Menu
Pentium 4 with FSB 533 MHz in the test: Intel sets new standards

Pentium 4 with FSB 533 MHz in the test: Intel sets new standards

Table of contents
  1. 1 Foreword
  2. 2 Technology
  3. 3 Bandwidth
  4. 4 Power consumption
  5. 5 Requirements
  6. Overclocking
  7. 6 Test system
  8. Benchmarks
    1. 7 Boottime
    2. Sandra 2002
    3. 8 PCMark 2002
    4. 3DMark 2001SE
    5. 9 GLMark
    6. 10 Quake 3 Arena
    7. 11 Sysmark 2002
    8. 12 Seti @ Home
    9. WinACE
    10. 13 FlaskMPEG
    11. Lame
    12. 14 Cinema 4D
    13. 15 ViewPref
    14. 16 Lightwave
  9. 17 Pricing
  10. 18 Conclusion

Boottime

The Bootvis program developed by Microsoft was used to measure the boot speed. 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, hard disk input/output, driver delay and much more are examined. As our experience has shown, the boot times between the individual processors on a platform differ only slightly from one another. For this reason we are content this time with two representative measurements. Bootvis determined a boot time of around 10.4 seconds with the 2.4 GHz Pentium 4 (533MHz). Our AMD test system with Athlon XP 2100+ needed 12.2 seconds for the start procedure and was thus prepared for work with a slight delay. However, since the two test platforms differ more than significantly from each other and especially thatMainboard plays an important role during the boot process, the result should not be transferred to any Pentium 4 or Athlon XP system.

Sandra 2002

Before we let the ranks of 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 !, SSE1 and the SSE2 expansion of the processors are correctly recognized and used accordingly.

Sandra Processor Test Sandra 2002 offers two benchmarks that are only intended to 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 considering the extended multimedia instruction sets. In addition, the SSE2 command extension was used in the Whetstone test.

Sandra 2002 - Dhrystone
Unit: MIPS
    • P4 2.53/533 - PC800
      4,895
    • XP 2100+ - PC266
      4,800
    • XP 2000+ - PC266 *
      4,642
    • P4 2.40/533 - PC800
      4,549
    • P4 2.40/400 - PC800
      4,513
    • XP 1900+ - PC266 *
      4,456
    • P4 2.20/400 - PC800 *
      4.211
    • P4 2.0A/400 - PC800 *
      3.723

Since the memory bandwidth in this test was not ais of outstanding importance, we have decided not to publish our test results with the fast PC1066 memory. In terms of the measurement inaccuracy, it makes no difference whether the PC800 or PC1066 Rambus is used. As we can see, the Athlon XP 2100+ and its predecessor cut a really good figure in this purely theoretical test. Not even 100 points separate the processor from the 800 MHz higher clocked opponent.

Sandra 2002 - Whetstone
Unit: MFLOPS
    • P4 2.53/533 - PC800 SSE2
      3.076
    • P4 2.40/533 - PC800 SSE2
      2,916
    • P4 2.40/400 - PC800 SSE2
      2,859
    • P4 2.20/400 - PC800 SSE2 *
      2,706
    • P4 2.0A/400 - PC800 SSE2 *
      2.457
    • XP 2100+ - PC266
      2.403
    • XP 2000+ - PC266 *
      2.320
    • XP 1900+ - PC266 *
      2.228
    • P4 2.53/533 - PC800
      1.316
    • P4 2.40/400 - PC800
      1.250
  • P4 2.40/533 - PC800
    1.247
  • P4 2.20/400 - PC800 *
    1.143
  • P4 2.0A/400 - PC800 *
    1.039
  • In the Sandra processor test Whetstone, this time again there was an ambiguous result. If the Pentium 4 processors are addressed with their SSE2 extension, they cannot be stopped by any AMD processor for the time being. However, as soon as the benchmark is forced to use SSE1, theThe performance of the Pentium 4 processors is not really convincing. This benchmark is only suitable for the Pentium 4 architecture if the SSE2 instruction set is also used.

    Sandra multimedia test In Sandra 2002's multimedia test, an algorithm is used that is also used to generate realistic natural objects such as mountains or clouds. We are talking about the chaos theory put forward by Mandelbrot. The extended instruction sets of the Pentium 4 or the Athlon XP are also taken into account in this benchmark. Since the implementation of SSE1 is better than that of 3DNow! is, we let the Athlon XP work with its SSE1 unit (i.e. 3DNow! Professional). In the Pentium 4, however, both SSE1 and SSE2 were used.

    Sandra 2002 - Integer
    • SSE1 and SSE2 optimization:
      • P4 2.53/533 - PC800 SSE2
        10.022
      • XP 2100+ - PC266
        9.530
      • P4 2.40/533 - PC800 SSE2
        9.491
      • XP 2000+ - PC266 *
        9.207
      • P4 2.40/400 - PC800 SSE2
        9.059
      • XP 1900+ - PC266 *
        8,839
      • P4 2.20/400 - PC800 SSE2 *
        8,692
      • P4 2.53/533 - PC800
        8.518
      • P4 2.40/400 - PC800
        8.087
      • P4 2.40/533 - PC800
        8.067
      • P4 2.0A/400 - PC800 SSE2 *
        7.913
      • P4 2.20/400 -PC800 *
        7.396
      • P4 2.0A/400 - PC800 *
        6.721

    In Sandra 2002's multimedia benchmark, the Pentium 4 with 2.53 GHz can call first place its own, but the gap to the Athon XP 2100+ is not particularly great. Here too, full performance can only be achieved with activated SSE2 command extension. What is particularly interesting about this test is the fact that the new 2.4 GHz Pentium 4 achieves 400 points more than its counterpart with a front-side bus that is clocked at the same time thanks to the system bus bandwidth increased to 4.2 GB/s of 400 MHz. However, the difference in performance is only 4.8 percent. The 2.53 GHz Pentium 4, which clocks 5.4 percent higher than the 2.4 GHz Pentium 4, can convert its 'more' clock rate into 5.6 percent more performance.

    Sandra 2002 - Floating Point
    • SSE1 and SSE2 optimization:
      • P4 2.53/533 - PC800
        12.422
      • P4 2.53/533 - PC800 SSE2
        12.209
      • P4 2.40/400 - PC800
        11,789
      • P4 2.40/533 - PC800
        11,759
      • P4 2.40/400 - PC800 SSE2
        11,594
      • P4 2.40/533 - PC800 SSE2
        11.564
      • XP 2100+ - PC266
        11.085
      • P4 2.20/400 - PC800 *
        10.784
      • XP 2000+ - PC266 *
        10,666
      • P4 2.20/400 - PC800 SSE2 *
        10,597
      • XP 1900+ -PC266 *
        10.238
      • P4 2.0A/400 - PC800 *
        9,807
      • P4 2.0A/400 - PC800 SSE2 *
        9.625

    During the multimedia test The Pentium 4 actually looks pretty good with floating point numbers. However, one has to admit without envy that the Athlon XP 2100+ can pass this test extremely well. It even ranks just ahead of the 2.2 GHz Pentium.

    Sandra memory test Sandra's memory test occupies at least 50 percent of the available RAM. This benchmark determines the performance of the storage subsystem as well as the caches. Both arithmetic and floating point operations are performed for this purpose. Since this test is very platform-dependent, let's just take a quick look at it.

    Sandra 2002 - RAM
    • Integer Data:
      • P4 2.53/533 - PC1066
        3.312
      • P4 2.40/533 - PC1066
        3.304
      • P4 2.53/533 - PC800
        2.802
      • P4 2.40/533 - PC800
        2.764
      • P4 2.40/400 - PC800
        2,491
      • XP 2100+ - PC266
        1.980
    • Floating Point:
      • P4 2.53/533 - PC1066
        3.312
      • P4 2.40/533 - PC1066
        3,303
      • P4 2.53/533 - PC800
        2.796
      • P4 2.40/533 - PC800
        2.764
      • P4 2.40/400 -PC800
        2,489
      • XP 2100+ - PC266
        1.900

    The Sandra 2002 memory test the Pentium 4 models that could be operated with the PC1066 Rambus clearly win. Of the theoretically possible 4.2 GB/s, the memory was able to send at least 3.3 GB/s through the lines. The Athlon XP with its PC266 memory really looked beaten in this test. This result in particular is responsible for the poor performance of the Athlon XP compared to the Pentium 4 in one area or another. However, you should be careful not to blame the memory performance for all bad results.

    On the next page: PCMark 2002

    Comments