NVIDIA Next-Gen ION Preview, Optimus Enabled

Article Index:   

NVIDIA Next-Generation ION 2

Ever since NVIDIA's original ION chipset proved it was possible to pair Intel's low-power Atom processor to a relatively powerful graphics core and attain acceptable multimedia performance, the market has been curious as to what NVIDIA's Ion 2 Next-Generation Ion product might offer. This has also been a topic of some discussion between Intel, NVIDIA, and the FTC. Because Intel's next-generation Atom platform is a system-on-chip (SoC), NVIDIA had to adjust its own plans for NG-Ion.


That's Ion on the left, NG-Ion on the right. Parse the diagram and you'll soon realize that NG-Ion is a discrete GPU, unlike its predecessor. Specifically, it looks to be a low-power, netbook friendly version of the GeForce 210M (although not all NG-Ion chips will have 16 cores). Like the 210M it's produced on a 40nm process (down from Ion's 55nm) and clockspeeds are rumored to be higher than the original model as well. Most significantly, NG-Ion packs a dedicated 512MB frame buffer where the original chip had to share main memory bandwidth with the Atom processor.

A 64-bit link to a 512MB frame buffer at 800MHz (that last is a guess) doesn't sound like much, but it'll give the NG Ion GPU 6.4GBps of memory bandwidth compared to the 3GBps available to a standard Pinetrail system. NVIDIA claims these improvements boost NG-Ion's performance up to 100 percent above its predecessor.

Oddly, NVIDIA doesn't specify if we'll see NG-Ion paired up with Intel's Atom Z550 or the N470. At 2GHz (Z550) and 1.83GHz (N470), these are the two fastest Atom processor's Intel has released. Given the diminutive processor's similarly diminutive performance the faster speed bins would seem to be quite desirable pairings for NG-Ion.

So far we have the NG-Ion GPU—a faster Ion manufactured on a more advanced process with a sizable frame buffer, DX10.1 support, and one more thing—a sexy new power-saving feature named Optimus.

Related content


blog comments powered by Disqus