[H]OCP does the FX5600 & FX5200 Ultras

First,

I have to make a public apology to HardOCP. So here it is: "I Apologize!"

That apology is for my harsh with respect to HardOCPs handling of the 3DMark03 epsiode.

I still strongly disagree with how they handled that, but after reading this review (and then going back and reading the 9800 Pro review), I'll have to withdraw my accusations of blatant nVidia bias and favoritism.

This preview, as did the 9800 Preveiw, showed a legitimate attempt at getting a "fair" comparison, and went specifically against the "suggestions" of nVidia to benchmark the "aggressive" mode.

Kudos to HardOCP. You've won me back as a reader.

In fact, if anything, I'd say this review is slightly biased toward ATI with the aniso comparisons. Why? Because although the mip-map blending aspect was well covered and compared, the other aspect of ansio filtering (the filtering algorithim itself) was not.

And in that aspect the "Application" mode of the FX cards is superior to the ATI card.

I have not seen both cards in action side-by-side, so I do not know which of the two comparisons:

1) FX Application vs. Radeon Qualitry

or

2) FX Balanced vs. Radeon Quality

gives the "best possible" apples to apples comparison. HardOCP says its 1), and that is definitely true with respect to the mip-map blending. However, I'm not sure if they also looked at the filtering algorithm aspect of each as well. (Rotation...) It is celar that NEITHER of those two comparisons is apples-to-apples, but if you had to pick one for comparison, you'd have to see them in motion in several scenarios to make a decision.

In any case, HardOCP benchmarked ALL THREE modes of the GeForceFX, which is something I've been wanting to see. !) Though at the same time, they didn't include benchmarks for the Radeon "Performance" mode, which I think would have been prudent for the same reasons.

The overriding disappointment is with nVidia.

I'm not disappointed that they are offering new filtering options that trade-off performance for quality. More choice is good. I am a bit disappointed that you can no longer "force" the high quality mode from the control panel. Obviously, that's implemented that way for the SOLE reason to make it less likely that reviewers will actually test that mode.
The reason (performance) is obvious.

I am most disappointed though, that nVidia suggested that the aggressive mode be used for benchmarks. That's only a slightly less repulsive than SiS having poorer texture filtering by default.

 

In-depth, but lacking some breadth: I was left wanting ATi vs. nV AF-comparison gif's. It does seem that [H] is swinging almost too strongly in ATi's favor (Sean's conclusion is pretty damned strong).

But a good review overall. [H] is becoming a great source of in-game IQ comparison screens. Good on ya, Sean.

 

Also I'd like to add that these were mid-range and low-end cards being tested. A more accurate portayal might have been Balanced v. Performance, because these are the settings a more budget oriented game will likely be using, as opposed to 4xAA and 8x high quality AF. When benchmarking a 9800 or GFFX then I will agree that the highest IQ is the main issue, but on a mid-range card, this may not be the case.

Infact, no real comparisons were made for ATIs Performance AF at all. Do these not also use Bilinear? (Just asking, I really dont know)

Hehe, I hate criticizing small things, as I do still think this was the best review to date. Just wanted to reitterate that

 

Just wanted to drop in and pay my respects to the [H] team if they're reading this thread. I'm really impressed that they took the time to really dig into the tech and do apples-to-apples testing. (the guys must have balls the size of Texas to go public with that review )

Cheers,

AJ

 

I don't understand this quote from http://www.hardocp.com/article.html?art=NDQ0LDI=

Actually, they are bound (i.e. forced) to use trilinear filtering. If the application (in this case Quake 3) sets the filtering mode to LINEAR_MIPMAP_LINEAR (trilinear filtering) then the driver and hardware should use trilinear filtering. End of story.

 

Yep, we're reading, thanks for the comments

I think Sean did an excellent review here, We all worked to make sure we were really testing apples to apples and came up with our own decisions about the quality and how to compare the cards.

As retail gffx's come in we'll do further iq testing, I know some people have mentioned wanting to see ATI's Performance AF compared to the GFFX's as well. So thats one thing we'll take a look at and try to get all the information out there we can comparing cards.

 

OpenGL guy, I think your assertion stands on its own wrt trilinear... as long we don't use all these nonsensical AF options. Once we use that, I think it's open game to what IHVs decide is "best" for every specific ("all the nonsense "Balanced", "Agressive" etc) instance/option. IIRC HOCP used various AF "options" in determining whether trilinear is in effect, where trilinear, IMO, shouldn't be a rule when you use these various "forced" AF options.

If only trilinear is asked to be used by an app, without any silly "Balanced"/etc AF being "forced", and trilinear is not used, then it is Wrong.

[edit]I'm more troubled by the Splinter Cell anomalies by the GFFX.

 

I'm not surprised at Pelly's review, as he struck me as willing to listen to criticism the last time I offered it to him.

Also, I'm glad I seem to have (finally?) found a fellow believer in Russ about the potential pitfalls of "pretty mip map color" comparisons which I've had issue with for quite a long while. An impression of "Balanced" versus "Performance" in motion would have been helpful indeed, as my recollection seems to indicate Brent said the transitions were readily visible for Balanced (could their blending method for balanced actually make things worse?).

It seems to me that there was too much emphasis on mip map color blending, and no emphasis on texture quality implementation in the same circumstances. I think comparison would have been better served by showing the mip map color blends, and then alternating to showing the texture details with no mip map color overlay...I think the mip map blends were presented as a short cut to evaluating both aliasing in motion and texture detail, and I don't think it is a conclusive evaluation of either. However, I get the impression that Brent is working on an in-depth article addressing this (which should offer analysis of the same type for the Radeon's filtering changes at angles), and perhaps that is why it seems incomplete?

I've already posted how I think the modes compare, and how I think they should be contrasted, and it would be interesting to see a fair minded and thorough presentation regarding a reviewer's take on the issues at some point.

 

i cannot see why you would not want to be able to force trilinar filtering in af; you are not an image quality fan i guess, eh Rev?

 

The quote that I referred to was from where HardOCP was strictly looking at trilinear quality w/o AF.

nvidia's current driver behavior is to default to "balanced" which means they aren't doing true trilinear by default.

 

did [H] get hacked... by nVIdia...
?

the main page is empty, reviews is empty, etc.

 

It appears [H] has gone soft--temporarily, no doubt.

 

Just would not be a proper day in the hardware world without a bit of drama.

 

^^^ Too funny