NVIDIA officially launched its Ion “2” platform at CeBIT. This evolution, which in fact will be grafted onto the Intel Pinetrail platform with its chipset integrated onto the CPU package, simply consists in the addition of a GT218 GPU (GeForce 210) equipped with either 16 scalar processors or half this number.
In contrast to the previous solution, this one implies a significant additional cost but may benefit from higher perforamnce (in the 16 “core” version) thanks to its dedicated memory. The Optimus driver will of course be used so as to keep the energy consumption envelope within reason.
This results in a potential problem in terms of memory bandwidth as Pinetrail only supports 4 PCI Express lanes that allow either 1 4x port of 4 1x ports configuration. What’s more, these are limited by the PCI Express 1.x clock. As a result, for example, if a wifi adaptor is used in the system, the GT218 will have to make do with one PCI Express 1x port at 250 MB/s in each direction. This is just enough to throughput the necessary data for a 720P 60Hz display and makes it hard for us to believe that there is no limitation in this respect, as claimed by NVIDIA.
However, when using HDMI connectivity, which doesn’t support Pintetrail, the GT218 will run in standard mode and pilot the display directly, thus avoiding the PCI Express limitation. Note however that 3D performances will still be limited by PCI Express bandwidth when it is used in mode 1x. NVIDIA says however that the level of Pinetrail graphics performance is so low that their solution still gives better performance.
At the end of the day, this solution truly looks like a bit of DIY that’s trying to edge its way onto a platform from which Intel aimed to exclude third party GPUs to prevent it from being used on standard PC territory. Paradoxically, to achive its ends NVIDIA has had to add the sort of complexity that it was congratulating itself on having removed on the first Ion version.