Why the New iPad 3 May Be Slower than iPad 2. Wait, What?
“Quad core tablet”, “four times the graphics performance of Tegra 3”, “the second coming of Jesus Christ – again”, and of course all the usual “amazing, spectacular, magical and all the other attributes we’ve heard before about any new Apple device, appeared once again in the Apple media a couple of days ago. Except many of those are not exactly true – well perhaps the “magical” one is – I mean who can argue with that?
We don’t have any actual benchmarks done by 3rd parties yet, but from what I can tell, the iPad 3 is first of all not a “quad core tablet”. Every single human that has ever bought a computing device knows that dual core or quad core refers to the CPU, not the GPU – otherwise we’d be calling the Asus Transformer a 12-core tablet, and our PC’s – “500 core PC’s”, since that’s how many cores a high-end GPU has on PC today. So let’s end this shameless misleading by “some” Apple media outlets, here and now: the iPad 3 is a “dual core tablet”, not a quad core.
Second, the “4x the graphics performance of Tegra 3” is extremely suspicious – to say the least. Nvidia believes so, too. But we didn’t really need Nvidia to tell us that. We can think for ourselves, too. It’s true that the A5 GPU tested in benchmarks to be about 30% faster than Tegra 3, but those are only some rather simple tests, and they don’t test the actual performance of a chip in a game.
Plus, a game is also bound to memory bandwidth and CPU performance as well, so for all we know, a game could actually run better on a Tegra 3. And so far I haven’t seen anything to make me decide that the iPad 2 graphics were absolutely better than Tegra 3 graphics. Anandtech also seems to believe that some games may actually run worse on iPad 3 compared to iPad 2 at native resolution.
But now we have the A5X GPU with 4 instead of 2 GPU cores, so clearly the iPad 3 graphics performance must be faster than iPad 2 graphics – right? Well, not so fast. While “theoretically” the iPad 3 GPU has double the performance of the iPad 2 GPU – in reality the tablet itself likely won’t have that kind of performance increase.
Why? Because the new GPU has to push 4x as many pixels on the screen. While this won’t affect regular use of the tablet too much because I’m sure the new GPU can handle normal tasks and apps well, otherwise they wouldn’t have released it, I’m also quite sure that the graphics performance in games will take a very serious hit when the games use the native 2048×1536 resolution and not the 1024×768 one and by upscaling the game.
What’s my basis for this besides normal common sense believing that a 2x increase in GPU performance won’t be enough to beat the older GPU when the number of pixels increases 4 times? Well, remember when the iPhone 4 also got the Retina Display? It actually used a more powerful overclocked PowerVR SGX535 than the one found in the iPhone 3GS, and yet in this test the iPhone 3GS is still faster than iPhone 4 in graphics performance.
I expect something similar to happen to the iPad 3, but even if it doesn’t – in worst case scenario (if I’m wrong) the iPad 3 will still be only just as fast, or slightly faster than iPad 2 in real world graphics performance, and will definitely not have the performance increase they are touting (4x Tegra 3, etc). But I don’t think I’m wrong, and iPad 3 should actually have slower performance in advanced 3D games than iPad 2 at native resolution.
Also, that 1 GB of RAM? Will not be enough to make up for the increase in resolution. A more appropriate and proportionate increase would’ve been 1.5 GB of RAM, and if you thought the 512 MB of RAM wasn’t really enough for it, then you’ll start feeling the same with the iPad 3 once apps start supporting the new resolution.
Same goes for the CPU, which apparently either got no improvement in performance, or it was only a slight increase so they would rather not mention it. The resolution is not generally bound by the CPU, but at some point the CPU does interact with data that uses the new resolution, so if it’s the same CPU as in A5, then it should experience some performance bottlenecks when dealing with the high-res apps.
My advice for those who are still in the Apple camp, and want an iPad – if you already have an iPad 2, then skip iPad 3 and wait for iPad 4, which might actually have a performance increase over the iPad 2. That’s without even counting the increase in thickness and weight over the iPad 2, which will probably disappoint many who want to switch from iPad 2 to iPad 3.