… discreetly. NVIDIA has introduced the GeForce GT 420 without any publicity. At first it will only be available for OEMs. This new entry level card is the first to be based on the GF108, the smallest GPU from the Fermi family. With this discreet arrival, NVIDIA is no doubt aiming to benefit from a particularity of this card which gives it an advantage over the competition AMD products…
While the latest entry level GPUs were interfaced at 64 bits, the GF108 has a 128-bit memory bus, which means budget cards can easily be marketed with 2 GB of video memory. This is of course a heresy as this much memory is completely useless on such a small GPU, which moreover has been cut down so that only half of its processing units (or 48 scalar units) are activated. Its fillrate is extremely low. With their 64-bit bus, the GT218s (GeForce 210/310/315) and Cedars (Radeon HD 5450) must “make do” with 1 GB of memory based on standard components.
Even the name of this GeForce GT 420 is arguable as, outside of DirectX 11 support, its spec is inferior and often greatly so, to that of the GeForce GT 320, which should continue to give significantly better performance in gaming.
As you can no doubt tell, this GeForce GT 420 is being marketed to attract the business of PC manufacturers who see in it the opportunity to attract consumers who all too often only consider the quantity of memory when choosing a graphics solution.