This is the first in a series of Platform Comparisons intended to show the effects of various components on the performance of the entire system. It is important to note that these articles are intended to determine which platform is better for business, rather than gaming or overclocking. This article compares the chipset/memory combinations available for use with a low-end Katmai processor (PIII 500) at the time these tests were performed (December ’99). Future installments will compare other platforms, including those supporting Athlon and Coppermine processors at 500MHz, 800MHz and up.
Background and Platform Definitions
A number of publications over the past several months have purported to show the benefits of various chipsets, memory and processors, however, the fact is that in most cases it is impossible to determine exactly what each component contributes to the overall performance. Many publications these days seem to eschew component level (synthetic) benchmarks as being ‘unrealistic’, focusing instead on games and application benchmarks. Yet, these synthetic benchmarks are the only tools available to the average user for isolating components. Unfortunately, even these tools can only provide a limited understanding because of the close interaction between components, requiring careful analysis in order.
As an example, the CPUMark99 score from the Winbench99 benchmark are intended to isolate the processor, biut in reality the numbers are somewhat dependent upon the memory throughput, and the cache performance in particular. Therefore a platform that has better memory performance (either system memory or cache) will skew the results, making the processor appear faster. Of course, one could argue that with both L1 and L2 caches integrated onto the processor package, or even into the chip itself, that this is appropriate. Regardless of whether it is appropriate or not, it is a fact and must be considered when doing testing and analysis on a component level.
One side benefit to this knowledge is that we can use the CPUMark99 scores as ‘circumstantial evidence’ of the memory performance, assuming we can keep the cache speed constant. In addition, by measuring the I/O, memory and graphics performance for each platform separately, we can gain a better understanding of what actually contributes to the performance gain (or loss) on a particular platform in application level benchmarks. Since various applications utilize the system resources differently, this is critical information when selecting a platform that is best for specific uses.
While many publications have focused upon the top-end components to do this type of comparison, I chose what we considered to be the most likely set of components a business user would purchase. The combinations used were Aopen AX6C (i820) and Apacer PC800 DRDRAM, Aopen AX6BC Pro Gold (i440BX) and EMS PC133 HSDRAM, AX63 Pro (Apollo Pro133) with both EMS PC133 SDRAM and NEC 133MHz VCSDRAM. The common components included a PIII 500MHz (Katmai), W.D. 8.4GB UDMA HDD and Diamond Viper V770 Ultra (32MB). Note that all platforms were run with a 100MHz system bus.
In the months since these tests were performed, processors speeds have continued to increase at a dramatic rate, Coppermine processors have become much more readily available and some new chipsets have been released, or soon will be released. Future comparisons will be performed on these platforms as well.
Be the first to discuss this article!