GTX260/GTX280 Performance

[maroon1]

CJ sent us a performance chart comparing Radeon HD 3870 X2 against the GeForce GTX 280 and GTX 260 on some of the games out there. Although, the benchmarks are done internally by Nvidia but it is probably conclusive to compare GTX 280 and GTX 260 here. The drivers for GeForce GTX 200 cards is the upcoming GeForce Release 177 while the drivers for 3870 X2 is a rather old Catalyst 8.3.

VR-Zone : Technology Beats - GeForce GTX 280 & GTX 260 Gaming Performance

So, based on those benchmarks GT200 is around 1.5 to 2.5 times more powerful than HD3870 X2. I hope that those benchmarks done by Nvidia are true.

Note: All the tests show the 3870x2 at 1.0, because it's showing a % relation to the nvidia cards. The bars for GTX 260 and 280 are several times bigger than because the bars start at 0.8

 

[MrCoffee]

Impressive, most impressive.

 

[Merkwürdigeliebe]

I really hate those graphs, the bars are visually double or eight times greater than the numbers on the Y-axis, they do not reflect the same ratio... they cut off the bottom of the graph... how misleading! The y-axis doesn't start at 0.0 it starts at 0.8 ... that's some shody marketing right there...

Look at the Crysis benchmark.. the GTX 280's bar is more than 5x the size of the 3870x2's bar... yet the real difference shown on the Y-axis is only an 80% increase - that's still impressive, but the graph's incredibly misleading...

Not as impressive now... and we can only wonder how unfair these tests were

 

[Trotter]

It's all in the name of spin. Even if the actual difference is miniscule, if you stretch the chart out enough it makes it look like so much, much more.

All of them do it, unfortunately. It is up to us to actually dig for the real numbers behind the visuals.

 

[veedubfreak]

I'd say an 80% improvement in crysis is great reguardless of how the graph is shown.

 

[Trotter]

That all depends on what the numbers represent. 1.0 vs 1.5 vs 1.85 is very vague... which is great when you are trying to sell something. What exactly does the "1.0" stand for? Or the 1.5? Or 1.85? We are not told. the same goes for the rest of the graph. We don't even know that the increments are supposed to be.

I see this stuff all the time at work. You skew the results by exaggerating the differences. It is used to justify changes and projects all the time. It's like that whole "For the price of a cup of coffee a day..." things.

 

[veedubfreak]

I would assume(yes i know what it makes me) that 1 is the control and the others are how they compare to the control. So 1.8x the value of the control would be 80% increase.

 

[Trotter]

Exactly. I would normally assume the same thing, but that is not always the case. Without seeing the actually results, we are taking someone else's abstract chart and being left to figure out what it actually means.

Were I the one trying to make the sell, i would do the same thing.

 

[Apokalipse]

I don't really trust benchmarks that are this vague, and without actual details of test conditions.
We know for example that R600/RV670 doesn't do as well with AA, so they could easily have just performed the tests with AA up high or something

Nvidia's marketing team have been less than honest before.

 

[Ste]

We should ignore everything until true benchmarks/reviews are released.

That way, we can focus all our attention on saving.

And remember, the GTX 280 needs to be at least 20% better than the 4870 X2 to warrant purchasing. (Dependant upon which benchmark or game you care most about, of course. )