Here is a Letter from Trident Systems in response to this article before it was published.

Hi Andrew,

(1) We appreciate very much that you had taken time to re-evaluate our XP4m32 graphics chip in the Toshiba Tecra M1 and gives a new assessment about our product.

(2) We agree that benchmark evaluation should be better regulated to give a fair assessment with respect to quality-versus-performance tradeoff.

(3) In case you are not aware of, driver optimization for popular benchmarks and leading 3D games has been a standard graphics industry practice for the last 16 years and will probably continue to be so in a foreseeable future. It is very unfortunate that Trident XP4m32 was put under the spot-light due to an earlier "bugged" driver. However, since it also impacts the reputation of our OEM customer Toshiba, we feel very bad about it.

(4) We don't intend to "hide" anything about the image quality versus 3D performance tradeoff, as this is a well-known optimization technique in our 3D graphics industry. Nevertherless, it is very unfair for us to be singled out under the microscope and being negatively labeled as "cheating". No other graphics suppliers offer an user-selectable, quality-versus-performance option to run the standard benchmark today. It is always up to the OEM customers and the end-users to decide whether they are satisfied with the quality-versus-performance at the particular screen resolution of interest.

(5) Driver optimization is a fair game because every 3D graphics vendor is doing it in this extremely competitive market. Furthermore, it is not really "cheating" since the user can actually see what they are getting, i.e. slightly lower image quality in exchange for significantly higher 3D performance. We would however agree that it will be better for ALL customers if the graphics industry can adopt such practice in offering quality-versus-performance option to the users.

(6) We sincerely hope that you will refine your article so that it would better convey our perspective and clearly reflects that we are equally treated as all other graphics vendors, especially ATI. We have strong reasons to believe that your reader Bill Smith is actually an ATI application engineer !

Much thanks.

Le

In response to the letter, I'm sure you are right in saying application detection goes on with other companies, but not in the same way you have. For instance, they optimize a program to work more effeciently with the way their card handles pixel shaders. Much like programs optimized for SSE instruction sets ran faster. What you have done is the equivalent of taking the seats out of a car to make it go faster (which there is nothing wrong with) but then claiming it still has all the luxury of a car with seats. As far as it not being cheating becuase customers can see what they are getting - that is also wrong. Many people go by the numbers they see on web sites and magazines and may in fact be surprised by the poorer image quality when their laptop arrives - as many people have been. 5400 3d marks should be 5400 3d marks. A benchmark should literally be that. How else can customers see what they should buy?

As far as Bill Smith being an ATI representitive, other readers have also uncovered this "bug".

I personally feel that everything I have said in this article has been fair and quite founded.

Spode

<< Back | 1 | | 2 | | 3 | Next >>