AMD to the future

At last week considering the 6th birthday of the central processing unit of treatment of Opteron of AMD, the core of its product of waiter. We were among an small group of journalists and analysts with the campus of Sunnyvale of AMD for the event - which included also the launching of a new generation of silicon, and the revealing of the roadmaps of Opteron of AMD during the next years.

Specifying that �the market of waiter is very different�, Nigel Dessau of AMD opened even with a glance with the manner the market of waiter had changed during the last years - with a movement into exit rather than the frequency of clock with the systems multinucl�aires, with the virtualisation, and to conceive for energy efficiency, all with an aim of changing the economic scenes of the computer centre. The result was the density of waiter and the increaded use - more blow for your male in less space!

Like occurs afterwards? Dessau suggests that �the manner that you assemble architectures is what makes the difference.� This is why AMD puts so much in its central processing units of treatment, with much of what we would normally consider the tiddlywinks of support - including/understanding controllers of memory - on same silicon. It is not completely a system on a chip, but it obtains very narrowly nowadays.

Stalemate Patla, the GM and VP of the unit of company of the waiter of AMD revealed the new material. It was the course of the product range of Istanbul, a processor of six-core for two existing, four and eight systems of casing with what he said were �30% more execution than the former generation with the same power�. Istanbul also introduces all technologies of management of the power of AMD into a place like AMD-P.

Istanbul is not the end of the history - the next chapter already is written inside. AMD takes already its processor of next generation, 12 the core �Magny-Course�, that in spite to be a terrible pun, is a piece of silicon rather energetic - we observed a system of demonstration of 48 cores much of its way by several reference marks. It is not that remote, either, and should be transported in 2010.

The virtualisation is a target for the next generation of silicon, and she will add the support of the virtualized input-output. The units of i/o being virtualis� there is much place for the new application and the new manners of work (as well as a chance of virtualiser applications and waiters which were previously locked in the existing material). It is all the part of infrastructure 2.0 model, where the management tools benefit from the material to provide the flexible individual-controlled virtualized computer centres.

The other part of the history of AMD is how that functions on the management of power, with products available in various thermal bands - including/understanding a weak range of power intended for the applications to high density of simple and duelle casing, ideal for the computer centres of cloud. It must there be known as much for this approach, particularly like all the devices of the central processing units of treatment of power end high and high are in the versions of ideal efficiency. The dense deployments need greater effectiveness, because the costs of the computer centre are an enormous proportion of the costs engaged while directing a service of cloud. With an average power of 40W, the processors of series of the EE will help to keep these cooling and of power of the costs.

Magny-course is only part of the roadmaps of AMD, and the current architecture of the company is a long-term play (which is good for the virtualisation, because it will allow asymmetrical migrations, leaving the use of businesses a material that older for the covering of the losses purposes). The next generation will be the devices 32nm in 2011, a Interlagos device called from 12 to 16 cores and a Valence device called from 6 to 8 cores.