The easiest way to see the "quality" between those three shots is simply to look at the center of the pic. Ideally, the image would never turn beige, but it does on each card at a different distance...the further, the better (basically, the further, the higher degree of anisotropy is being used).
State of the Graphics Industry Rant
- Thread starter Sabastian
- Start date
Joe DeFuria
Legend
Right...you can see GeForce and 9700 have similar "quality", while the 8500 is lacking. If those pics were not "rotated", they would all look very similar. That is showing the "worst case" for the 8500.
Given this, why is Russ asking if the "fixed" 8500 drivers decrease performance? (Because nothing has changed to my knowledge wrt 8500 aniso.) Apparently, Russ doesn't actually have any quality issue with 8500 aniso, so why is he arguing about aniso implementation in the first place?
Given this, why is Russ asking if the "fixed" 8500 drivers decrease performance? (Because nothing has changed to my knowledge wrt 8500 aniso.) Apparently, Russ doesn't actually have any quality issue with 8500 aniso, so why is he arguing about aniso implementation in the first place?
Joe DeFuria
Legend
It's admirable that you two are covering for him
, but here's Russ' original "unanswered" question I assume he's referring to:That would not make sense if he was talking about the R300. Clearly, he's asking the question about the R200 (Radeon 8500 output).
I also don't believe anyone (R200 users or otherwise) claiming that R200 aniso was "fixed." I can only conclude he reached that conclusion himself based on the picture?
*scratch* Fuz said "seems fixed to me", which I took to mean the 8500 was, but I guess he must have been talking about the 9700.
If it wasn't fixed, then ignore the question.
If it wasn't fixed, then ignore the question.
Joe DeFuria said:
why, what would make those IHVs deliver that variety? why would they need conforming to the specs if their non-comformant product received higher fps benchmarks ratings and generally was selling well? what options would fligh-sim players have in this case?
there's a difference - quake visual requirements were met and the basic quake speed requrements were met too, those >100 fps would give you more comfort. whereas the visuals for those flight-sim players were never met, they never saw what they liked in the first place.
read my really basic explanation on rudimentary defintion consistency in one of the previous post. if any of the api docs had explicitly stated 'pixels of the same anisotropy degree may actually get different anisotropic filtering across the same hw implemtation' then my rudimetary definition consistency would not hold. still, they state so nowhere (from all dx and ogl specs i've read), and a developer would expect the hardware to behave consistently. which would not happen in the r200 af case.
why, you want my permit to quote it?
are you evading my point or you just missed it?
let's repeat it - hw is made with both consumers and developers in mind. developers are a numerically-small but significant part of IHV's consumers. JC is a guy who can say 'this board will sell, this won't'.
as re his position on correctness - i don't remember him being particularly fond of the early r200 drivers - and i don't think that was because they would give him too many a BSOD, more like there must have been something with correctness..
my, why all of a sudden this change of heart? according to all you said up to here as long as the red sphere gives you the higher fps and is ok with your personal taste then it should be ok, no?
darkblu said:
You are treating this like it doesn't render at 45 degree angles, atleast it seems so to me.
Their options are to live with the tradeoffs. The same as with the other tradeoffs mentioned in this thread. What about the GF user that has to go to a lower resolution to use anisotropic filtering at all and get acceptable performance? Your example ignores that aspect of the issue. That is my understanding of Joe's point for the user's viewpoint.
See, you are proposing a statement like "the visuals for those flight-sim players were never met" as if that statement is a valid objective criteria.
Does the definition even cover whether they do or do not have to consistent anisotropic filtering coverage? Does it say anything about angles? I thought it was established that rip mapping (which the 8500 doesn't do), which has a much more noticeable set of limitations for rendering, fits the definition of anisotropic filtering...this seems to make your argument a bit invalid doesn't it?
To me it reads like you are using definitions backwards. In any case, I think this is a separate argument from the user's visuals "never being met". I'd be interested in your take on my rip mapping comment, however.
Thread cleaned up a little.
Let's keep this civil, gentlemen.
Let's keep this civil, gentlemen.
Joe DeFuria
Legend
I already covered that. The same thing that makes the IHVs deliver the variety of products we have on the market right now. Competition. Every IHV is looking to differentiate the product, while making it as attractive as possible to the widest audience as possible.
To do that, you
1) Lower cost
2) Increase rendering quality
3) Increase performance
4) Add additional "features"
Note that these things compete against one another. Increased quality usually dictates higher cost, as does increased performance. Higher quality usually dictates lower performance.
Thus, decisions need to be made (based on market research, or whatever) as to which things take priority, what trade-offs will be made, what features can and can't be included, etc.
It's all about market share. If "every" IHV decided to go the Radeon8500 aniso route (as you imply would happen), then that would by definition create a market for a new vendor to create a card that would get 100% market share of the flight-sim market, by virture of having a differentiated product.
That's the "magic" of a free economy and competition. It regulates itself (for the most part).
Says who? I ould just as well argue that "basic flight sim" quality requirements are handled with bilinear filtering. Ansio just gives "comfort".
There's no difference. There's no way that you can make blanket statements about the "acceptability" of performance / quality for all users.
Edit...(hit "submit by accident before finishing the post)....
As I said...hardware is made with developers in mind to the extent that developers use them to also sell products to consumers.
JC can say all he wants...that doesn't seem to impact the tons and tons of TNT's, Rage Pros and GeForce MX cards being sold....
Uh...yes. Correctly rendering the console. I wouldn't be fond of the drivers either if they couldn't get past the console without breaking down. Or do you think R200 didin't "aniso" the console correctly, and at this, Carmack wasn't pleased?
Technically, yes, it would be OK. But will consumers be "pleased" when they fire up Doom3, and are greated with red sphere's, no textures, etc? No.
Aside from me thinking you would apply some common sense to the issue, my point is, there are of course limits to what consumers would be "pleased" with. If one IHV believes that just skipping textures and replacing all meshes with rectangular red blocks at 10000 FPS would be "pleasing enough" to consumers, best of luck to them.
Again, the market will take care of itself...
Joe DeFuria
Legend
Heh..we did..but then you brought it up AGAIN practically demanding for it to be addressed.
Nah, my fault for wading into the discussion assuming facts based only on this thread.
I don't have a radeon, so I'm pretty ignorant about the driver of the week for ATI cards. (Though, quite honestly, I'm pretty ignorant about the NVIDIA driver of the week also, even though I own one) Plus, I try to stay away from who's card has a bigger penis type arguments if I can, so I'm not completely up to speed on who's card has which [hyperbole]earth shattering[/hyperbole] problem.
I don't have a radeon, so I'm pretty ignorant about the driver of the week for ATI cards. (Though, quite honestly, I'm pretty ignorant about the NVIDIA driver of the week also, even though I own one) Plus, I try to stay away from who's card has a bigger penis type arguments if I can, so I'm not completely up to speed on who's card has which [hyperbole]earth shattering[/hyperbole] problem.
Chalnoth said:I think I just figured something out.
It has to do with the moire patterns seen in Unreal Tournament on a GeForce4 with anisotropic filtering disabled.
It turns out that it's directly related to texture compression. Without texture compression, there is next to no aliasing. With TC, there is lots of moire apparent (in certain scenarios, anyway) when aniso is disabled.
The only thing that I could imagine this could be a result of is that with TC enabled, MIP maps smaller than 4x4 are not created. 4x4 is the size of one block of a DXTC compressed texture. It should be apparent that MIP maps smaller than that should be created in an uncompressed format, though I'm pretty certain that UT does not do this (the only way it could would be within the driver...).
Now if only I could post some screens to show the issue
Compressed mipmaps smaller than 4x4 are stored as full 4x4 blocks, except that only the upper left part of it is used. You can't use different formats for different mipmaps.
A little late jumping in the fray but the following is the only thing that struck me as relevant :
Personally, I think the 8500's limitation of bilin + aniso is not an "issue" to be hyped, not when the resultant displayed image really don't show the mip bands associated with bilin (which, the mipmap bands that is, really should be the only complaint about the bilin + aniso limitation... then again, what's the complaint for when you really don't see the mipmap bands?!). The Z rotation ''issue" is an annoying one however but only so when it applies to specific types of games as well as specific instances in all games.
Sometimes when issues are not noticed by the majority, they may be non-issues.Chalnoth said:
Personally, I think the 8500's limitation of bilin + aniso is not an "issue" to be hyped, not when the resultant displayed image really don't show the mip bands associated with bilin (which, the mipmap bands that is, really should be the only complaint about the bilin + aniso limitation... then again, what's the complaint for when you really don't see the mipmap bands?!). The Z rotation ''issue" is an annoying one however but only so when it applies to specific types of games as well as specific instances in all games.
Yes, I'm sure there are many people who don't notice the problems. That doesn't change the fact that they are there. I feel it is important, at least for those who want to know, to educate consumers as to all the advantages/drawbacks of the hardware.
Besides that, I'm also of the opinion that image quality is only as good as the worst-looking scenario. This was why I liked 32-bit color (I refused to run in 16-bit even on my TNT 16MB). The drawbacks in most games at the time weren't all over the place, but I didn't like them when they were made apparent.
This is also why I've spent a fair amount of time attempting to solve the DXTC issues in Unreal Tournament, as well as adding alpha blending. While neither problem looked bad except in just a few scenarios, I didn't like it.
And back to anisotropic filtering. I find that I can play for a couple of hours before realizing that it's not enabled at all at times. It doesn't surprise me that many don't notice the 45 degree angle games, especially if they only play somewhat older "boxy" FPS games, or racing games, or other titles that usually don't have slanted surfaces. Right now, I'm sure I'd notice the problems in Morrowind from time to time. It wouldn't be often, but it would annoy me. That is enough.
In sum, even if it isn't "much" of an issue that's rarely seen, if the problem is seen at all, it should be fixed.
Besides that, I'm also of the opinion that image quality is only as good as the worst-looking scenario. This was why I liked 32-bit color (I refused to run in 16-bit even on my TNT 16MB). The drawbacks in most games at the time weren't all over the place, but I didn't like them when they were made apparent.
This is also why I've spent a fair amount of time attempting to solve the DXTC issues in Unreal Tournament, as well as adding alpha blending. While neither problem looked bad except in just a few scenarios, I didn't like it.
And back to anisotropic filtering. I find that I can play for a couple of hours before realizing that it's not enabled at all at times. It doesn't surprise me that many don't notice the 45 degree angle games, especially if they only play somewhat older "boxy" FPS games, or racing games, or other titles that usually don't have slanted surfaces. Right now, I'm sure I'd notice the problems in Morrowind from time to time. It wouldn't be often, but it would annoy me. That is enough.
In sum, even if it isn't "much" of an issue that's rarely seen, if the problem is seen at all, it should be fixed.
Most of your post is about individual (=yourself) priorities.
The most important point to make, when it comes to websites/press writing reviews, is to state whether something "out-of-the-ordinary-in-specific-situations" (i.e. 8500's bilin + aniso and Z rotation) is "important" enough. More importantly, it is how a website/press presents such "findings/studies"... whether they want to "stress" on it or not.
No doubt, once someone finds out about any anomaly, it gets hyped. And when it gets hyped, it gains "recognition". Which means to say, as is obvious, once a "fault" is found, it is hard to ignore.
BUT only in the context of what you want to talk about. You want to talk about 3D, fine, the 8500's "faults" is noticeable (because you look for it or it has been brought to your attention and not because you found out about it yourself ... should you curse or praise the websitye/press for such "revelations"?) and should be discussed at length.
You want to talk 3D, fine. You want to talk games (which drives 3D forward), it can be very different. I did not care for aniso nor AA of any kind when I first played MOHAA (resolution and other IQ tidbits mattered more). But if I bought the game because it is popular and I don't really game but bought it because I wanted to use it for video card "studies/reviews", I start looking for : (a) how good the game looks OOTB; and (b) start looking for rendering problems.
As a video card reviewer, especially nowadays when there is no such thing as a "bad video card", the only thing you get to do if you want "recognition as a journalist" (=page hits) is go looking for problems.
That does more to promote arguments and quarrels in forums than the advancement of 3D or "journalism".
The most important point to make, when it comes to websites/press writing reviews, is to state whether something "out-of-the-ordinary-in-specific-situations" (i.e. 8500's bilin + aniso and Z rotation) is "important" enough. More importantly, it is how a website/press presents such "findings/studies"... whether they want to "stress" on it or not.
No doubt, once someone finds out about any anomaly, it gets hyped. And when it gets hyped, it gains "recognition". Which means to say, as is obvious, once a "fault" is found, it is hard to ignore.
BUT only in the context of what you want to talk about. You want to talk about 3D, fine, the 8500's "faults" is noticeable (because you look for it or it has been brought to your attention and not because you found out about it yourself ... should you curse or praise the websitye/press for such "revelations"?) and should be discussed at length.
You want to talk 3D, fine. You want to talk games (which drives 3D forward), it can be very different. I did not care for aniso nor AA of any kind when I first played MOHAA (resolution and other IQ tidbits mattered more). But if I bought the game because it is popular and I don't really game but bought it because I wanted to use it for video card "studies/reviews", I start looking for : (a) how good the game looks OOTB; and (b) start looking for rendering problems.
As a video card reviewer, especially nowadays when there is no such thing as a "bad video card", the only thing you get to do if you want "recognition as a journalist" (=page hits) is go looking for problems.
That does more to promote arguments and quarrels in forums than the advancement of 3D or "journalism".
I disagree there Rev. Case in example I haven't read so far anywhere about that one:
http://users.otenet.gr/~ailuros/RtT.jpg
Most probably a driver "quirk" (at least in D3D), but in a generally antialiased scenery just one little shadow or car mirror being a jag-fest sticks out immediately. That's with AccuView of course.
Similar threads
