The Take on Tech

nVidia’s Fermi line of graphics cards’ launch date has been set at March 26, a long ways away from AMD’s launch of its Evergreen architecture last year. This means that Fermi had better be better in quite a few ways. There are a few glaringly obvious things that nVidia must do with Fermi. For example, Fermi must support DirectX 11, Microsoft’s newest graphics API. This point is so important that it is virtually a given that nVidia has incorporated support of DirectX 11 into Fermi. Fermi had also either be faster than AMD’s flagship 5970 and 5870, or a better deal than AMD’s products, preferably both.

Perhaps a summary is required for those who have not been keeping up with the latest developments in the graphics cards realm. AMD released its 5870 and 5850 GPUs in September of last year.

Yet, apart from those basic points, is there anything that Fermi could do to make for that huge time gap between its release and the release of AMD’s 5xxx series lineup?

1. nVidia could incorporate its Optimus technology in a desktop setting. nVidia developed Optimus, in short a system which allows for a laptop to switch seamlessly between its integrated and discrete GPUs. Apart from the energy savings that such a system could realize on a desktop (which for many users would admittedly be only a relatively small, but not necessarily insignificant, monetary benefit, probably much less than $100 a year), such a system could allow for quieter systems. A passively cooled integrated processor could take the burden of less graphics-intensive tasks. When a user actually initiates a graphics-intensive task, the task itself may require enough of the user’s attention that he or she does not notice the additional noise of the actively cooled discrete graphics processor (e.g. a computer game). Of course this may not hold for tasks such as GPGPU computations, but at least a significant number of situations could be made quieter.

2. nVidia could increase support for GPGPU (General Purpose GPU) tasks., i.e. using the GPU to go beyond simply graphics processing to running other programs. Right now nVidia has the lead in GPGPU programs over AMD. If nVidia and its partners bundled useful GPGPU programs with their graphics cards, it could lead to an uptick in consumer interest in GPGPU computing, a boon for a company already ahead of its competitor in this respect.

3. Focus on value rather than power. This was the strategy AMD took to beat nVidia with the 4xxx series of Radeon HD cards. The 4870 wasn’t the fastest card on the market when it came out, but it had an unbeatable value. Although it is definitely way too late for nVidia to do this now, there’s still hope that nVidia will be coming out with lower-binned cards that provide more bang for the buck than their 5xxx Radeon counterparts currently out there.

nVidia better get its act together with Fermi. A duopoly in the discrete graphics market already limits the competition. Hopefully, that duopoly won’t shrink into an AMD monopoly some time in the forseeable future.