The confusing number scheme from Nvidia and AMD can be really frustrating when you are buying a new laptop and trying to figure out which has got the best GPU onboard. On top of that, vendors always confuse the customers advertising the memory on their GPUs instead of bus size, speed and things that actually matter. As such it results in scenarios where your brand new laptop with 2 GB graphic memory is not able to run even 2 year old games. To avoid this, here are a few key points to keep in mind.
More RAM doesn’t mean more performance
Companies advertising laptops with as much as 2 GB of graphic memory are only trying to dupe the customers and it is nothing more than a market gimmick.
Under normal circumstances it would seem that more amount of RAM on the GPU should definitely add to the performance. And it does, but only up to a certain limit. After that it depends on the bus width on the chip. Bus width allows data to get transferred through it and thereby get accessed by the RAM. Bus width come in sizes of 64 bit, 128 bit, 256 bit and so on. So say for a GPU with bus width of 128 bit, memory above 512 MB would be useless because it wouldn’t be able to utilize all of it at the same time.
An analogy to this would be the car and road example. Suppose you have a Ferrari that can go up to 300km/h. However, if the road is only two lane wide you wouldn’t be able to go above a certain speed due to all the traffic and curves proving to be a bottleneck. Similarly a smaller bus width would bottleneck the amount of usable memory, rendering the extra RAM useless.
So it might be that you have a laptop with 1 GB memory from 3 years back and it would still perform poorly while running recent games.
Bandwidth is the product of bus width multiplied by the memory speed. It is how fast the memory is communicating with the GPU. Just like RAM on your PC is classified as DDR1, DDR2 depending on the frequency (higher the better), the same theory goes for GPU memory. GPUs with GDDR5 have been available for sometime now.
Bandwidth directly equates to data transfer rates. And the more, the merrier it is.
Nvidia 6xxM is not always better than 5xxM or the ATI 6xxxM is not always better than 5xxxM.
This one is pretty easy to get confused with. The first number in the naming represents the generation that the graphic card belongs to. The rest correspond to the performance. The low end cards have a lesser number than the high end cards.
For example, an Nvidia 280m belongs to the high end of this particular generation and say an Nvidia 320M that belongs to the low end segment of the next generation are pegged against each other. Even though the 320M is of a newer generation, the 280M will beat it by quite a margin in all benchmark tests. Similarly an ATI 5770M is better than 6330M.
If you have taken care of all these things the next advisable thing to do would be to check out benchmark scores on various sites on the net and compare performance between the cards you’re trying to choose from. Sites like http://www.techarp.com or http://www.notebookcheck.net offer a wide variety of benchmark scores.