A5 beats Tegra 2 in benchmark tests, thanks to larger size

These test are clearly not optimized for each of the devices. If you add the iPhone 3gs to the line up and compare it with the iPad2 and the Motorola Xoom, you can clearly see this. Just examining the first 8 stats makes the iPhone 3GS look comparable to the Xoom. This clearly gives a false impression.

The best metric I see for cpu performance listed on that chart are the CPU Floating Point performance test. Regardless of which devices you choose, the comparison makes more sense and scale across all platforms. FP calculations are generally straight math computations leaving very little in code that needs to be optimized for each platform.

So TUAW realizes that part of having benchmarks is knowing how to read them right? First of all, this is a GL benchmark, so it's testing how fast the GPU and drivers are, not how fast the CPU is. Both the Tegra and A5 processors have both a CPU and a GPU in them.

The tests that aren't based on FPS aren't always so black and white. The Tegra CPU is waaaay faster. Comparing the Xoom scores considering that it has a lot more pixels to draw, it's not really that far behind either. So the iPad 2 will have much prettier games because it has two GPUs, but don't pretend like this one benchmark says everything about it's performance.

I don't have any problem believing that the SGX543 in the A5 is much faster than the Tegra 2 CPU, the specifications for that thing are pretty impressive, and Imaginations simply has much more experience in building mobile GPU's than NVidia. Their design is almost definitely many times more advanced than NVidia's. But looking at some of the other GLBenchmark I'm seriously doubting the usefulness of this benchmark. For example the iPad 1 and Galaxy Tab have better CPU performance than the iPad 2, both in integer and floating point? And Tegra 2 scores almost 3x better than the A5 on the floating point test, even though it only has VFP and no NEON?

To me that indicates a very badly optimized benchmark that doesn't say much about real-world performance differences.

Umm no. The important quote is:

"The major difference in specification between the chips is that Apple uses Imagination Technologies‘ PowerVR SGX543 dual-core GPU in the A5 and Nvidia uses its own GPU called GeForce"

The article is say the better performance is because of the GPU.

The A5 is bigger because the manufacturing is 45nm and it accommodates the PowerVR chip (which is better and happens to be bigger).

All things being equal the physical size of the chip doesn't matter. It is the number of transistors on a chip. (If you can pack the same number of transistors into a smaller package i.e. using a 40nm that is usually even better)

I've never quite understood everyone saying "A5 is better than Tegra 2 because it's bigger," then going on to talk about how the A5 is bigger because it uses 45nm manufacturing instead of 40nm. Am I totally misunderstanding things? I always see reports with standard CPUs talking about how this one is so much better because the old one was x nm but the new one is x-10 nm.

So why is it in other cases the physically smaller chip is better, but here the physically larger one is?

Have in mind that the density of the TSMC process nVIDIA is using is better than what Samsung Foundries manufacturing the Apple A5 provides.

Of course that difference doesn't make up for the fact that the A5 has more transistors than Tegra 2 to begin with.