The world of silicon has been an interesting place of late.
Yesterday, we heard that the x86 market share is significantly increasing to eclipse RISC style processors in the mission critical arena, where UNIX has been dominant for thirty years.
But it is in the smaller applications that the most significant changes are likely to be made.
First of all, Microsoft announced that it would adapt a version of Windows 7 to run on ARM processors. This is a noteworthy departure for Redmond. Just as significant though, is today’s news that Intel will pay around $1.5 billion to Nvidia for its graphic processing unit (GPU) technologies. The deal will also see Nvidia get some Intel tech for what is surmised to be a low power, low end processor line.
GPUs have been getting more powerful in pure maths terms for some time now, eclipsing CPU capabilities, as multicore designs have gone in a different direction. The result is that for pure number crunching, GPUs are by far the most cost effective option. This has resulted in super computers being cobbled together from banks of GPUs. Single purpose machines of this ilk have been employed in star gazing, medical imaging and other compute intensive niches. Nvidia has been promoting the use of GPUs for such tasks, beyond the one off and custom projects that have proved the principle.
The question is whether these developments are likely to see a major shift in the balance of power in the processor world? Well, the answer is likely to be no, in my opinion. The current wave of virtualisation and cloud computing is based firmly on the venerable x86 architecture and its current multicore incarnation. As such, these side deals extending platforms to be compatible with previously exclusive operating systems (OS) and technology sharing, are more of an acknowledgement that these patches are secure and can thus be extended with other technologies without undermining their current status.
So, the rush towards desktop virtualisation might benefit from the graphic boost of GPU technologies, or maybe an encryption function might be better served from a GPU type processor without having to put a multicore x86 chip behind it. In any case, such a move would not be likely to threaten the incumbent technology.
In fact, it is difficult to see where the next major computing development will come from, in terms of processors, as the current multicore technology seems to be extensible for a few years yet. It is possible that these peripheral deals will only secure current technology lines for several years to come, meaning that the next major leap will likely be something quite radical, perhaps optical processors or the like.
In the meantime, we get to enjoy the old favourites with capabilities extended by these partnerships and technology sharing.
Subscribers 0
Fans 0
Followers 0
Followers