Originally posted by: Rollo
"If nVidia has stopped smoking there crack and there act together"
Er, yeah. they're smoking crack, and having the largest video market share, the largest console contract, the best AMD motherboard chipset, and the highest profitability. Of course, none of this compares to having a card that appeals to about 2% of the market and is sometimes faster than your competitor.
Sure, it's sometimes faster than your competitor. In fact, in very specific situations (in UT2003, essentially only two of the six tested setups).
Let's face the facts shall we:
Based on Anand's benchmarks, we know :
1. The GeForce FX5800 Ultra offers performance relative to the R9800 Pro of (UT2003 without AA/AF):
at 1024x768x32: 95.6%
at 1280x960x32: 103.1%
at 1600x1200x32: 109.7%
2. The GeForce FX5800 Ultra offers performance relative to the R9800 Pro of (UT2003 with AA/AF):
at 1024x768x32: 67.7%
at 1280x960x32: 66.8%
at 1600x1200x32: 59.7%
Now, the scores without AA/AF are such that the Radeon 9800 Pro still scores 120FPS at 1600x1200x32, a more than playable speed.
The scores with AA/AF are such that at 1600x1200x32, nobody will consider the cards to be very playable, but at 1280x960x32, the R9800 Pro's 110FPS score means that it is now pheasible to play at 1280x960 with The Radeon's Performance 8xAF + 4xAA.
In case you don't quite get it, the Radeon 9800 Pro makes it very possible to now play UT2003 fully AA/AF enabled at REASONABLE SPEEDS (Key words). The GeForce FX5800 Ultra can't.
So it's a tradeoff: get an at most 9% lead in non-AA/AF setups at 1600x1280x32, or a 33.2% lead in AA/AF at playable speeds at 1280x960x32?
If I were upgrading now, I know what my answer would be.
Granted, ATI has problems with the R9700 Pro and wavy lines, but I haven't seen anyone confirm whether this is a problem with the R9800 pro as well.
Hopefully nVidia can keep things interesting by releasing another one of their magic driver updates to increase performance by a sudden 25% (which has always made me wonder why the drivers are written so poorly at release as to cause a 25% performance loss on the cards).