tequila bomb

We had decapsulated the A5 a couple of days ago, but as you could see in those early pictures, you can’t tell much of a chip’s layout from the top metal – it’s all power and ground buses.  So we have to de-layer the chip down to a level where we can see the block layout of the chip; not an easy thing when there’s nine layers of metal!  In fact, these days it’s easier to go in from the back and remove the substrate silicon, and look at the gate level from below.  Then we can identify the circuit blocks that make up the full device.

[W]e’ve labeled the key blocks; the ARM cores are in the right half of the die, with ~4.5 Mb of cache memory each.  We can also see the USB interface at the top, and the DDR SDRAM interfaces at the bottom right, for the memory in the top part of the package-on-package.  Other I/O blocks are strewn around the edge of the die.

Interesting stuff.

An amazing array of magnets, both in the iPad 2 and the Smart Cover, with polarities carefully organised, and each magnet positioned, to ensure that when you snap it on it “Just Works”. Very cool. Very Simple. Pretty much exactly how I assumed it worked. But the amount of R&D and practical experimentation that went into making sure the concept behaved as expected is obvious.

“To those who have been affected by the earthquake and tsunami, we send sympathies from our heart – In this deep …sadness, we are praying for the victims and their families.”

*You can donate to the Red Cross via iTunes here. Apple has of course waived its 30% cut.

*Unfortunately this link does not work in some iTunes regions, like for example Thai accounts. Please use a US account to make your donation until Apple fixes this.

This is the first definitive set of GPU benchmarking for the iPad 2, courtesy of AnandTech…

Developers with existing titles on the iPad could conceivably triple geometry complexity with no impact on performance on the iPad 2.

With a lit triangle, a simple test for fragment shader performance…

While the PowerVR SGX 535 in the A4 could barely break 4 million triangles per second in this test, the PowerVR SGX 543MP2 in the A5 manages just under 20 million.

Texture fetch…

5x increase in texture fetch performance. This has to be due to more than an increase in the amount of texturing hardware. An improvement in throughput? Increase in memory bandwidth? It’s tough to say without knowing more at this point.

Those of us working with multiple FBOs for post processing effects, should be very pleased with both the increase in fragment shader performance and the apparent increase in texture fetch throughput.

GLBenchmark 2.0 app & Infinity Blade tests…

While we weren’t able to reach the 9x figure claimed by Apple (I’m not sure that you’ll ever see 9x running real game code), a range of 3 – 7x in GLBenchmark 2.0 is more reasonable. In practice I’d expect something less than 5x but that’s nothing to complain about.

There are more in depth details in the full article, and they promise follow ups.

But for a simple head to head comparison running GLBenchmark 2.0 Egypt…

Apple iPad 2 (1024 x 768) : 44 FPS

Motorola Xoom (1280 x 800) : 11.8 FPS

Apple iPad (1024 x 768) : 8.1 FPS

As usual. Excellent work by the guys at iFixit.

Apple 1GHz A5 dual-core Processor with a 200MHz bus and 512 MB of RAM.

The A5 processor has manufacture dates of late January and mid-February 2011. Production was clearly ramping up through the last minute. It looks like the A5 processor is the APL0498, replacing the A4/APL0398 seen in the iPad 1 and iPhones.

Of course what we really want now is the Apple A5 ripped apart in one of their friend’s labs.

Digital Foundry talks with Guerrilla Games’ technical director Michiel van der Leeuw about the development of Sony’s latest first person shooter, Killzone 3.

One of the things we’re very proud of is one that nobody sees: We’re doing a full depth rasterisation of tens of thousands of triangles in a software rasteriser on SPUs so we can do occlusion culling against it. This allows us to do much more aggressive culling, which in the end allows us to do more complex scenes and further draw distance.

…

With MLAA we reduced our load on the RSX quite a bit, we halved the size of the G-buffers and some depth/stencil re-primes between passes were no longer necessary, which gave us additional speed boosts. In the end it gave us quite a bit of RSX cycles and memory back, which we invested back into texture streaming, draw distance, texture sampling quality and – more indirectly – the variety in characters, environments and so on.

We were already doing most of our post process effects on the SPUs, but with MLAA moving to SPUs as well we had the entire post chain running there.

The whole article is full of fascinating little bits of insight into how Killzone 3 ticks under the hood.

If you’re not already doing, or at least looking at doing the things mentioned in this article, and you’re working on the PS3, then you are behind the times.

When an entire generation of computer users first poked our doe-eyed faces onto a young internet, many of us were greeted with a single, encompassing, monolithic face peering back: the AOL Home Screen. To call it a young internet isn’t even fair—it was a mature, thriving AOL. It was ubiquitous, it was powerful, it was everything—and it ended up destroying itself, too flawed by design to last. And someone’s trying to rebuild the Death Star.

From Gizmodo.

Interesting piece from Sam Biddle. I, and others, have been banging on about this for a while now.

Microsoft Corp. will pay Nokia Oyj more than $1 billion to promote and develop Windows-based handsets as part of their smartphone software agreement, according to two people with knowledge of the terms.

Nokia will pay Microsoft a fee for each copy of Windows used in its phones, costs that will be offset as Nokia curtails its own budget for software research and development, said one of the people, who declined to be identified because the final contract hasn’t yet been signed. The agreement runs for more than five years, the people said.

At a $15 per phone royalty that Nokia will be paying Microsoft, Microsoft is betting on Nokia selling 60 Million handsets running Windows Phone 7 for the deal to break even.

For some perspective: Apple passed the 100 Million iOS devices sold mark in the summer of last year. That’s iOS devices, which includes the iPod Touch and iPad, not just smart-phones.

We think the A5 is likely not built around Cortex A9 cores, but instead probably uses two [of] the same custom low-power A8 cores used in the A4. If Apple had indeed used two Cortex A9-based cores, raw performance should be more than double that of a single core A8-based design.

This makes a lot of sense. I noted in February that Apple had received custom silicon for what we expected to be the Apple A5. But that it had not had enough time to transition that silicon into iOS (or ongoing iPad 2 manufacture) for an early 2011 launch. So my best guess was that the iPad 2 would ship with an interim SoC. I suggested something like a beefed up A4, with a faster ARM Cortex A8, and a much better GPU most likely making up an iPad 2 specific Apple A4-and-a-half.

To be honest, until someone (iFixit and friends) rips the silicon in the iPad 2 apart and sticks it under a microscope, none of us will have much more than guesses to go on about what exactly the Apple A5 is. But it seems very likely that Apple has made expedient decisions to maximise performance as well as keep battery life gains.

I’ve have always maintained that Apple’s mobile silicon lineup is more than powerful enough in the CPU department, and what it really needed was a kick on the GPU side. Even the CPU in the original iPhone is still very capable. But the GPUs in all current iOS devices are constantly fighting an uphill battle with fill rate.

It remains to be seen if the iPhone 5 will get this Apple A5, or a further iteration.

Google removed a number of malicious applications from the Android Market last week.

In addition to preventing further infection by removing the malicious applications from the Android Market, Google will also be using its remote kill switch to forcibly uninstall the application from infected handsets. The company is also pushing out an update to the Android Market that can reverse the exploit, thus preventing the attackers from using it to cause further damage. Google has already started to send out e-mails to affected users in order to explain the situation.

Although Google can deploy software to undo the damage caused by the malware, the underlying vulnerability that the attackers exploited can’t be closed so easily. Google says that the bug is fixed in Android 2.2.2 and later, but there are still a large number of users at risk because their handsets runs a previous version of the operating system. Google is making a patch available, but it’s going to be up to the carriers and handset makers to make sure that the patch gets deployed. In light of the mobile industry’s poor track record updating Android phones, it’s possible that this flaw will continue to be exploitable on a considerable number of handsets.

Another problem with Google’s “Open” approach to their smartphone OS.

We’ll see just how much service providers care about their customers’ security versus control over, and monetization of user features, in the coming days.