Conclusion

So, there we have the R420. This part is certainly not revolutionary, it's hardly even evolutionary, but more of a refinement on R300’s weak points, placed on a more advanced process with double the number of pipelines and effectively we have an R300 on steroids! This speaks volumes for R300’s initial design in that it can be scaled to such a degree that it can offer well in excess of three times the performance in many cases of its original incarnation.

While the scalability speaks well of R300’s initial design, you might think the lack of Shader Model 3.0 support doesn’t speak so highly of ATI – seeing as the Shader 3.0 specification had resided, unchanged, in DirectX9 from its introduction you’d probably be forgiven for thinking it would be supported by ATI with R420. From the software perspective ATI’s position is that the vast majority of developers are hardly pushing the limits of Shader 2.0 yet, not for titles in 2004, and even many developers who are working to much longer timescales now, and hence they feel there is still a fair bit of headroom in both what can be done with Shader 2.0 and educating developers on shaders in general. Should developers be providing Shader 3.0 code in the short term, ATI feel confident that their ISV team will be able to provide Shader 2.0 alternatives that have the same output and without loss of performance – so they say.

However, given the choice they probably would have wanted to support Shader 3.0 now, but quite evidently there were some business drivers that suggested they couldn’t – are these drivers down to lack of resources, or simply feeling that what they believe to be true Shader 3.0 support can’t be implemented in current processes? We’ll let ATI’s new CEO, Dave Orton answer that one. However, one thing is clear is that ATI have set up a lot of their engineering resources to be focused on the PCI-Express transition such that they hit the market with an entire line of PCI-Express solutions as soon as PCI-Express is ready to be released.

Looking at the performances of the Radeon X800 boards, in the case of the X800 PRO we can see that even had the X800 XT Platinum Edition not been here the performance increase would have been about what we would see from many previous next generation / refresh launches. In some gaming instances, at least those that are quite intensive, performance increases of beyond 70% are offered and in some shader tests over double the performance can be seen, which bodes well for more shader intensive titles. So. While this may make a respectable upgrade from even a 9800 XT, it would certainly make a huge upgrade in term of for those who are still clinging on to their 9700 PRO’s as the X800 PRO doubles, or even greater, the performance of this board in a wider range of cases.

One thing that you may have noted with the X800 PRO, though, is that the theoretical performance difference to the X800 XT PE is lower than the performance difference seen in some of these tests – indeed, in one or two cases it wasn’t able to show an improvement over the 9800 XT. ATI can map different memory timings within the driver to different modes of operation (for instance, whether FSAA is enabled or not) and understanding the optimal setting for a new chip and board configuration can take time. Obviously the lions share of the understanding of the optimal modes have gone into the high end X800 XT PE and ATI still have to spend more time understanding how the 3 quad operation of the X800 PRO best maps to the memory interface. However, as time goes on ATI will be able to work on these timings more and in a number of cases the performance should move closer to the X800 XT PE and further away from the 9800 XT.

As for the X800 XT Platinum Edition, ATI have managed to repeat the performance improvement that they had with Radeon 9700 by more than doubling the performance of their previous generation. While limitations in the software prevent them from showing that performance differential in many cases, and lack of higher speed memory prevents it with FSAA, it is illustrated in many of the shader tests. Hopefully performance increases like these will spur developers on to have the confidence in offering more immersive, shader intensive titles than there currently are today.

One thing that we also see with both of these boards in a number of cases is that the performance gains over the previous generation are often greater than the shear memory bandwidth increases would suggest. R300’s memory bus was fairly efficient, but relatively slow so one of the key tasks for R420 was to increase the speed, however it also needs to maintain and increase the efficiency. The performances appear to bear out that this has occurred.

Whilst the lack of Shader 3.0 is somewhat disappointing, regardless of how much difference it will make in the short term, the performance to power utilisation is less so. To double performances from the previous generation whilst reducing power consumption is fairly impressive. This also keeps the board to single slot solution which should please Small Form Factor PC owner that still want high end gaming performance.

We are of the understanding that Radeon X800 PRO boards are in production now, and could be shipping this week. Radeon X800 XT Platinum Edition boards are presently scheduled to ship two weeks after the PRO.

While this review has been looking at things very much from the isolated view of what ATI has managed to do relative to its previous offerings, how this equates in a competitive environment is a whole different kettle of fish. Hopefully we’ll soon be able to compare and contrast the two architectures of both ATI and NVIDIA’s latest offerings to see where both their relative strong and weak points may lie.