/article-new/2012/09/apple_a6_chip-150x168.jpg?lossy)
We've learned more details about Apple's new 'GPU Design Center' in Orlando, Florida, following our reporting from earlier this week.
Sources told MacRumors that the engineers Apple hired recently were not laid off from AMD, but were instead actively recruited -- largely via their LinkedIn profiles. Apple is said to have learned that many of AMD's 3D graphics patents were issued from its Orlando offices and targeted this area specifically. AMD has job listings for its Orlando offices to fill several of these recently vacated positions.
The temporary office space that Apple has leased for the new team in Orlando is located very close to AMD's campus in the city, though Apple is reportedly building permanent offices as well. Apple hired more than twenty employees from AMD and recruited more than that, with Apple reportedly looking to build the GPU team up to roughly forty engineers.
The newly hired employees are said to be reporting to the Austin-based former Intrinsity team that Apple acquired two years ago. Intrinsity technology was used in the A4 processor, and their expertise has contributed to Apple's more recent chips as well.
With its hiring of these 3D graphics specialists, Apple is likely working to redevelop its 3D graphics capabilities in its iOS devices. The company has made a number of acquisitions in recent years to revamp its chip design capabilities, and this new team would seem to supplement those efforts.
Top Rated Comments
(View all)ARM is intently competitive. Both reference design chips and custom chips.
You are mistaken to think that ARM's improvement shows that it is competitive vs Intel or AMD.Thats why we have had such an insane speed growth in ARM since Iphone. It was an 416mhz ARM.
Today we have 4+4 cores/4 cores ARM 15 at 2 to 3 ghz. Upwards 20 times faster in 7 years.
Compare that with X86 since 2006. Less then 200% speed increase till 2013.
The reason why ARM performance has increased so quickly is the same reason why CPU performance improved so quickly during the 90s and early 2000s: because they were able to take advantage of pipelining, superscalar, out of order, and speculative execution. ARM couldn't take advantage of those things until recent years because of power and size requirements of the chips. Intel and AMD already had these things implemented in their x86 chips, so it's been much harder to improve single threaded performance. ARM will soon face this same problem once it implements all the "low hanging fruit" chip features.
When it comes to competing in the high performance market, the advantage of the ARM architecture declines dramatically. The debate over what instruction set architecture is best should be left in the last decade. Instruction sets don't matter much anymore. The future is about who implements heterogeneous computing the best.
Have you actually read what is all part of what you call "XBox division"? One part of that division happens to be the Macintosh software unit. So every time someone buys Microsoft Office for the Mac, it adds profit to the so-called "XBox division".
Macintosh software is a part of their entertainment and devices division? News to me.
ARM is intently competitive. Both reference design chips and custom chips.
Thats why we have had such an insane speed growth in ARM since Iphone. It was an 416mhz ARM.
Today we have 4+4 cores/4 cores ARM 15 at 2 to 3 ghz. Upwards 20 times faster in 7 years.
Compare that with X86 since 2006. Less then 200% speed increase till 2013.
Its incredible that Apple choose to compete in this market when there are so many vendors to pit against each other. But this is Apples unique position: they control both hardware and software. This gives Apple the ability to put custom DSP and stuff into their SoC's just for their OS.
The A5 SoC had 30% die area with Apple specific stuff. Like the "Apple visual processor (in real world NOVA SIMD), noise cancellation DSPs for Siri, Dual channel RAM and so on.
Apple will continue with their custom SoCs. One thing that is real interesting is Anobit and their/Apples SSD controllers. We will move to fast NAND FLASH chips in iPad/iPhone. Imagine having 550 megs/sec bandwidth. Its possible with lightning to thunderbolt connector or gigabit WiFi.
Remember that Apple's goal is to replace Intel with their own SoC. Todays AIRs/Macbook pros the Intel CPU/Motherboard is 50% of BOM. Apple and the consumers could save huge by dumping X86. A 1970 technology.
I think you're over-romanticizing the rise of ARM and exaggerating x86's faults. ARM's rise owes a lot to all the money pouring into it. More money means more engineering and it also means higher demands. TDPs have also been going up, based solely on the principle that your processor is on less and your average usage is still constant.
Meanwhile, x86, an already mature technology, has always had the high performance requirements in a relatively static TDP. It's also been converted into a mobile platform that can finally compete with ARM on a performance/watt basis.
Well one thing for sure people expecting to see the fruits of these hires in the next few months are way off base.
Yes, it took several years for Intrinsity and PA Semi to crank out Swift. Expect the same time-frame for a custom GPU implementation.
They should move to Melbourne where authentec is. Much cheaper and less traffic.
As someone who lives near Authentec in Melbourne, I would love to have a bigger "Apple design center" here :D
He's talking about Melbourne, FL silly. Its an hour away from Orlando.Australia is quite expensive nowadays due to the exchange rate.
So I wonder what spurs this move for Apple to create its own GPU design center. What are they looking for that Imgtech has not provided them? What are they going to use these resources to do?