| To: Jozef Halada who wrote (24344) | 1/9/2001 7:49:33 PM |
| From: Jozef Halada | Read Replies (3) of 264558 |
| |
From Microprocessor watch e-mail newsletter:
*** Athlon Edges Out Pentium 4
Peter N. Glaskowsky {1/08/01-01}
Our first hands-on tests of Intel's 1.5GHz Pentium 4 and AMD's 1.2GHz Athlon have borne out the predictions we made four months ago based on Intel's preview of the P4's microarchitecture. The P4's higher clock speed gives it impressive peak throughput on selected tests, but overall, on a variety of tests, it comes in slightly behind its primary competitor.
Our AMD Athlon test system-configured with 133MHz double-data-rate SDRAM-demonstrated excellent performance on both productivity software and multimedia content-creation applications. Despite a 20% deficit in raw CPU clock speed, the Athlon system was 5.3% to 12% faster than the RDRAM-equipped Pentium 4 system on several industry-standard application-based benchmark suites.
The results of 3D-graphics benchmark tests were mixed. The Athlon-1.2 system was slower than the P4-1.5 system when running Quake 3 Arena and the processor tests in Mad Onion's 3DMark 2000 v.1.1 and Ziff-Davis Media's 3D WinBench 2000 v1.1. The Athlon system was slightly faster than the Pentium 4 machine, however, on the rendering portions of those benchmarks.
It is generally assumed that the Athlon architecture is much more efficient at executing highly conditional code, and that Pentium 4 is more efficient when performing parallelizable tasks on large data sets. The reality revealed by our testing is more complex than these generalities. We are left with the classic benchmarking paradox. If we summarize a processor's performance using a composite result, we gloss over the consequences of significant microarchitectural decisions. Any single test component, however, has little predictive value for as-yet-untested applications. We can say that today, AMD's Athlon at 1.2GHz is generally faster than the 1.5GHz Pentium 4, but this is not a very strong statement.
Over time, the Pentium 4's NetBurst microarchitecture should have an advantage as compiler and software developers adapt to its unusual characteristics. It is likely to take a long time for mainstream applications to reflect the results of this adaptation process. Many such applications have yet to benefit from Intel's MMX extensions, which are now some five years old. Certain applications and many device drivers, on the other hand, will improve quickly over the next year as small, well-understood sections of code are made P4-aware.
In the meantime, Intel must price the Pentium 4 according to the value it delivers to customers. P4 prices are already unusually low for a new Intel processor, especially one as large and expensive as the P4. This situation can't be the outcome the company envisioned when it began the Pentium 4 project, but until Intel can offer higher clock speeds, further architectural enhancements, and lower system costs, it must make the best of the situation.
Microprocessor Report readers can read the full story on this controversial subject at:
http://www.mdronline.com/mpr/h/2001/0108/150201.html |
