ltcommander.data
Regular
http://www.anandtech.com/show/4064/glbenchmark-20-released-modern-socs-benchmarked
So Anandtech did a comparison between various smartphones using GLBenchmark 2.0. The results aren't too surprising with the SGX540 being the fastest and the Adreno 205 coming second. Qualcomm certainly did a good job going from the Adreno 200 to 205. Some interesting points include that there doesn't seem to be much optimization going from Android 2.1 to 2.3 seeing Hummingbird's results are unchanged, the Adreno 200 is surprisingly competitive in OpenGL ES 2.0 with the SGX530 but falls behind in OpenGL ES 1.1, and the SGX 540 really flexes it's lead over the Adreno 205 in OpenGL ES 1.1. Apple also seems to have a well optimized OpenGL ES 1.1 path, with the iPhone 4 and iPad managing to overtake both Adreno 200 and SGX 530 devices in OpenGL ES 1.1 vs. being last in OpenGL ES 2.0.
The other obvious result is how poorly the iPhone 4 and iPad do due to their high resolution. While these results aren't surprising, I'm unsure how come they aren't reflected in actual games. Infinity Blade which is 3rd gen device and up, so no excuse of the 3rd gen devices sharing code paths with older devices due to the common resolution, looks much better on the iPhone 4 than the iPhone 3GS. The GLBenchmark results should indicate that 3rd gen devices should be capable of additional shader effects and AA over the iPhone 4 and iPad, but that doesn't seem the case in actual games. Are developers just leaving performance/features on the table or are their other factors that GLBenchmark isn't capturing? Obviously the iPhone 4 has the additional advantage of more RAM, but wouldn't that be offset by the higher resolution? And the iPad has no such advantage over the iPhone 3GS.
Seeing Apple's focus on high resolution isn't likely to back off, and things will only get worse depending on what resolution the iPad 2 comes with, the new SGX544 seems pretty tempting if it doubles the TMU and ROP count without having to resort to MP configurations which presumably are more complicated. I guess it's too soon for a possible A5 next year to use the SGX544?
So Anandtech did a comparison between various smartphones using GLBenchmark 2.0. The results aren't too surprising with the SGX540 being the fastest and the Adreno 205 coming second. Qualcomm certainly did a good job going from the Adreno 200 to 205. Some interesting points include that there doesn't seem to be much optimization going from Android 2.1 to 2.3 seeing Hummingbird's results are unchanged, the Adreno 200 is surprisingly competitive in OpenGL ES 2.0 with the SGX530 but falls behind in OpenGL ES 1.1, and the SGX 540 really flexes it's lead over the Adreno 205 in OpenGL ES 1.1. Apple also seems to have a well optimized OpenGL ES 1.1 path, with the iPhone 4 and iPad managing to overtake both Adreno 200 and SGX 530 devices in OpenGL ES 1.1 vs. being last in OpenGL ES 2.0.
The other obvious result is how poorly the iPhone 4 and iPad do due to their high resolution. While these results aren't surprising, I'm unsure how come they aren't reflected in actual games. Infinity Blade which is 3rd gen device and up, so no excuse of the 3rd gen devices sharing code paths with older devices due to the common resolution, looks much better on the iPhone 4 than the iPhone 3GS. The GLBenchmark results should indicate that 3rd gen devices should be capable of additional shader effects and AA over the iPhone 4 and iPad, but that doesn't seem the case in actual games. Are developers just leaving performance/features on the table or are their other factors that GLBenchmark isn't capturing? Obviously the iPhone 4 has the additional advantage of more RAM, but wouldn't that be offset by the higher resolution? And the iPad has no such advantage over the iPhone 3GS.
Seeing Apple's focus on high resolution isn't likely to back off, and things will only get worse depending on what resolution the iPad 2 comes with, the new SGX544 seems pretty tempting if it doubles the TMU and ROP count without having to resort to MP configurations which presumably are more complicated. I guess it's too soon for a possible A5 next year to use the SGX544?
