We are all eagerly awaiting for the launch of the Asus Transformer Prime next week on the 9th, to see what it’s capable of, but until then we can check out these leaked benchmarks using the AnTuTu benchmark tool.
As you can see in the benchmark, the Tegra 3 powered Transformer Prime reaches over 10,000 points, or almost double than current tablets and phones. But as with any synthetic benchmark scores, you have to dig deeper and understand what they really mean.
You have to look at what exactly it’s testing, and see the individual score for that, too, especially with these “overall” benchmark scores. The CPU, which is a quad core 1.4 Ghz one, gets over twice as much the score for CPU integer and floating points. That’s inline with what you would expect from it. Other devices today have either a dual core 1 Ghz or a dual core 1.2 Ghz Cortex A9 CPU, and Tegra 3 has a quad core 1.4 Ghz CPU it seems, so the score should be more than twice as much.
The RAM also seems twice as fast as any other device in the market today, which is more than welcome, and also expected. If the CPU gets twice as fast, you have no choice but to use twice as fast RAM, too, otherwise what’s the point? The CPU wouldn’t be utilized with a slower RAM, if the RAM becomes the bottleneck of the sysem. It’s been reported before that Tegra 3 will work with DDR2-800 (800 Mhz) RAM, while the RAM used in dual core devices today is DDR2-400 or slower.
Where things get a little strange, is at the other benchmarks, which seem about the same with current devices. The reason I say strange is not because Tegra 3’s GPU would be about as powerful as Exynos, because I would expect that in benchmark tests, although I still think Tegra 3’s GPU would perform better in real games.
The 3D scores shouldn’t be the same, and the reason why they are is probably because the benchmark tool for 3D graphics got “maxed out”. Remember the old Neocore tool, where more recent GPU’s could all score 60 FPS or so? That’s because they were maxing out the test. It’s very possible that this happened here, too.
Also, I remember reading Anandtech’s review on iPhone 4S and his benchmarks of the GPU. The GPU scored higher than any other GPU indeed, but it seems that score may be useless in real world testing. That’s because the games would still be bound by slower memory speed. So when he tested the games in real world, he only got about 2-2.3x more FPS. So even if the GPU is 7x faster than the old iPhone 4 GPU, it can still run games only 2x faster, because of the RAM speed bottleneck.
The good news is that Tegra 3 won’t have this problem. As long as the GPU is still fast enough, and I believe it is, the games should run significantly faster or look significantly better because of its DDR2-800 RAM (assuming it uses either the same RAM or the same type of RAM for its GPU). Maybe this is the reason why Nvidia’s glowing ball demo looked much better than anything we’ve ever seen on a mobile device.
Either way, we should learn more about all this starting November 9th, or whenever Tegra 3 starts shipping.