Duron 950 to Athlon XP 1.7+ in the test: AMD processor Roundup
- 1 Foreword
- 2 Overview
- 3 Technology
- True Performance Initiative
- 6 Requirement
- 7 Overclocking
- 8 Known bugs
- 9 Test system
- 19 Pricing
- 20 Conclusion
Since AMD also states the 'extremely fast boot' of the computer as one of ten reasons for buying an Athlon XP, we naturally wanted to get to the bottom of this statement. For this purpose we have organized the Bootvis program from Microsoft. This program was actually developed to optimize the boot process and for this purpose it analyzes the system start up to the smallest detail. Among other things, the boot activity, CPU usage, the hard disk input/output, the driver delay and much more are examined. With the boot time in general, it has been shown that this is hardly dependent on the clock frequency of the processor becauseusually other components delay the system start unnecessarily. While the boot time on the Athlon XP 1700+ averaged 31.87 seconds, a Duron 950 (Spitfire), for example, booted up in 31.86 seconds. Of course, these values vary from case to case by up to two seconds, so that in summary one can say that the system start is not significantly accelerated by the Athlon XP.
Before we let the ranks of a total of 12 processors compete against each other in real applications, we would like to take a closer look at the theoretical performance values of the bolides . We used Sandra for this purpose, as both the 3DNow! as well as the SSE1 expansion of the processors is correctly recognized and used accordingly.
Sandra Processor Test
Sandra 2001te offers two benchmarks that are exclusively the Determine the performance of the processor. On the one hand, the Dhrystone benchmark is used, which was originally developed by Siemens to measure the performance of the main processor. On the other hand, the performance of the co-processor is determined via the Whetstone benchmark. Both tests are carried out without taking the extended multimedia instruction sets into account. This result is also confirmed when using floating point numbers.
As you can see quite clearly, the processor speed is primarily responsible for the processor performance at this point. The size of the L2 cache, which is known to be much smaller in the Duron, does not matter. However, the enlargement of the translation lookaside buffer (TLBs) and the hardware prefetch seem to be at this pointhaving no influence. The Whetstone co-processor test illustrates this result again.
Sandra multimedia test
In Sandra 2001te's multimedia test, an algorithm (chaos theory by Mandelbrot), which is used, among other things, to generate realistic natural objects such as mountains or clouds. This benchmark also takes into account the extended instruction sets of the Pentium 4 or the Athlon XP and the new Duron.
While the old Duron and Athlon with Thunderbird core with the 3DNow! Extensions were controlled, the SSE extension was used for the Athlon XP and the Duron with Morgan Kern. As you can see, the multimedia test with whole numbers (integer data) benefits slightly from this.
Sandra memory test
Sandra's memory test occupies at least 50 percent of the available memory. This benchmark determines the performance of the storage subsystem as well as the caches. Both arithmetic and floating point operations are carried out for this.
As you can see here, the new cores do extremely well in this test . Since this test does not use the MMX, 3DNow! orSSE extension is made, this result is based solely on data prefetching and the improved translation lookaside buffers. Using floating point information, in which the internal processor cache plays a larger role, the Athlon 'C' can outperform the Duron with Morgan core again due to the higher front-side bus of 133MHz and the larger L2 cache. It should be noted, however, that the results fluctuate within a certain range, as always. In general, however, a staircase effect can be seen between the various processor cores.
On the next page: 3DMark 2000