LOSTCIRCUITS

Conclusions

What was intended here as a standalone review of the nVidia Quadro FX3000 has more or less evolved into a comparative analysis of the performance of different cards from different manufacturers and models from ATI to Wildcat and from current street pricing of $400.- up to whopping $1,700.-. Needless to say that every card included here has its own areas of excellence, there are benchmarks where the Wildcat VP990 prevails, others are taken by the FireGL X1 - - but the majority still goes to the FX3000 and there is not even a shadow of a doubt about that.

There is no shadow of a doubt about the price point either but overall, our benchmarks show that there is still some truth to the old saying you get what you pay for, meaning that if you want top performance, you will need to shell out some top $$$.

A separate issue that only emanated during the later stages of this review, specifically the data analysis is once again the inadequacy of the benchmarking procedures. Part of this we need to throw back into the ballpark of the graphics card manufacturers since there is, unfortunately, very little support with respect to relevant benchmarking software.

Relevant in this case encompasses two separate issues, the first being the fact that the performance measured is, in fact, reflecting some real life applications or requirements on graphics cards. The second issue here is that the results actually make sense and this is where we came to some gripes with SPECapc's in their various iterations since the scores generated were far from a realistic reflection of the benchmarked system configuration.

Admittedly, we are only looking at the composite scores but the mere fact that one and the same CPU can achieve a score of either 3.24 or else 10.5 doing the exact same work casts some serious doubts with respect to the validity of the overall result metrics. Moreover, it stresses the point that it is not possible to use these CPU scores for any system or processor evaluation if there is even a single setting that is changed, not to mention a variation within the hardware configuration.

Similarly, SPECapc for Maya appears to be geared more towards the overall system performance than towards evaluation of graphics adapters, there are nVidia scores, and there are ATI scores and if we still had access to the Wildcat, there would be 3DLabs scores as well, demonstrating something that we don't know what it is but certainly not the fundamental performance differences between the individual graphics cards.

A third issue we have been running into with this and the last few reviews has been the question of what the rendered output should look like. The case in point is the 8spot application in 3ds max and unfortunately, there are no references provided, meaning that it took a flood of emails and phone-calls to get to the bottom of this. The problematic here is that the differences were so obvious that one would have to be blind to not see them but what about more subtle changes? On a side note, the differences between the speed and the quality settings in MAXtreme were noticeable and on a scale of 1-10 from the speed to the quality settings, the FireGL X1 would be somewhere between 4 and 7.

In conclusion and despite all the confusion surrounding the different drivers and benchmarks, I believe we have still been able to paint some relatively accurate picture of the different graphics cards that have been running in this review and their performance, and how the rankings change depending on the different applications and that is all we tried to convey. In other words, the trade-off in performance compared to the trade-off in price should allow some judgement as to what is an appropriate solution for the one or the other customer. Keep in mind also that we have been looking at a very selective sample of professional graphics cards, encompassing the top and the obsolete end of the Quadro line and some upper middle class ATI, spiced up with the 3DLabs top model. There are more cards out there and they are difficult to get and almost impossible to put into a comprehensive perspective but this article should provide some orientation and a frame work for some basic understanding of what the general issues are.

Next Page:    => More =>