X-bit labs

AMD’s Fusion On Track for 2009 – AMD’s Chief Technologist [CORRECTED].

Category: CPU

by Anton Shilov

[ 01/16/2007 | 11:45 PM ]

CORRECTION: A spokesperson for AMD said that InfoWorld web-site incorrectly quoted Phil Hester as saying that the first commercial Fusion processors emerge in late 2009. Insted, the company stressed that the first products to combine general purpose and graphics processors will be produced in early 2009. InfoWorld web-site, which incorrectly quoted Mr. Hester, has corrected its interview.

The chief technology officer of Advanced Micro Devices – Phil Hester – said in an interview that the highly-discussed microprocessors that integrate both general processing and graphics processing capabilities will be in production only in late 2009, which essentially puts their availability into the next decade – the 2010 and beyond – later than the company originally said.

Fusion Comes in 2009

“The first Fusion CPUs will be available in prototype in late 2008 and in production in early 2009,” said Phil Hester, chief technology officer of AMD, in an interview with InfoWorld.

Earlier the company indicated that Fusion processors “are expected in late 2008/early 2009”, and the company anticipated to use them within all of the chipmaker’s “priority computing categories”, including laptops, desktops, workstations and servers, as well as in “consumer electronics and solutions tailored for the unique needs of emerging markets”. This time Mr. Hester confirmed that the first applications to use Fusion would be notebooks, but declined to indicate, when the company plans to offer something for other computing platforms.

“We’ve already, through discussions with our customers, reached agreement that the best place to initially focus this is the notebook space. That is where you get the most benefit of the efficiencies from a power standpoint and also from a physical area standpoint,” Mr. Hester said.

3D Graphics Processor’s Destiny Is Inside CPUs

Advanced Micro Devices, the world’s No. 2 maker of x86 microprocessors, believes that graphics processing units (GPUs) now are in the same position as floating point co-processing units (FPUs) two decades ago and are meant to be integrated into central processing units (CPUs).

“If you went back to the early PC platforms in the mid-1980s with the 286 generation, you had an open socket next to it for the 287 math coprocessor, and at that time, a very small fraction of the population ever bought that math coprocessor because the applications didn’t demand it. The same thing happened with the 386 processor, although you had 20% of people using it. And then the 486 integrated that as a standard part of every microprocessor, driven by the fact that the operating system and the applications demanded it,” Mr. Hester said.

The company’s chief technologist believes that while standalone graphics cards – such as those developed by AMD’s graphics processor division ATI Technologies – have a lot of power, the vast majority of consumers hardly need them, but will certainly need 3D graphics going forward, particularly due to rather intensive usage of graphics capabilities by operating systems like Microsoft Windows Vista.

“Essentially every client you see now is going to have a decent level of 3D graphics capability… If you look at the things driving that, whether it’s business information rendering or entertainment, by the 2009 time frame, just about every PC is going to have to have some level of 3D graphics. Not just the top end of the enthusiast market, but it is going to be a standard PC experience without having to incur the cost and efficiency hit of using a separate graphics card,” he said.

Mr. Hester was not questioned what level of performance could be expected from CPU-integrated GPUs and how will it compare to discrete GPUs. Currently there are chipsets – which are used by the majority of personal computers sold – with built-in graphics core, but performance of 3D graphics offered by such core-logic sets is much lower than that compared to standalone GPUs.

Though, if we take a look back at the add-in FPU market, we will remember that the massive deployment of built-in FPUs instead of add-in FPU took no less than four years. Among Intel 80486 microprocessors family – introduced in 1989 – there were 80486DX and 80486SX chips with and without FPU respectively; whereas all Intel Pentium processors – unveiled in 1993 – included FPU, which essentially forced company called Weitek, which was one of the largest suppliers of add-in FPUs, to merge with Rockwell Semiconductor Systems (now known as Conexant) in 1996.