SuperPi 32M

32M is one of those benches one could waste an eternity in trying out various tweaks on both the board and in the operating systems. We decided to keep things short and simple for this review by optimizing the basic memory options and Bclk rates to show differences in how each board sets up sub-timings using the same processor and memory clocks on a given multiplier. All the boards managed the same 22x 234 BCLK using the 2:8 memory ratio.

ASUS trails by virtue of a very loose set of default sub timings. Unfortunately, we lost our ASUS ROG board to the Foxconn socket meltdown problem at this point. This also damaged our best CPU in the process so were unable to push further in this particular benchmark as our replacement processors did not clock as well.

The EVGA E659 did not fare as well as the other boards in this benchmark using the 2:10 memory ratio. No matter what we tried we could not get the board to pass 32M any higher than 226~227 BCLK in this bench. Moving over to the 2:8 ratio the E659 managed to out clock the other boards for raw CPU frequency. However, raw memory bandwidth and tight latencies favor this benchmark over the small differences in CPU frequency in our results so the lack of stability with the 2:10 memory ratio is adversely affecting the E659.


Super Pi 32m - Max CPU Clock
WPrime 1024M What is RTL?
Comments Locked

52 Comments

View All Comments

  • SpaceRanger - Friday, November 6, 2009 - link

    [quote]Gigabyte’s top end offering comes in at $*** and touts their answer to competing boards as having ’24-Phase’ PWM circuit (using a multiplexed ISL6336 6 phase controller).[/quote]

    So all it takes is 3 stars to get the top end offering from Gigabyte?
  • Rajinder Gill - Friday, November 6, 2009 - link

    Sorry, that skipped through edit - corrected.

Log in

Don't have an account? Sign up now