you would see both polls in favor of can't tell the diffrence .
But hey whatever floats your boat.
Evildeus said:I disagree.
The fact that people can't say which one is the R420 has nothing to do with their preferences
if they all look the same or they can't tell the diffrence then how would they have a preference ?
If two cerals tasted the same and looked the same to me then how would i have a prefrence ?
That is why it doesn't make sense .
Humm, no, they are not.
They don't look the same or even they look almost the same, but you don't know or can't tell which one is the X800, that means you don't have the background to make a judgement.
On the other hand, you can't tell which one you prefer in IQ, that has nothing to do with the fact you can't tell/ don't know which card produce the image.
They don't look the same or even they look almost the same, but you don't know or can't tell which one is the X800, that means you don't have the background to make a judgement.
On the other hand, you can't tell which one you prefer in IQ, that has nothing to do with the fact you can't tell/ don't know which card produce the image.
Evildeus said:
I disagree.
I voted the third option because by studying the shots i could see no diffrence.
But now from not seeing a diffrence suddenly I like one more than the others ?
If they all looked the same to me how could i pick 1 to be my prefrence.
Even if you take the full number of diffrent votes out of i don't know. Its still the majority in that post.
So something is obviously wrong with one of the polls .
As i said i think people who knew which was which voted towards what suited thier agenda (and that goes both ways for nvidia and ati users )
You are one of the guy who voted D in both because you don't see a difference, but many people did so.
that being said, other people can't say or don't know which one was rendered with the R420 but can also have a preference in IQ.
I really don't see what's difficult to grasp over here.
For exemple, you have 3 bottle of wine. It's like saying which wine is a Bordeaux, and which one do you prefer. The 2 question have not the same implications.
that being said, other people can't say or don't know which one was rendered with the R420 but can also have a preference in IQ.
I really don't see what's difficult to grasp over here.
For exemple, you have 3 bottle of wine. It's like saying which wine is a Bordeaux, and which one do you prefer. The 2 question have not the same implications.
I disagree. Because obviously a portion voted for they can't taste a diffrence (wine thingy here) So either the option isn't clear enough because now you have to possible things and it needs to really be to options. or the result is skewed.
If a portion voted for I can't taste a diffrence and yet some of those went on to pick wine 3 as a prefrence tha wouldn't make sense now would it .
Although tasting isn't that great because can't tell would lead to the same thing. But your right with the graphics card they can still have a prefrence. The only question is how many foted for d 1 or d2
If a portion voted for I can't taste a diffrence and yet some of those went on to pick wine 3 as a prefrence tha wouldn't make sense now would it .
Although tasting isn't that great because can't tell would lead to the same thing. But your right with the graphics card they can still have a prefrence. The only question is how many foted for d 1 or d2
Bambers said:bloodbob said:I gotta love how at the start of the POLL brilinear was winning
it might be due to that less aliasing in from the nv40.
Remember it uses a different LOD calculation with 8 segments and not 4, and the 45° angles (and also towards the sides of the screen on a flat surface) get a fairly notable +ve bias compared to the refrast and any other card. With a texture as prone to aliasing as this one this extra bit of blur does actually help it look a bit nicer.
Well the angles didn't really come to much into effect since the plan was flat horizontal and the lod calculations aren't that much different ( . I think it was because the NV40 does less optimisation by selecting the higher quality mipmap then the R420 does so therefor thier was less aliasing.
Anyway on to those pictures I promised. ATI selected a texture with an equal amount R,G,B but when those colours are used they appear to be used at maxium saturation. So what happens is as the LOD decreases the saturation will decrease ( when you do an average against all pixels of the same LOD ). So a good way to look at the mipmap is to use saturation. so here is the saturation component of image when converted into the HSL colour space.
http://www.lexicon.net/mccann/asat.png
http://www.lexicon.net/mccann/bsat.png
http://www.lexicon.net/mccann/csat.png
I hope you find them useful.[/img]
But those people voted con't tell/don't know in both. So what the problem? But you can prefer one, and still don't know which one is Bordeaux.jvd said:
Sure, but that's not what i said, and you can't prove they didn't have a preference
Thanks for saying that i'm rightAlthough tasting isn't that great because can't tell would lead to the same thing. But your right with the graphics card they can still have a prefrence. The only question is how many foted for d 1 or d2
Evildeus said:Yeah, 9800 first, which is normal IMO, second brilinear
What that doesn't show is that, looking over the thread, most of the people who picked B listed C as their second choice. Wonder what that tells you, eh?
I picked B, but I don't know about anyone else, but I had to carefully study the images before making my choice. If this is worst case, then I'm happy. I'd really want to see a video clip of this in motion (or several clips), I mean, after all, that's why people really disliked brilinear.
Something else to take into account here is that people at Beyond3D know what to look for (which in itself somewhat invalidates any attempt at a blind test). The results are interesting, but hardly conclusive.
JVD you are wrong. I could not tell which was what card but I stil have a preference. It is more like you hear two songs and do not know who sings which one but you can still prefer one to the other quite easily. I thought B was the best on this on BTW. I simply put different ones up and walked a bit away squinted and so forth
bloodbob said:
ok oops i was talking bull a bit however my point is still relevant to an extent.
Although both nv3x and r420 produce the pretty much same image in a tunnel on a flat plane the nvidia cards will produce a curve while r420 produces a kind of v shape and so even with the bri of nv38 the edges are still using a lower detail mipmap than r420, removing some of the aliasing.
your pictures show this pretty well, where the cross shapes produced by the moire at either edge near the first mip level are. With r420 at those points the mipmap levels are further back than on nv38 leading to more aliasing and so c being the least popular.
however if the textures used were less prone to aliasing but showed up blurring well then r420 could be the preferred image.
on the other point here it's quite possible for people to not know which is r420 and still have an IQ preference. That's because really it takes extra knowledge to work out which card is r420 and even then it's still a little hit and miss since we had no idea what cards were being used to compare against. Similarly someoen could be able to work out which was the r420 but not think the difference was large enough to actually have any preference in image.
Huh?
Read.
you spent alot of time saying "Althornin, if people can tell a difference then they must prefer trilinear" even though that wasnt part of the poll. Now that these two polls are done, giving you the information you pretended was present in the first, you have changed to saying exactly what i was saying. Nice. All that without ever admitting you were wrong in the first place. Good job, but it aint fooling anyone.
Read.
Of course, in the other thread (on the poll that just asked people which one the r420 is, that had no "preference" component) you spent alot of effort telling me how i was a moron for saying "hey Evildeus, the poll doesnt say anything about which is better!".
you spent alot of time saying "Althornin, if people can tell a difference then they must prefer trilinear" even though that wasnt part of the poll. Now that these two polls are done, giving you the information you pretended was present in the first, you have changed to saying exactly what i was saying. Nice. All that without ever admitting you were wrong in the first place. Good job, but it aint fooling anyone.
Similar threads
- Locked
