Random Phrase: Home sweet home...


Read Me First!

This is the Spode's Abode archive. The old articles and forum have been kept here purely for historical purposes.

This section of the site is not kept up to date and there is no guarantee that all portions of the site transferred to this sub-domain intact.

Please visit the new site!



Trident Cheating Benchmarks? (Part 1)
Written by Bill Smith (11/May/03)
Page 2 of 2

Untitled Document

Note from Kaitain

Within hours of this article being published, I was contacted by Le Nguyen at Trident Microsystems, who argues that this selective acceleration of certain applications is a necessary feature of the Xp4 chipset. To ensure fairness, Le's response is recorded here:

Hi Spode Abode Team,

It seems that your reader Bill Smith has misunderstood Trident XP4 product greatly ...

In case you don't know, Trident XP4 hardware architecture is designed to require only 50% the number of transistors as compared to the competition and yet still deliver 70% of the competition's performance.

This unique capability has been achieved by the use of proprietary 3D algorithm optimized for multiple pixel rendering. In addition, Trident XP4 also implements proprietary software + hardware support to accelerate special 3D graphics features that could become the performance bottlenecks in various 3D graphics applications.

In order to selectively enable this performance acceleration, Trident XP4 driver is designed to recognize the names of these popular applications of interest. Therefore when you change the names of the applications of interest, the optimizations will not be enabled and the performance is reduced as the result.

This is not cheating in the driver, it is selective performance optimization of popular applications of interest by using an unique combination of software + hardware algorithm proprietary to Trident.

Of course, the secret here is knowing what these software + hardware optimizations should be :)

Please let me know if you have any additional question.

Thanks.

Le Nguyen
VP, Graphics Marketing
Trident Microsystems

In response I asked why these performance optimisations he mentioned are only selectively enabled, rather than being used for all D3D or OpenGL applications:

On Tue, 2003-05-13 at 00:05, Le Nguyen (Old Email address) wrote:

Hi Kaitain,

The reason why we don't enable these performance optimizations at all times is because each 3D application has different types of performance bottlenecks.

In order to achieve this 50% reduction in transistor count, our design strategy is to optimize the performance of some of the key apps that account for the majority of the customer uses.

Please let me know if you have any additional question.

Thanks.

Le

Finally, I asked whether it was possible for us to be told which applications are accelerated and which are not. The answer was, well, as expected ;-)

Hi Kaitain,

I am very sorry I can not tell you that because it is our trade secret !

Also, this list changes with different product family as well.

Thanks.

Le

From which it can be concluded only that some applications are optimised and some are not, with 3DMark being one of the applications receiving an additional performance boost from the Xp4 hardware.

Update from Spode: Le has told me that this apparrant problem is a software bug in early drivers. Bill tested this and said that the performance was still poor (quality wise). So Le is sending a Laptop our way for us to do some independent testing.


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



Copyright Andrew Miller
Please read our disclaimer

Search the site:

Random

Enermax CS527 Case