As of today, Nvidia's second-generation of mobile, Fermi
-based parts is complete--at least, in theory. The 570M and 580M will only be available from select OEMs (at least initially), with the 570M reportedly available "immediately" from MSI as a configuration option on the GT780R gaming notebook. Alienware M18X systems can be ordered today with a GTX 580M, but "should ship in a week or so."
According to Nvidia, the two new high-end GPUs deliver both speed and more efficient power consumption. It describes the 580/570M as "up to 20 percent better performance per watt than the previous generation," while simultaneously claiming that the 570M is "20 percent faster... than the previous generation" it replaces. (aka the 470M).
The two new parts. The 570M offers quite a few more cores than the 470M did.
Like the 5xx desktop series, these new notebook parts are built on 40nm technology. If you're wondering how the 580M breaks down against NV's previous heavy-hitter, the 485M, the differences are subtle. The 580M is a new chip and incorporates the efficiency improvements that distinguish the desktop 5xx cards from the previous 4xx series. The 580M is also clocked slightly higher--its core, shader, and memory clocks are 620, 1240, and 1500MHz. That's eight percent higher than the 485M on the first two figures, identical on the third. Both the 485M and the 580M offer 96GB/s of memory bandwidth and a 256-bit RAM bus.
The 570M is made of sterner stuff. While its core speed is just 4.5% higher than the 470M's, the 570M contains significantly more shader cores (336 vs the 470M's 288, a gain of ~16 percent). Like the 580M/485M, the 570M/470M have the same amount of memory bandwidth and a common 192-bit memory bus.
As far as GPU
launches go, Nvidia isn't blazing a new trail here. That's not NV's fault as such--the best/most memorable mobile launches occur when a GPU architecture debuts on a new process technology. That shift typically allows NV and/or ATI to significantly improve performance and simultaneously slash power consumption. Everyone trots out agreeable analysts to exclaim over how the mobile 3D market is improving.
Given the less-than-positive reception DNF has garnered, this might not have been the best marketing move
Without a 28nm die shrink handy, NV had to rely on rearchitecting the 4xxM hardware just as it overhauled the 4xx series to create the much-improved 5xx cards. We suspect, however, that some of the power improvements we saw on the desktop were baked into 4xxM hardware last year. Notebook TDPs are strict--NV may not have had enough wiggle room to more sharply differentiate performance between the old series and the new. But we'll know more when we test some actual hardware, which should be happening soon.
Make no mistake, the 570M and 580M are upgrades from their 4xxM counterparts, but no one who bought the older high-end cards needs to feel performance anxiety quite yet. The delta between the GT200 series to Fermi was much more significant than this update--the 5xxM cards are nudges in the right direction, not game-changing leaps.