DDR2-1000 tested: Crucial Ballistix and Super Talent

Memory by KeithSuppe @ 2007-05-25

Today we take two popular kits PC2-8000 from Crucial (Ballistix PC2-8000) and Super Talent (PC2-8000 T1000UX2G5) for a spin. Tests will focus on the tightest manual timings at several speed grades across AMD and Intel systems.

  • prev
  • next

Intel Test Results

E6400 - Asus P5W DH (BIOS 1901) Testing:


Madshrimps (c)


Orthos stability under such and such system settings, usually guarantees those settings will pass just about any benchmark. The parameters below were used throughout the Intel phase of our benchmark suite.

  • 8x266FSB = 2136MHz / DDR2 - 800MHz / 2:3-Divisor
  • 8x300FSB = 2403MHz / DDR2 - 1000MHz / 3:5-Divisor
  • 8x266FSB = 2136MHz / DDR2 - 1066MHz / 1:2-Divisor
  • 8x270FSB = 2163MHz / DDR2 - 1080MHz / 1:2-Divisor (Super Talent only)


  • Each kit was tested under the FSB + bootstrap combinations and then at the tightest timings at stock (memory manufacturers) voltage. Crucial Ballistix were not stable beyond 1066MHz on the P5WDH and at that speed were most stable running SPD. Super Talent overclocked slightly higher to 1080MHz running from SPD.

  • Sandra XIandra XI Memory Bandwidth Benchmark - is probably the most widely used to determine Bandwidth. This is a synthetic benchmark and most often inflates real world performance. To the author's credit Bandwidth is a difficult value to measure since it varies from application to application.

    Madshrimps (c)


  • SuperPi 2M / 1M - this benchmark is especially sensitive to speed and the lowest scores were usually held by AMD system owners. Processors such as the Core Duo and AMD's with on-die memory controllers still do better with this benchmark due to it's use of the moth co-processor.

    Madshrimps (c)


  • Everest Ultimate - is a benchmark, diagnostic and monitoring suite which truly livesd up to its name. Unfortunately its freeware/shareware sibling Everest Home Edition has been discontinued to depriving DIY hobbyists of a valuable benchmark suite. Unfortunately Lavalys mis-interpreted the term "valuable" without the qualifier Freeware as in valuable freeware. Politics aside Everest Ultimate / Home Edition features a latency benchmark among its memory tests which is sorely needed among today's freeware benchmarks.

    Madshrimps (c)


  • PCMark05 - is a multi system test which includes HDD and XP start-up, 3D tests, 3D pixel shading, Web Page rendering, 2D graphics memory, multi CPU and system memory tests. The shareware version doesn't allow for scrutiny of specific test results but gives a great all around picture.

    Madshrimps (c)


  • 3DMark2001SE - has been discontinued by Futuremark, although it is still widely used. Ironically for the reason this benchmark is no longer supported is precisely why it serves our purposes. 3DMark2001 was introduced when graphic sub systems albeit add-in card or on-board regularly shared main memory.

    Madshrimps (c)


  • 3DMark06 - has less to do with system memory, however; CPU tests integrated into the benchmark suite reflect main memory performance.

    Madshrimps (c)


  • Cinebench - the rendering element in CineBench exemplify how well the CPU and main memory sub-system work together.

    Madshrimps (c)


  • DOOM3 - Timedemo is a prima facie playback excerpt form the game measuring FPS (Frame Per Second) benchmark, however; DOOM3 places large demands not just on the GPU but the memory sub-system as well.

    Madshrimps (c)


  • FEAR - benchmark is another internal benchmark although it differs from DOOM3 which is essentially a "play back" from one of the game's levels where frame rate is measured. FEAR employs a unique script in which the game's graphic engine is working with lighting and physic's effects as if you were playing the actual game. This provides a much more realistic scenario and more accurately represents how your system would perform during game play.

    Madshrimps (c)


    The benchmark results for Intel show these kits work best at CL4-4-4-12 and 1000Mhz DDR2. I was pleasantly surprised both kits held these timings despite their specified timings of CL5-5-5-15. They also held these timings and speeds without increasing voltage beyond manufacturer's specified 2.2V

    Onto AMD testing ->
    • prev
    • next

    No comments available.

     

    reply