14 DAY TRIAL // Though Fermi is, currently, NVIDIA's biggest pride and joy when it comes to GPUs, it looks like this architecture won't hold this position for very long, at least not if the recently revealed graphics processing unit roadmap is anything to go by.
As one may well know, mostly everything that came from NVIDIA's labs recently has been one iteration of the Fermi GPU architecture or another.
This technology has mostly allowed it to reclaim the performance crown on the single-GPU consumer front.
However, it seems that Fermi's glory may be less long-lived than one would expect, apparently set to be deposed by its successor in 2011.
Said successor has the codename of Kepler and, according to what NVIDIA reportedly said during its GPU Technology Conference, its engineers are already working on it.
The main asset of the new graphics processing unit is that it will be constructed on the 28nm manufacturing process technology.
This will actually let it deliver a performance-per-watt improvement of about three or four times compared to Fermi while, hopefully, not creating as much heat.
Nevertheless, what will truly cause a crater upon impacting the market is the so-called Maxwell GPU, slated for launch in 2013.
Basically, this processor will have a per-watt performance increase of a full sixteen times over today's Fermi, which is quite impressive knowing its arrival is set for just two years after Kepler.
What's more, the Maxwell is expected to be able to perform certain processes completely independent of the central processing unit. It will use the 22nm process.
Of course, NVIDIA also reaffirmed its intention to finish the top-to-bottom refresh of its GPU lineup within three months.
Finally, the outfit offered the assurance that it will unleash "mid-life kicker" products in between these major releases, such as a Fermi refresh next year.