The GPU: PowerVR SGX 543MP2

There are no spoilers here but the PowerVR SGX 535 in the original A4 is out, replaced by the new PowerVR SGX 543MP2. The 543 as a building block has a bit over twice the peak floating point throughput of the SGX 535. The MP2 just means there are two of these 543s working in tandem. The result is what Apple uses as the basis for their "9x faster" GPU claims. In practice the improvement should be less than that, but it's still enormous.

We've got GPU performance data coming, but I thought I'd take a slightly different route and show what one developer is using the extra horsepower offered by the A5 for.

Epic Games is a familiar face around these parts. We've used games powered by their Unreal Engine in our reviews for years now. More recently Epic has made a splash in the iOS world with the release of its Citadel demo and eventually Infinity Blade.

In time for the launch Epic updated Infinity Blade to have some special enhancements for iPad 2 owners. Rather than simply deliver a higher framerate for iPad 2 users, Epic enabled higher resolution textures and anti-aliasing. The resulting boost in image quality is astounding, particularly on the iPad 2's 1024 x 768 screen:


Mouse over to see Infinity Blade on the iPad 2

There's far more detail in the character models as well as the environment. Lighting looks improved and the AA is definitely appreciated.


Mouse over to see Infinity Blade on the iPad 2

The gallery below has a bunch of side by side shots showing the improvements made to Infinity Blade for the iPad 2 vs. what you get when you run the game on a first generation iPad.

As we mentioned before, the A5 (and iPad 2) are about enabling developers. In a year's time the PowerVR SGX 543MP2 will be found in the majority of iOS tablets on the market - and games are just going to get prettier from there.

Update: We've published a look at the performance of the PowerVR SGX 543MP2 running GLBenchmark 2.0 here. If you want a teaser, it's good:

The CPU: A Dual-Core ARM Cortex A9 The Display: Multiple Vendors, Nearly Identical to iPad 1
Comments Locked

82 Comments

View All Comments

  • Brian Klug - Monday, March 14, 2011 - link

    That's weird that the apple rep would say those things. The current MacBook is entirely hard plastic, and lacks the matte soft-touch material that used to stain. The current iPad is actually indeed covered in glass the same as the black model, so it will absolutely not stain.

    -Brian
  • podpi - Sunday, March 13, 2011 - link

    Hey Anand

    Let me preface this by saying I'm a huge Anandtech fan, and really trust and appreciate your word and the deep, thorough reviews you do. That's what sets you apart from the other tabloidy, first out, sort of content. They do anything to have content out first. I believe some write the reviews before they test the equipment.

    Now, whenever any big product comes out, I may glance at the Engadget/Wired reviews, but I keep checking Anandtech everyday, waiting eagerly in anticipation. The other sites try and up the number of views, so they launch bit by bit and try and draw it out, but also try and be out first with a brief headline-like review. Quality suffers. So I keep checking AT, and waiting for a thorough, unbiased, great reading review. And when it comes out, sometimes up to a week later than the rest. By then the hype had died, down, and it has a chance to come in clean, and see what someone who put the effort in actually thought of the product. The fact that your reviews tend to be a little later show to me how much effort you put in, and how much you care about quality. And this is what I appreciate.

    Now, with the iPad 2 review, it seems to be coming in drips and drabs, to try and what appears to be to keep up with the competition? I think it's natural to get into a race with those sites. But how can you win a race against tech-tabloids? If you blend in with the crowd, then there is little differentiation. I don't think you can have best of both. In my opinion, taking your time and giving a full long in-depth review, like a short story, provides a unique, quality characteristic to your reviews, site and brand.

    Thanks for the great work.
    Terence
  • Brian Klug - Monday, March 14, 2011 - link

    Terence,

    I actually agree with you that we really excel at the longer reviews, and it's good to hear that even though we're sometimes much later than everyone else, that it's still useful. I know I enjoy spending a lot of time being exacting about things and sometimes going behind schedule just to uncover absolutely everything.

    We were able to get a preview out on day one this time around just because we have about four people with iPads this time around, where with most things we'll only have one (or in extreme cases, two) units. Each of us working in parallel really sped things up and we thought we'd just share some quick impressions.

    The big review is still coming (today actually - 3/14) and will be the usual length/depth for certain ;)

    -Brian
  • Belard - Sunday, March 13, 2011 - link

    As an owner of a Galaxy S series, with poor support from Samsung/at&t and my experience of using an Android device phone for almost 6 months.

    Using the iPad2 was a breeze, its was fast and fun to use. Yeah, the missing USB & SD slots still sucks... but that won't effect sales much.

    We then went across the isle to the Samsung Galaxy Tab which costs the exact same $500. It was easily heavier, thicker... the GUI looks and functions mostly like my own phone so I had no problems "using it".

    So after that experience, we both agreed... why spend $500 on an Android, when we could get a much snappier, larger screen and much easier to use device. BestBuy was sold out.

    If the Galaxy Tab was $300~350 - it would be a more fair deal.
  • podpi - Monday, March 14, 2011 - link

    Thanks Brian - looking forward to it :)
  • ssvb - Monday, March 14, 2011 - link

    VFPv2 unit in ARM11 is actually pipelined: http://infocenter.arm.com/help/topic/com.arm.doc.d...

    And yes, naturally it is much faster than Cortex-A8 on floating point code.
  • NCM - Monday, March 14, 2011 - link

    AT: "Laying in bed and reading is probably where the difference becomes most apparent."

    You should never read in bed while laying—your partner will be offended.
    It's perfectly all right while lying in bed, however.
  • eanazag - Monday, March 14, 2011 - link

    I really appreciated the mouse over images changing comparison. It really gave a clear picture of the difference in the images.
  • retnuh - Wednesday, March 16, 2011 - link

    So.... being the 16th should we assume its getting the super mega awesome thorough testing treatment?
  • UNLK A6 - Thursday, March 17, 2011 - link

    I'd like some clarification about LINPACK and Geekbench. Are these benchmarks created by compiling some portable code for each platform as a measure of floating point performance? Or, is this supposed to be some measure of how fast one can do linear algebra or DSP on the platform? On Mac OS and iOS, one wouldn't compile say LINPACK for this but use the hand-tuned LAPACK/BLAS and DSP routines built into Apple's Accelerate Framework. The difference between the two can be huge. Which do these benchmarks purport to supply--generic floating point performance or available linear algebra and DSP performance?

Log in

Don't have an account? Sign up now