backtop


Print 32 comment(s) - last by jurassic512.. on Mar 22 at 7:38 PM


  (Source: iFixit)
Benchmarks show Apple's chip is generally the more fit competitor, however the Tegra 3 still manages to impress

Apple, Inc.'s (AAPL) new iPad 3 processor, the A5X, has been pitted against a top-of-the-line Android chip from NVIDIA Corp. (NVDA), the Tegra 3 in the real world.  The Tegra 3 is found in the Transformer Prime, the flagship Android 4.0 Ice Cream Sandwich (ICS) tablet from ASUSTEK Computer, Inc. (TPE:2357).

Early reviews and analysis commentaries from various parties were rather light on hard performance numbers for the CPU.  But the folks over at Laptop Magazine have dropped a full set of CPU and GPU benchmarks on the iPad 3 and evaluated how it performed versus the Tegra 3.

In GLBenchmark 2.1, the A5X is a beast.  Apple's tablet system-on-a-chip (SoC) is thought to pack a PowerVR SGX543MP4+ (the same chip found in the Sony Corp.'s (TYO:6758PlayStation Vita).  Apple wasn't kidding when it said the chip was four times as fast as the Tegra 3.  In a huge win for Imagination Technologies plc. (LON:IMG), the GPU is actually close to 5 times as fast as Tegra 3's GeForce derived mobile GPU.

Tegra3 v. A5X GLBenchmark
[Image Source: Laptop Magazine]

Ironically, while graphics veteran NVIDIA languishes in graphics, it hits back equally hard in the CPU department.  The Tegra 3 outperformed the A5X more than 2-fold in the GeekBench benchmark, which examines various CPU performance metrics.  The only victory for Apple came in the stream test.

Tegra3 v. A5X GeekBench
[Image Source: Laptop Magazine]

Overall, these results are not surprising.  Both the A5X and Tegra 3 use the licensed Cortex-A9 design from ARM Holdings Plc. (LON:ARM).  And the A5X has two cores, while the Tegra 3 is a quad-core design.  Now, the comparison isn't exactly apples and apples, as the Tegra 3 in the Transformer Prime is clocked at 1.3 GHz, where as the iPad's chip is clocked at 1 GHz.  Thus the A5X may be actually a bit more efficient on a per-core basis, albeit weaker overall in raw power.

Unfortunately for NVIDIA, those results don't necessarily translate into real-world performance.  The stock ICS browser ran the Sunspider Javascript test suite slower than the stock iPad 3 Safari variant.  Ultimately, this reminds us of Windows Phone, which runs silky-smooth on a single core.  The lesson here is that software is equally important to the raw computing power.  ICS appears to be either less efficient or insufficiently threaded.  Count that as another win for the A5X.

A handful of games such as Shadowgun and RipTide are using the Tegra 3's extra computing muscle to pull off slick effects like physics and reflections.  While the iPad 3 lacks these niceties, it does offer much crisper rendering, thanks to its higher resolution 2048x1536 pixel screen.  That's a lot of pixels, hence the need for that beefy GPU.

Shadowgun wide
The iPad 3 (right) reportedly offers a crisper image, but the Tegra 3 powered Transformer Prime (left) offers better lighting and fix-driven objects like the flags hanging overhead.
[Image Source: Laptop Magazine]

In terms of gaming the iPad 3 and Transformer Prime seem to be neck and neck -- the iPad 3 offers gaudier resolutions, but the Transformer Prime offers unique visual effects in certain apps.

ASUSTEK is also hoping to level the playing field a bit with its coming Transformer Prime "Infinity" (TF701), which sports a 1920x1200 pixel resolution, up from the current model's 1280x800.  The real question is what GPU will be inside that.  After all, the Tegra 3 doesn't seem like it has the muscle to power that high a resolution display.  One has to wonder whether ASUSTEK won't jump ship to a new PowerVR equipped SoC to quench its mobile graphics thirst.

Source: Laptop Magazine



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: Not a horrible GPU exaggeration after all
By Solandri on 3/19/2012 4:28:34 PM , Rating: 4
It's a fill test. Probably has more to do with graphics memory bus speed than the actual GPU. As with all benchmarks, it's important that all these components tests be considered as part of a whole, not piecemeal. The whole is what's important to the end-user.

The component benchmarks are more useful for helping the hardware designers decide which parts to improve in the next gen product. It's misleading and I would say irresponsible for marketing to use them to sell to end-users.

It'd be like Toyota advertising "The Prius uses no gas!*"

Strictly true, but incredibly deceptive the way it's presented.

* When stopped and idle for more than 10 seconds.
quote:
At the end of the day though software makes the hardware irrelevant. iOS and WP7 just operate smoother than Android does. Android is a warm knife through butter, iOS/WP7 is a hot knife.

I think this is Android's Linux roots showing through. The Linux (and Unix) community has never really cared much for user interfaces. That's why CLIs (character line interfaces) are still very strong in that community. That's why they scoffed at the Mac and Windows when they first came out. Even X-windows is just a regular program.

iOS prioritizes user interface interactions and animations. That's what keeps it silky smooth. But the tradeoff is that if there's something CPU-intensive going on in the background, the UI can starve it (e.g. your music playback can skip). Android does not prioritize the UI, hence its multitasking is better, but the interface doesn't appear as smooth to the end-user.

I'm with Apple on this. In old days with single-core processors, equality of priority made sense. But now that dual cores are standard and quad cores are showing up, there's no reason not to prioritize the UI to keep it smooth and responsive. At a very minimum, the UI should be able to do as much as it likes whenever it likes on one core. Let the other apps and processes fight over the remaining core(s), as well as use unused cycles on the UI's core. But give the UI absolute priority over one core.


By MGSsancho on 3/19/2012 9:45:01 PM , Rating: 1
Unix and Linux have no GUI real because of tits application. A computer with no display does not need a fancy user interface. Hand held devices and tables which are nothing but display need great user interfaces. I think you are forgetting that you have to think about your devices application. Cores and flops are cool but only matter if the software can utilize it. iPhone has a slower CPU then rest of its competitors yet the software running on it does just fine. As far as roots are concerned, the *nix operating systems are very modular and cheap to obtain and license.

Stop thinking about the hardware too much and focus on the total experience, Currently Apple has things figured out well from purchase show room experience, buying, opening packaging, first boot, ergonomics, all the useability stuff, customer support, ponies etc. Others do well is most respects. What apple does better than no one else is ship few products with great components. Notice I did not say best.

We on this site need to stop harping on pissing wars and try to device at a given price point, what is the best deal. Maybe Anand or the anandtech employees can use tables like for the GPU stuff and try to do comparisons on price brackets. with many tablets being around $400 maybe there is not much of a table but for smart phones, laptops, netbooks, ultrabooks, desktops and workstations maybe we can.


RE: Not a horrible GPU exaggeration after all
By sigmatau on 3/19/2012 9:49:06 PM , Rating: 2
So you like skipping music?


By testerguy on 3/20/2012 3:59:46 AM , Rating: 1
I'm sorry but a GL fill test is not comparable to Toyota saying 'the Prius uses no gas!'. It's a legitimate measure of the power and speed of a GPU when in use and needs no such disclaimers such as 'when the GPU is not being used'. It's not misleading any more than any of the CPU benchmarks are - which as the article states weren't all won by the Tegra.

I agree with you in general that how a device is put together overall is the most relevant metric but there are no easy benchmarks for that, taking FPS, for example, is not a proper benchmark because it doesn't take into account the resolution of that FPS. There's no benchmark to demonstrate that even with a slower CPU the iPad 'feels' faster when in use, either. To try and stay objective, review sites benchmark each individual component, and then offer their analysis on what the overall effect feels like.


"Well, there may be a reason why they call them 'Mac' trucks! Windows machines will not be trucks." -- Microsoft CEO Steve Ballmer














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki