Corsair TWINX1024-4400C25PT DDR Memory Review

Memory by JNav89GT @ 2004-11-23

Corsair?s second foray into Samsung TCCD modules brings about higher levels of performance onto the DDR scene. Validating the ram for both Intel and AMD Athlon 64 platforms at timings of Cas 2.5 4:4:8 with 2.75v, there is certainly the promise of memory Valhalla.

  • prev
  • next

The belly of the beast!

Test system setup :

JNav89GT's Test Setup
CPU AMD Athlon FX55 S939
Mainboard MSI K8N Neo2 Platinum
Memory Corsair TwinX 1024-4400C25
Video EVga Geforce 6800 Ultra (overclocked to Core:425/Memory:1200)
Cooling Modified Prometia Mach1 R402a by Chilly1

Benchmarks Used:

  • Futuremark PCMark 2004
  • Futuremark 3DMark 2001SE(build 330)
  • Super Pi 1MB Test-Lower is better
  • Cachemem
  • Sciencemark 2.0 Beta
  • Doom3 640x480 Low Detail(default low details, no optimization otherwise)
  • SiSoft Sandra Memory Bandwidth-Buffered

    BIOS/System settings and remarks:

    Memory timings:
    200MHz Cas 2 2:2:5 2.5v
    275MHz Cas 2.5 3:3:7 2.8v
    275MHz Cas 2.5 4:4:8 2.8v
    285MHz Cas 2.5 4:4:8 2.8v

    HT Multiplier:
    5x for testing at 200MHz HTT, 3x for testing at 275MHz and 285MHz HTT

    CPU: Multiplier:
    15x for 200MHz HTT Multiplier.
    11x for 275 and 285MHz HTT

    CPU Final Speed:
    3GHz for 200MHz HTT
    3025MHz for 275MHz HTT
    3135MHz for 285MHz HTT

    I attempted to stratify CPU speeds between 200 and 275MHz HTT speeds. At 285MHz HTT speeds I could not stratify CPU speed to stay close to 3GHz so I choose to retain the 11x multiplier to show potential at higher CPU MHz and memory speeds. The system was somewhat unstable at this greater speed though, so please use these scores as a reference only. In addition, due to the instability at the higher CPU and HTT/Memory speeds 3DMark 2001SE would not complete a full benchmarking run, as it would continually error out in Lobby High.

    Prior to formally reviewing the ram I was able to "play" with the memory for approximately 3 weeks. I can say that I had no problems at all with the memory that were not self induced or in other words, overclocked too far. After my CPU and platform related experiences that resulted in a rather steep learning curve in relation to possible limiting factors on HTT and memory speed scaling with my G.Skill PC4400 Memory Review, I was better prepared with this memory and had far less headaches along the way.

    In addition, I had a brand spanking new Athlon FX55 that I was using in a modified Nventiv (ex-Prometeia) Vaporcooling setup. This CPU was willing to travel to the heights of 275-280 MHz HTT/Memory MHz which I used for system testing with an 11x CPU multiplier to produce a final clock speed just over 3GHz. Much past that and I began to experience limitations. Specifically, which parts were limiting me I am unsure of. The CPU was good for a few more MHz but I cannot confirm that I could go higher than 280 MHz on the memory at lower CPU multipliers. Logic would have necessitated I test lower multiplier and higher HTT/Memory speeds. However, ego prevailed in this case, and I could not bring myself to lower the clock speed of the CPU below 3 GHz, having successfully achieved a longstanding goal of mine. I plan to VDD mod my motherboard in the coming weeks and I will update this article should this provide expanded headroom on the memory.

    Subjectively, the ram performed well and up to my lofty expectations. Fit and finish of the modules were up to Corsair's high standards, and the ram looks nice tucked away amongst the other parts in my PC. Furthermore, I was pleasantly surprised that the Corsair modules continued a tradition of exceeding rated performance levels. I was able to run the Corsair TwinX 1024-4400C25 modules at a more aggressive Cas 2.5 3:3:7 versus rated timings of Cas 2.5 4:4:8 at 275 MHz memory speeds. I did need a slight bump in vdimm from the rated 2.75 up to 2.80v. However, this could also be attributed to production variance in motherboards and Corsair would still warranty these modules up to 2.85v. Past 2.80v on the vdimm I experienced no further improvement, and actually 2.85v produced declining results in clocking of the memory modules. This was my specific experience and I would advise that each person experiment with different voltages to produce the best results for each independent platform.

    Madshrimps (c)
    System ready to go!

    So after a relatively uneventful "getting acquainted" period, I was set to let the benches begin and see what this setup could do. Please keep in mind that the Athlon64 with its integrated memory controller will experience increased memory performance and bandwidth with higher clock speeds. Since we are using AMD's flagship CPU at a rather high MHz level, please do not try to generalize the results reported here to those you might experience with a slower speed CPU. In addition, the results reported here cannot be compared to other memory reviews, even my G.Skill article, again due to variations in platform, and more specifically CPU speed.
    • prev
    • next