We all know how Apple said that the A5X will cream Tegra 3 in graphics. My question is whether or not you seriously believe them. On another note, which one do you think will be faster in general?
Anandtech.com did a comparison between the Tegra (or GeForce?) GPU and the current A5 dual-core GPU found in the iPad 2.
There is a story and article over there for you to read. Very interesting. However, Anandtech's numbers and benchmarks (even real-world testing) was inline with Apple's claims, months before today's presentation.
Nvidia was a bit crap with their Tegra 2 chip, so I don't really believe the Tegra 3 is going to be anything great, other than being one of the first to the market of quad-cores.
The awesome Apple-alternative will be Snapdragon S4 Pro with Adreno 320 which comes out this summer.
From a very technical standpoint - it's also true that the PowerVR SGX543MP2 found in the iPad2 clock for clock seems to best the Tegra3 and Tegra3 came out 3/4 of a year after the iPad2. When more crazy layers, graphics/effects are involved, Tegra3 can handle more features at one time than the SGX543MP2 can due to having more cores. The raw power of the SGX543MP2 is obviously stronger in simpler tests, being limited by 2 cores.
You have this kind of the wrong way round. The tegra may actually be more powerful in simple tests (like drawing a single 'layer' with a complex effect), but in complex scenes (say a game like infinity blade) the PowerVR has a huge advantage. It's because of the architecture difference, and has nothing to do with the number of cores. In fact doesn't the tegra 3 just have a single powerful GPU core?
The difference is that the powerVR is "tiled". When the tegra renders a scene, it has to draw it in one go, all of the polygons (there might be 100,000 of them) at once. When the PowerVR renders it, it splits it into lots of very small tiles, and does one tile at a time. Because the tiles are small, a tile might only have 10 polygons in it. That's nothing, so you can afford to figure out which ones are in front of others, and only render the visible bits.
There might be 3x more surfaces that are hidden behind things than there are visible surfaces, so if it does this it only has to draw 1/4 of the amount the tegra does. That means it might be slower than the tegra, but it will seriously outperform it because it does far less work. It also makes it much more efficient, and therefore better for battery life.
I call this BS. They are naming the new ipad's gpu while no one knows what gpu it has.Analysis of the new Apple iPad.
Anand has the answers.