[H]OCP does the FX5600 & FX5200 Ultras

AFAIK Sean Pelletier did that review. Kudos to him.
(and nevertheless Kudos to Brent for his efforts on the other reviews)

 

Yeah it was Sean. Very well written review indeed. Almost Beyond3Desque!

 

One question: all of his "iq" investigations used the color bands as "proof" that one is better than the other.

How does that bear our in practice?

Assuming your mip map sampling mechanism provides precise and continuous results, you could have single pixel transitions (i.e. no blending) between the mipmaps when using anisotropic, no?

Or is it too difficult to obtain that precise and continuous result?

This kind of relates to this thread by Reverend.

 

Well of course I found that rather refreshing compared to other reviews. The comparison using 3DMark2003 as apposed to 3DMark2001 most appropriate for these DX9 cards. Nice objective comparisons on the IQ and clearly they point out that indeed ATIs DX9 products from last year are better of course we all know that anyhow. But it was nice to see a non PR type of review of these cards. Also I can't be more happy about the noting of nvidia marketing of "cinematic" rendering for $79 but am kind of disappointed the $79 card didn't make an appearance so we can see just what kind of "cinematic" experience people will get for $79 as opposed to the $150 "ultra" gffx 5200. Clearly ATi pulls ahead in all areas with the Radeon 9500 pro that sells in relatively the same price range of the NV31"ultra", NV34"ultra". I think they did a good job as well.

 

Can't really argue with that at this point. I would really like to see some objective comparisons between the Radeon 9600 pro and the Radeon 9500 pro. The reduction in transistors on the RV350 compared to the R300 core is so much it is almost alarming but the clock speed should make up for quite a bit.

 

You missed my point.

I'm asking if the color bands are really indicators of "IQ". Sure, they show the shape, and also the blending transition between the mip maps, but does that really translate to anything in particular, particularly the blending transition?

Wouldn't a more telling test be a clean checkerboard and/or grid pattern? With that you could see that the samples line up properly and (hopefully) you cannot see any abrupt transitions.

 

Russ,

they do comparison shots with UT2003 between aggressive and application mode as well without showing MipMap lines. You can cleary see the difference in Ansio being performed.


but yeah, basing it strictly on mip map lines would be bad.

 

I agree, I kept waiting for him to prove to me that Balanced actually looked worse that ATIs quality "in game". Not only that, but from what I understand, a lot of people use ATIs "performance" while playing their games. An investigation of this seems necessary considering its what almost all other reviews used when compairing balanced.

Dont get me wrong, I loved the review, but it almost seems that now we're performing "synthetic" IQ compairisons instead of real ones. No mention was made of ATIs degree sensitive AF usage either.

 

Martrox, please don't turn this into an ATI vs. NVIDIA thing, and particularly don't try to make this into a fan boy thing.

I'm interested in this because of Dave's poll here, Brents useage of the colored bands as an "proof", Reverend's thread mentioned above, and even Beyond3d's useage of those bands as "proof". It suddenly has become the test de riguour, and I'm wondering if the "measurements" are really valid and if and how they could be misleading.

 

I also agree, compairing the 9500Pro to the 5600Ultra isnt necessarily "wrong" because they are in the same price range, but it certainly is akward to say the least. Not only that, but people keep making references to the DX9 for $79 when mentioning their reviews on the 5200Ultra, when its obvious the $79 itteration will have a significant performance decrease compaired to the Ultra.

I think we'll just have to wait till the 9500Pro really gets phased out completely until those reviews will really be accurate though. Im sure phasing it out will be good for ATI, but not for us poor consumers

 

You have a good point, since still shots can't show whether it is a problem or not.

So I asked Brent earlier whether the boundries between mip-maps are easy to notice in motion in Balanced mode and here was what he replied:

 

I'm sorry, Martrox. The statement of yours I quoted above (questioning why I didn't ask about this when NVIDIA was pointing out no trilinear w/anisotropic on ati) sounded like you were suggesting I was only interested in this because of the parties involved and the positions they were in.

If that wasn't the case, mea culpa.