Performance of GPU in Civ V

[Vimes]

Hello

I am reading this comparison between the GTX480 and GTX560Ti in Civ V...

http://www.anandtech.com/bench/Product/309?vs=330

(Blue for 480 and Black for 560)

I have a GTX 460 and with that card against the 480...

http://www.anandtech.com/bench/Product/309?vs=313

(Blue again for the 480)

why does the 560Ti seem to perform quite a bit better than the 480 at 1680x1050, even the 460 that I own seems to do better than the 480..?

I am using Civ V as my example as that is the game that I'm playing a great deal at the moment. I was hoping that if I could justify any change at all in the GPU to try and improve the performance at the late era gameplay then the 480GTX would have offered that over the 560Ti (I play at 1680x1050). But it seems that at that resolution the 560Ti would be a better choice and yet that would have given me the impression of just a small incremental difference of not being worth it.

 

[MooMoo444]

Those benchmarks are quite odd. As the resolution rises, the FPS increases which suggests a CPU bottleneck. Therefore, you'd expect a 560 to score pretty much the same as a 480 at 1680x. Assuming the same CPU and clocks are used for each bench.

I'm baffled.

 

[Vimes]

Confuses me as well, unless it is something to do with the increased texture fill rate for the 560Ti but that would not explain why my 460 shows as performing better than the 480.
I'll pop a post over at Anand's to see if they can shed light on the testing methods etc.

 

[ThomasTheTankEngine]

Civ probably isn't the best game for benchmarking.

The thing that's different compared to most games is how it utilises the cpu. When you hit the end of turn button, cpu usage spikes up to 100% for a few seconds whilst it's doing all the end-turn calculations. As a result, the FPS drops because there isn't enough cpu power available to let the GPU do it's thing (technical, I know ) Perhaps this has a greater effect on more powerful cards.

I guess you'd have to be very careful when running any kind of bench: Making sure you load the same game (a game that's just started in 4000bc is very different from a late-game one where there are units on every tile), and making sure you push the end-turn button the same number of times whilst the bench is running.

 

[Vimes]

Cv V has three benchmark modes built into it. I'm not sure how Anandtech calculated their findings, as it is not mentioned.
At late game era I find that my overclocked E8500 running at 4Ghz is utilised at around 80% (at most) and the RAM used by the game rarely goes above 1.7GB of my 4GB installed. The panning around the screen can at times be problematic, especially on large and huge maps.
My overclocked 1GB 460GTX is being used up to 85% utilisation and all its 1GB of memory is being taken.
I just couldn't understand from those graphs how well the 560Ti performs against that of the 480GTX.

 

[man_from_uncle]

Those are awful benchmarks and I really don't think they can be trusted. The GTX480 shows increased performance at higher res, not just reletive to the GTX560Ti, but reletive to itself at lower resolutions. That's just not very likely at all.

And the scales are so far off too, makes me think whomever created those had quite an agenda.