LOSTCIRCUITS

Platform Comparisons

I raised this point earlier but one problem with this kind of review is that the only thing that can really be shown in terms of performance are frame rates. There is a very limited possibility to show quality or rather "precision" of the real-time rendering in the form of dropped pixels or transparencies instead of textured surfaces that can make a huge difference in the appearance of games or CAD applications as rather eloquently pointed out by Johan but there is hardly a way of quantifying these issues. Because of these issues we need to rely on fps and those are important for playback even if they may be irrelevant for earlier stages within a design process. Now the only thing to keep in mind is that in this specific benchmarking environment CPU power is a rather lost cause. The GPU is taking over a lot of the workload and the CPU only acts as a "data server", limited by the system memory bandwidth.

ViewPerf 6.1.2

Let's start with the second to last version of SPEC Viewperf, revision 6.1.2 consisting of Advanced Visualizer-04, Design Review-07, Data Explorer-06, Light-04 MedMCAD-01 and ProCRDS-03. Please make sure when comparing the benchmark results to results posted elswhere that the revision of the individual applications is the same since there are huge differences in the overall complexity of the application between revisions.

AWadvs-04

FPS comparison between two Athlon and two P4 platforms. As mentioned earlier, the FireGL 8800 thrives on memory bandwidth and AWadvs-04 is no exception to the rule. Note that the dual AMD system falls behind the single Athlon on a VIA KT333 platform.

DRV-07

In Design Review-07 (a CAD model of an oil-rigging platform) the two P4 systems get the better of the two Athlon systems. Regardless of memory bandwidth differences between the A7V333 and the A7M266D, the two systems score roughly the same.

DX-06

Data Explorer-06 visualizes a set of particle traces through a vector flow field. The width of each tube represents the magnitude of the velocity vector at that location. Data such as this might result from simulations of fluid flow through a constriction. The benchmarks show a very similar picture as DRV-07, only that the Dual Athlon system falls behind even further. Keep in mind that these applications are not optimized for SMP operation, even though there is a limited amount of dual processing possible.

Next Page:    => Part Deux =>