Entry level graphic cards and integrated chipsets (page 1: Introduction, Intel)

Not everyone is prepared to spend several hundred euros in the acquisition of a graphic card. This is something that GPU manufacturers know and this is the reason why they also have cheaper solutions. But is it really possible to play with them? What about Intel's integrated solutions?

Intel, the 3D leader?

Intel's graphic integrated solutions are generally well known for their lack of performance or incompatibility with some games. Their integration in many computers, whether they are desk or laptops, make Intel the leader on the graphic market with a 40% share while ATI and Nvidia represents more or less 25%.

The integration of a graphic core in the chipset reduces fabrication costs, but of course this doesn’t give you a very complex chip for a low price. Intel could have only included the support of 2D and abandoned 3D for graphic cards, but because of the necessity for compatibility and for a more attractive solution compared to the competition, integrating 3D became almost unavoidable. However, it’s difficult to integrate the performances of a 150€ (or more) graphic chip to a chipset. Also, it’s no secret that an integrated chipset that costs a few more euros than the standard chipset will not replace a graphic card.

However, as long as chipsets are sold as integrated graphic solutions that allow you to play, even if they are entry level, it is only logical for us to expect some performance. Are these graphic cores as bad as everyone believes them to be?

From an architectural point of view, Intel's graphic cores aren't as bad as we could think and they are often more advanced that they look. For example, the 915G with a GMA 900 core has a similar calculation power as the GeForce 6600 GT. In practice, however, it’s far behind for many reasons. The first is that the memory bandwidth is shared between the components of the configuration and consequently is quite low. Also, Intel's graphic cores generally lack the minor functions that help to save resources here and there, and which lead once combined to significant performance gains. The last factor, the driver, is the most important and developing efficient and complete versions requires a lot of work and large teams.

This was something that has been partly neglected by Intel, probably because it wouldn't pay off. Why invest huge amounts money in drivers that are intended for low income products? This is an important point that will probably change in the future because of the recent officialisation of Intel's ambition to sell higher end graphic chips. As most development of drivers is similar to the product line, Intel could finally make the necessary investments to have better versions. You should note that when we say “better” drivers, we actually mean higher performances as well as the support of all capabilities of graphic cores. We are not referring to stability issues.