XGI first test

[bsdgeek]

http://www.tomshardware.com/graphic/20031107/index.html

THG preview. Looks great in some cases, abysmal in others.

[gangster]

Originally posted by bsdgeek
http://www.tomshardware.com/graphic/20031107/index.html

THG preview. Looks great in some cases, abysmal in others.

It's wild seeing that much variation.

[Helevitia]

Hopefully it can be chalked up to early drivers that need to be tweaked.

[gangster]

Originally posted by Helevitia
Hopefully it can be chalked up to early drivers that need to be tweaked.

I agree. I would be more than willing, even look forward to, trying a new card unless the price was astronomical.

[Helevitia]

Originally posted by gangster
I agree. I would be more than willing, even look forward to, trying a new card unless the price was astronomical.

They are saying $300 for there top of the line card. that is for the DUOV8 or whatever it is called.

Dave

[gangster]

Originally posted by Helevitia
They are saying $300 for there top of the line card. that is for the DUOV8 or whatever it is called.

Dave

Surely they realize that with only 2 major card makers now a third could be an incredible success if it's done first class, imo.

[Indiana]

Originally posted by gangster
It's wild seeing that much variation.

My guess would be cheating drivers

The fact that all games are forced to a bilinear filter while ONLY the one most used app to test the filtering is rendered with correct trilinear very strongly hints into this direction: i.e. they made the driver detect the test prog and then use trilinear and force bilinear in all other cases - maybe they thought they'd get away with this.

Considering the bilinear filtering and the BAD texture-quality (now that's much worse than even the worst NVidia-cheat driver ever has gone) renders the benchmark results invalid and doesn't shed a good light on the card: if performance is only at the level with this reduced texture details and as it looks reduced rendering precison, plus a forced bilinear filter, then the card wil most likely only perform at Radeon9600Pro level with full detail.
And then it's simply much too expensive...

A bit "strange" or should I say suspicious is also the fact, that the card does fare the best in the most used benchmarks and delivers abysmal results whenever custom benchmarks are used - sounds a lot like cheating to me...

[gangster]

yup

[knirfie]

Does beta hardware with beta drivers say anything to you?

[Admiral]

You might let them get away with the bilinear filtering though the custom versus preset benchmarks look mighty suspicious as Indiana mentioned.

Let's take UT2003, it does very well in Antalus, better than a 9700 Pro and closely trailing a 9800 XT, yet when you bench it in the custom recorded Magma and Inferno demos, it falls to almost half of what a 9700 Pro can do.

[bsdgeek]

It would have to be some extreme cheating for that much of an increase, it's like the 2nd chip isn't even on in some cases.

[K6-III]

I would wait for final drivers and hardware before throwing around cheating allegations...

[TransformX]

It's not like Tom can't be bought... Maybe they just didn't pay enough to *fix* all the benchmarks

"Hey, I told you those custom benchmarks'll cost you an extra..."

[gangster]

somethings still odd though, the variation seems to be from an optimization or just plain crappy drivers altogether and that seems odd to.