Are iPad’s A5x graphics really 4x faster than NVIDIA Tegra 3?

When Apple unveiled its the third generation iPad earlier this month, the company made the bold claim that the new tablet’s new Apple A5x processor offers 4 times the graphics performance available from NVIDIA’s Tegra 3 processor, which is one of the fastest chips available for Android phones and tablets.

iPad 3 v. Transformer Prime

NVIDIA expressed some skepticism at the claim, but only the folks at Apple know how they tested the graphics performance.

But now that the new iPad is available to the public, anyone can run their own tests. And that’s just what Australian blog Ritchie’s Room has done.

As it turns out, the iPad 3 does come out way ahead of the Asus Eee Pad Transformer Prime in the GLBenchmark graphics test. The Prime has NVIDA’s latest chip while the new iPad has Apple’s.

And while the iPad didn’t quite offer 4 times the performance of the Transformer Prime, it kind of trounced the Asus tablet with scores ranging from 2.3 to 3.5 times as high.

Update: Gizmodo ran similar tests, as did Laptop Magazine. with different results. The iPad still came out ahead, but not by nearly as great a margin.

But that’s just graphics performance. Tests that look at other factors aren’t as clear-cut.

Ritchie also ran web-based benchmarks to look at HTML5 and JavaScript rendering, and the scores were much closer, with the iPad winning one test and losing the other.

Benchmarks like these aren’t always indicative or real-world performance, and there are many differences in the iOS and Android operating systems that can affect real and perceived performance. But in a few tests where both tablets are competing on the same playing field, it looks like the latest iPad is more than capable of holding its own against the competition.

Long story short, the NVIDIA Tegra 3 may have four primary CPU cores and 12 graphics cores while the Apple A5x has 2 CPU cores and 4 graphics cores, but you can’t judge a chip by its core-count.

  • allanmac

    Until GLBenchmark is updated to render at the iPad3’s 2048×1536 this is a flawed comparison.  Rendering at the old 1024×768 and scaling up will generate exactly what you would expect given the new faster A5X CPU and somewhat beefier GPU.

    • http://profiles.yahoo.com/u/6IARER67M77XYXJTBCOMQ2OCEY X-Man

      But on the other hand the GLBenchmark is said to be very well optimized to run on iOS system which isn’t the case for ICS and Tegra3.

  • Cobey

    Odd, I don’t see a link on 
    Ritchie’s Room for said test results, but I do see that Giz did the very same tests 
    ( http://gizmodo.com/5893970/ipad-test-notes-speed-versus-tegra-3 ) and called it closer to a draw over all.

    • CyberGusa

      System components like the GPU vary in what they are good at and what they’re not so good at depending on how they are designed and often it’s what they’re not good at that is often marketed as the difference to exaggerate the differences of any one solution over another.

      So the Giz account is probably more accurate for overall performance.  Benchmarks done for the iPad 2 vs the Tegra 3 by Anandtech and others for example indicated that the Tegra 3 is significantly weaker in terms of texture rendering but in other things like shaders they were much closer.

      Meaning Apple’s claim is likely based on only the benchmarks that go heavily in their favor but ignore the others that don’t.

      • Rstisme

        Its easy too see u guys know a bunch of technical jargon but it it also easy to see which side of the fence you stand on. I never listen to those who don’t stand on a neutral sides and too tell you the truth there isn’t too many neutrals out there. So I say check it out for yourselves! You’ll defiantly see a difference.

      • CyberGusa

         The so called technical jargon is what separates facts from arbitrary opinions.

        Those who ignore facts are the ones who aren’t neutral and just pushing what they want to believe.

        Though I agree on having everyone check it out for themselves.  Just don’t confuse opinions with facts!

        Products are what they are, nothing more and nothing less.

      • Mr. T

        CG, if I could “Like” your comment 10^9 times, I would.

  • http://etiforum.com/ Mike Walker

    Given equal speed hardware, Apple will always have somewhat of an advantage in that they control and manufacture all the hardware their software runs on, and do not have to content with the multitude of platforms Android has to run on. Thus tuning the software (and especially the benchmarks) is much easier for Apple.

    • Ziomarco

      Apple does not have their own manufacturing. Foxconn makes the products for Apple.

      • Bervick

        He meant the design. The actual mfgr’ing is immaterial.

      • CyberGusa

         Not entirely, the capabilities of the manufacturers Apple chooses in many ways also determine the design limits and built quality of their devices.  Along with what technology they can license in addition to their own patents.

        For example, neither the GPU or the LCD of the new iPad was designed by Apple.

      • Bervick

        That’s not correct; at least as far as the GPU goes. Apple does customize their design based on an architecture. Then they simply hand out their specs to a mfgr with mfgr’ing capabilities to build it.

        http://en.m.wikipedia.org/wiki/Apple_A5 

        But yes, it’s not like they built the GPU entirely by themselves. I was not getting at that.

      • CyberGusa

         Customizing a SoC, or any configuration, is not the same as designing the technology used in the SoC.

        Apple didn’t design the GPU, Imagination did and Apple just licensed it from them and integrated what they wanted into their SoC. 

        Even going dual and quad core are because those are options that Imagination provides for the specific PowerVR GPU series version they licensed from Imagination.

        What Apple really designs is how they choose to configure their SoC but they don’t design all the technology or components used in it.

  • Someone

    Umm slower CPU parts, faster GPU. The A5X is four Imation 543s bolted together so that each one renders a 1024×768 quadrant of the screen. The only Tegra3 im aware of on the market is pushing a 1280×800 screen. Even throwing out the OS issues, there is really no way to compair the two rigs. Maybe by measuring texels or gflops, but you’re going to see a lot of skewing based on the differences n architecture. In the end I would be shocked if nVidia doesn’t have the faster single chip solution, that said ANYONE can technically do what Apple has done since this their chip is made with ‘off the shelf’ parts. So what is impressive is how little ‘special sauce’ nVidia seems to be bringing to the table. At least to me.

  • Sonone

    Something tells me that this new Ipad is gobbling at least twice as much juice as Tegra 3.  Is there any way to measure this?

    This is certainly not an apples-to-apples comparison if one chipset is drawing twice as much power!

    In any case, I’m happy that this new iPad trounces Tegra 3.  Competition is always beneficial to the consumer!  And Apple has raised the bar again!

    • CyberGusa

      Safe bet considering they needed to increase the battery capacity from 25WHr for the iPad 2 to 42WHr to keep about the same run time for this new iPad.

      However, it also means it takes a lot longer to charge, and according to netbooknews that’s up to 7 hours when off and up to 18 hours while in use.  While continuous movie playing only lasts up to just over 8 hours.

      Also, according to the iFixit tear down, the battery capacity density is the same.  So they did manage to reduce the weight of the rest of the system for the larger battery and lay that out thin enough to prevent too significant increase in system thickness.

      However, this basically means most of the iPad is a battery and hopefully the system mass reduction doesn’t mean any reduction in durability as the last thing anyone wants is the lithium battery, especially one that size, getting ruptured…

  • Mr. T

    “but you can’t judge a chip by its core-count.”

    Of course you can’t. Clock speed still matters, and even more so the nature of the problem you’re looking at. If the problem is inherently serial, a billion cores won’t do you any better than a single core.