Intel (NASDAQ: INTC) has made no secret of plans to integrate graphics with the core CPU but has not gone into detail on layout or exactly when it will ship. The chips are reportedly planned for Q2 2009, which would put it ahead of AMD’s planned launch of Fusion in the second half of 2009.
The processors are being developed under the code names “Havendale” (desktop
processor) and “Auburndale” (mobile processor).
Intel declined to comment on “rumor and speculation.”
According to reports, Intel will combine its forthcoming Nehalem
processor with a next-generation graphics processor from its current G45 chip.
The G45 is a decent performer, with support for DirectX 10 and Shader Model 4.0 as well as HD decoding and HDMI
The exact architecture is not known. A Japanese news site, PC Watch, has plenty of architectural
designs that show Havendale/Auburndale as a dual-core processor with integrated graphics and a 4MB cache for the two cores.
Havendale will be use a new desktop-socket design, LGA-1160, and Auburndale will use the new mPGA-989 mobile socket. The desktop chip will have a 75-watt power envelop, while the mobile part will be between 45 and 55 watts.
It’s quite a change for Intel to react to something AMD still has in the labs. It was dismissing 64-bit x86 extensions back when AMD first introduced them, then belatedly came to market a few years later with 64-bit chips.
“Intel is never going to let AMD get too far ahead of them again,” said Jon Peddie, president of Jon Peddie Research. “They learned a hard lesson. I think Paul Otellini is probably one of the most enlightened CEOs Intel has had for a while and he’s just not going to let that happen again.”
Peddie believes that Intel will take the same strategy it took with its first dual-core and quad-core chips, and put two parts on the same die before integrating it.
“The pros of this approach are that it allows you to test the concepts and get more yield by using proven parts, and it allows you to make more SKUs because you can mix and match more easily,” he told InternetNews.com. “The cons are there is an inherently higher silicon cost, and there is a higher power consumption due to the interface logic between the CPU and GPU chips.”
AMD, being the No. 2 contestant in the field, needs to take a bigger risk for the bigger reward, so it’s merging of CPU and GPU is more ambitious, added Peddie.
The question then becomes, what good is it? Fred Zeiber, president of Pathfinder Research, wasn’t exactly sure himself. “You do get some advances in I/O but it’s not like it’s the Holy Grail,” he said.
“They [Intel and AMD] probably feel that there’s a market there. What they see there, I’m not sure but you have to take into account changes in technology and Internet delivery. There may be markets coming down the road.”