Clearifying DDR2 performance



For a final comparison, I've morphed results from both systems at 800MHz and 1000MHz from two primary benchmarks. As stated in our intro comparing data across platforms is inherently difficult due to number of variables. Therefore benchmarks were chosen which tend to isolate memory and memory sub-systems. Sandra XI Memory Bandwidth while artificial does a decent job at predicting bandwidth. SuperPi taxes the CPU math co-processor, memory and in some cases system stability above 4M, we chose 1M/2M since we’re more interested in determining system speed. Memory latencies along with CPU and memory frequency will have a measurable impact on SuperPi calculations. This information is universally utilized to compare all types of desktop systems. We begin with Sandra below.



In the area of Bandwidth the Opteron trounces Core Duo; this is where the on-die memory controller still show its strengths. Unfortunately that seems to be the only area and adding insult to injury as AMD processors are basically trapped by virtue of their 200MHz FSB, which then requires one to overclock their AMD device to a 266MHz Reference Clock in order to keep pace with Conroe. .

SuperPi circumvents the graphic system focusing on the processor and memory sub systems, this is where we see the AMD's once un-beatable on-die memory controller showing its age.



Overclocking results for these two kits favor Intel Core Duo only if you’re strictly planning to run your system at 1066MHz which is easily attained by overclocking each kit at default voltages. If however you’re planning on reaching the upper echelons of overclocking these kits are better suited to AM2 systems. There's no doubt their default speed of 1000MHz does "fit" well with the 200MHz Reference Clock speed. My Opteron 1212 based system still packs quite a performance punch when overclocked to 3.0GHz from its 2.0GHz reference speed. It's certainly a little more labor intensive to overclock an AM2 system, if the Planets do not align just so it's BSOD time. Intel systems are much easier to deal with.

Conclusion:

Both kits tested today mark a departure from DDR2 speeds running CL3 and even CL4. Recently Bearlake made its 1333FSB debute. The chart results are prima facie insofar as which system out-performs the other. To get the most out of this review it's important to keep the focus on each memory as it performs on a specific platform. Testing reveals both kits to be excellent performers and while overclocking was somewhat limited, I could have easily received two kits which overclocked much higher, as we know this aspect is difficult to predict since RAM is unlike a CPU in terms of identifying a specific stepping known to be a decent overclocker. While each kit is rated to 1000MHz both easily reached 1066MHz speed and at tighter timings then advertised at their default speed. While Crucial Ballistix had some problems this was the first kit I've had from Crucial which was a relatively poor overclocker.

Nonetheless the kits both ran at tighter timings which does boost performance somewhat. Perhaps most important there was no need to increase DDR2 voltage and doing so usually had a negative effect.

Crucial Ballistix PC2-8000
Pros
+ Excellent value - $119 Tiger Direct (rebate)
+ Versatile, AMD or Intel platforms
+ Holds 800MHz CL3, 1000MHz CL4 at default voltage
+ Runs at 1066MHz SPD default voltage

Cons
- None for the price


Supert Talent PC2-8000
Pros
+ Excellent value - $137 Compuvest.US
+ Versatile, AMD or Intel platforms
+ Holds 800MHz CL3, 1000MHz ~ 1066MHz CL4 at default voltage
+ Strong overclocker

Cons
- None for the price

I would like to thank both Crucial and Super Talent for their support. This review took much longer the expected due to an illness in the family, I am grateful for their patience.