Its nice to see a review of delayed product in question, thought I can't help but notice how the review compares it with an 8800 Ultra, even though you clearly mention that this product is direct competition for a GTS?

Perhaps a GTS would have been a better card to benchmark it against?

Actually, it'd be better to describe the 360's graphics card as "DX 9.5", not DX10. And there's a fair amount of its architecture which wouldn't translate to a PC part.

We didn't have an 8800 GTS to hand, although we tested the Sparkle late last year, however we did have an Ultra, and considering the pricing we felt it made perfect sense to compare an Ultra with HD 2900XT in CrossFire.

Let's face it, AMD would have dearly liked the HD 2900XT to beat Ultra. The fact that it falls a long way sort is a bit of a mystery.

...but what about in-game/in-application/movie playback (etc.) graphic quality itself?

There is no comparison and that's quite disappointing, because I represent those, who prefer graphic quality and the hyped numbers are second or third on the list. And ATI cards have always been better in that since Radeon 7XXX and that's not just my opinion, I believe.

I hope this is only a quick review and you're going to return to that subject later, this article doesn't contain all the information I find important.

Regards from your regular reader,

Frantisek Janak

Frantisek makes a good point about the R600 range but I'm not sure how much it bears on the 2900XT. I take the view that gamers might spend £249 on a 2900XT but the man in the street has no need to spend anything like as much to get decent movie playback. And they would want a quiet graphics card while they were about it, which sounds like a cue for a 2400 or 2600.

Of course image quality matters in games and both the 2900 XT and the 8800 Ultra look superb in Oblivion and Half-Life 2 but the real test will come when we can get our hands on DX10 games.

Fair enough, I respect your view and find it meaningful, thank you very much for reply.

I have seen some image quality comparisons lately and the usage of CFAA with Wide Tent in Oblivion with 2900XT seems to be useful a lot for overall quality, I'm looking forward to your further testing.

Regards,

Frantisek

Two things spring to mind:

- If you used Crossfire for the ATI, surely you should have included an SLi config for the nvidia setup, for a more balanced report?

- Is it just me or does Crossfire's performance benefits look feeble (on a %age increase in frame-rate basis) compared to typical performance gains from Sli?

Frantisek, HardOCP has done a bit of a demolition job on the 2900XT partly with regard to the different AA filters including the different tents.

http://enthusiast.hardocp.com/article.html?art=MTM0MSw1LCxoZW50aHVzaWFzdA==

They have included a number of screengrabs to illustrate the quality of AMD versus Nvidia and frankly they are pretty much identical in most respects however when it comes to AA the AMD card images look blurred. I have no intention of taking issue with HardOCP - in the main they do excellent work - however I have been in this situation way back in the days of Radeon 9800.

This was a time when AA was quite novel and I simply wanted to illustrate what it could do so. I used Microsoft Train Simulator as an example because the train tracks suffer from horrible jaggies as they converge towards the horizon. Enabling AA smoothed the image considerably so I set about taking screen grabs to illustrate the success.

In the images the rails looked good with AA but the trees that line the track looked like green lollipops with no distinguishing features and the grabs with AA therefore looked terrible. Much like the images on HardOCP..

Of course I was looking at stills from a moving image and was effectively missing the point. As you drive past a tree you can't see the leaves and branches. Stand under a tree in Oblivion and you get the benefit of all sorts of eye candy.

Ever since I have been very wary of taking screen grabs as they often show visual information that is out of context.

As for the most recent comment, yes it would have been 'fair' to run a pair of Ultras in SLI however I only had one Ultra - does anyone have two? - and simply wanted to illustrate how £500 of AMD hardware compared with £500 of Nvidia hardware.