Originally posted by: HurleyBird
I made no such claim. The claim was that a few significant gains plus smaller gains across the board put the 4800X2 on equal footing with the GTX 295. Look at computebase.de again, on the page where they show the average performance between the different cards. If you compare cat 8.12 with 9.1 beta, add up all the differences and divide to see the average, you'll see that 9.1 outperforms 8.12 by 7.125% -- that's, well, significant overall. Hell, if you do the math across all of the resolutions and AA/Aniso levels they tested the GTX 295 beats the 4870X2 by a whooping 0.375% on average, and if you throw away the dubious 1280x1024 results the 4870 takes the lead overall. See? I've Objectively shown that the computerbase.de numbers put the two cards in a dead heat. This facet of our conversation should be over unless I made a mistake in my calculations.
You keep pointing to the performance rating aggregates but still don't understand those numbers mean very little when the individual benchmarks tell a very different story. What you'll find is a green bar followed by 2 red bars in the majority of the benchmarks showing the 295 is faster than the 4870X2, and that the 9.1 drivers aren't enough to change that.
The reason the performance rating numbers are more or less meaningless as an aggregate is just as you stated, you'll have outliers that skew results but at the same time, mean very little by themselves For example, STALKER Clear Sky at 2560 with 4xAA. 4870X2 scores 8FPS and GTX 295 scores 2.8FPS. The 4870X2 is over 200% faster! Similarly, you'll have games that perform much better with 1 vendor, like RS:Vegas2 (or Dead Space for NV, which wasn't tested). While that doesn't invalidate the findings that the 4870X2 is faster in that title, using those results in an aggregate clearly inflates any % using those numbers.
Also, how do you think they factored in some problems with the 4870X2, like Bioshock with 4xAA where the 4870X2 registered 0s. Looking over the results I'm not sure how they compiled those performance rating %s as the results just don't seem to add up.
I'm not going to bother adding up all the benches to see who won what, but what's your point? How many benches one company wins isn't as good a metric as average performance. Sure, when you only have a few titles and one card fails/dominates (behaves as an outlier) on one of them the results can get heavily skewed, but when you test a large amount of titles like computerbase did, the outliers tend to be less of an issue.
The point is the 295 is faster in more settings/titles by a quite a bit. I browsed through page by page but I'm not going to compile a list because its obvious. Sure the 4870X2 is faster in some games and settings but if the 295 is 10% faster in 9 titles, the 4870X2 being faster in one title by 100% doesn't put them on equal footing. Go through game by game and you'll clearly see, the GTX 295 is the faster part in more games, resolutions, and AA settings.
No, it takes a fanboy to warp a conclusion into something he wants it to be. If Kyle thought that the GTX 295 took the performance crown, he would have came out and said so outright. Saying that the extra money wasn't worth the performance after a couple of paragraphs explaining how the results were too close to call is probably his way of trying to not sound too harsh about the 295. Even giving you the benefit of the doubt, say he thought it was a sliver faster.... in that case both are still neck and neck and still around the same performance. You're grasping at straws here.
Kyle didn't write the article and HardOCP is intentionally vague with their benchmark style of comparing playable settings. The 295 might only be slightly faster than the 4870X2, but it still is faster and HardOCP's conclusion does make that distinction. While they're not explicit about the 295 being faster, they're certainly not anymore explicit about the 4870X2 being on par with it, they just say they both provide the same "gameplay experience", which equates to +/-10 FPS ~40FPS according to their playable benchmarks.