Menu
FSB400: Setting new records with the nForce 2

FSB400: Setting new records with the nForce 2

Storage benchmarks

SiSoft Sandra 2003 Pro

  • SiSoft Sandra provides information about the system in abundance and is also able to check the most important components of the PC for their speed. The so-called 'memory benchmark', which determines the interaction between memory, chipset and processor, turns out to be particularly useful. So it is not just a matter of the memory clock and the maximum memory bandwidth. Since all data is also sent via the processor, northbridge and frontside bus play an equally important role. For example, DDR400 with a bandwidth of 3200MB/s in this test will not exceed its transfer rate of 2700MB/s with a front-side bus of 166MHz, such as the AMD Athlon has so far. Dual Channel DDR with 6.4GB/s shouldn't change anything here either. The SiSoft Sandra Version 2003 Pro was used.
  • Further information: SiSoftware.demon.co.uk
  • Download: ComputerBase.de
SiSoft Sandra 2003 Pro
  • Int Buffered:
    • XP2400 + FSB400/DDR400
      3.071
    • XP2400 + FSB333/DDR333
      2,577
    • XP2400 + FSB333/DDR400
      2.503
    • XP2400 +FSB266/DDR266
      2.048
    • XP2400 + FSB266/DDR333
      2.029
  • Float Buffered:
    • XP2400 + FSB400/DDR400
      2.870
    • XP2400 + FSB333/DDR333
      2.445
    • XP2400 + FSB333/DDR400
      2,364
    • XP2400 + FSB266/DDR266
      1,977
    • XP2400 + FSB266/DDR333
      1.956

Such a three-way division was certainly foreseeable, but interesting trends can already be deduced here. On the one hand, the FSB400/DDR400 with almost 3100MB/s in the Int Buffered Test comes quite close to the maximum possible 3200MB/s. The fact that no higher values ​​are achieved here, although Dual DDR400 could theoretically deliver 6.4GB/s, is due to the bandwidth of the FSB400, which is limited to 3.2GB/s. The clock rates with an FSB of 166MHz are a good 500MB/s behind the leader. It seems interesting here that the asynchronous operation is clearly behind the synchronous one. The same picture also emerges at 133MHz FSB.

Cachemem

  • Cachemem seems to be made for exploring the memory latencies (delays). In addition to the pure bandwidth, these represent the most important factors for the speed of a system and are mostly decisive for the final ranking of the motherboards.

Cachemem
  • Latency times:
    • XP2400 + FSB400/DDR400
      215
    • XP2400 + FSB333/DDR333
      220
    • XP2400 + FSB333/DDR400
      276
    • XP2400 +FSB266/DDR266
      283
    • XP2400 + FSB266/DDR333
      334

Cachemem seems to present the cause for the failure of the asynchronous clocks in Sandra. The enormous administrative effort involved in this operating mode has a massive impact on memory latencies. FSB400/DDR400 is once again at the top, closely followed by FSB333/DDR333. The asynchronous cycle ratios and FSB266 follow clearly behind. If you give your Athlon XP (FSB266) a synchronous memory, you save a good 18% delay - even if DDR333 might have sounded promising. It's definitely not worth it! Just like the FSB333/DDR400 operation.

On the next page: Games