Our Duron Review (Score 3)

We've put up a review of Duron at AMD Zone, the largest AMD site on the net. We also hit a few of your problems (less pages, more paragraphs) and use more benchmarks. For example, Currently, there are five standard SPECopc viewsets: ProCDRS viewset is intended to model the graphics performance of Parametric Technology Corporation's CDRS industrial design software. IBM's Data Explorer (DX), which has 10 different tests, is a visualization application. Intergraph's DesignReview (DRV), which has 10 different tests, is a 3D computer model review package. Alias/Wavefront's Advanced Visualizer, with 10 tests, is an animation application. And Discreet Logic's Lightscape Visualization System, with four tests, is a radiosity visualization application. All five viewsets represent relatively high-end applications. These type of applications typically render large data sets. They almost always include lighting, smooth shading, blending, line antialiasing, z-buffering, and some texture mapping. So we use Specview, Seti, Distributed, and a few other benchmarks made by a friend of ours who attends the University of Illnois. QMC is best described in the words of Tim QMC tests cpu power, cache quality to some extent depending on the file Input1D0, bus io and ram. It calculates the ground state properties of a particular lattice model using a Quantum Monte Carlo algorithm. Linpack is a very interesting benchmark for comparing the Tbird to the classic Athlon because you can watch the data get too big for the CPUs caches and watch performance fall. Here is Tim's description of the benchmark. For a matrix of size N you use 8 * N * N bytes to store that matrix. For 25 sites you use 5KB. Each time the matrix is increased by 25 the memory consumption rises by 4 fold. So you use 5KB, 20KB, 80KB, .... , 5.12 MB. At some point the L2 is overflowed. At that point performance will depend on how quick data can be accessed from ram. This is where you will see a noticable performance drop, or at least you should depending on the latency of getting data from ram. So for matrices that are larger than L1(data) + L2 you are really stressing the quality of the cache and the bus speed. Also you pay a tremendous performance hit for using CAS3 ram. I think it's a very good benchmark. Also to those who commented that Athlon is about the same as Duron's price. That is off Price Watch, which is not the MSRP which was released as always in press releases, and was OEM. Duron will cost much less at Price Watch when it is readily available than its MSRP just like Athlon does now. Hope that helped.