Setup:
OS: Windows XP


Testing:

All results are dual 32M SuperPI stable. This is a rough test of stability, and is by no means guaranteed stable for 24/7 use. Multiple tests were performed, with several applications that were mainly affected by the ram.

Everest Ultimate 3.00.630 read and write tests - Everest is quite common amongst bandwidth junkies; it gives the option to do read and write tests for the ram and measures the results in MB/S.

SiSoft Sandra Lite 2007.6.11 INT and Float bandwidth tests - Another great program used to measure bandwidth. (INT and FLOAT, both measured in MB/S)

CrystalMark 2004R2 - CrystalMark 2004R2 is a program which is used to benchmark memory and several other computer components. It was developed by "hiyohiyo", an enthusiast in Japan, and is widely used amongst several benchmarking forums.


Cinebench R10 - Cinebench R10 is a real-world application. It is based on MAXCON's software, which has been used in movies such as Spiderman and Star Wars. It's a benchmark which stresses many components of your computer, including memory.

A test at stock speed was made to be sure that these sticks were not defective, and then I proceeded onto overcocking. Benchmarks were done with stock speeds to compare with overclocked speeds; they are included in the results.

The ram, as expected, handled DDR2-800 CL3 without a hitch. It would only boot up with 2.35V and timings 3-3-3-9. This was the only frequency attained with CL3, because the frequency increase was not significant with the extra voltage added. You will see little to no performance gain with CL3 over CL4 or CL5 using Intel's P965 chipset, so take these results with a grain of salt.




These results from the Everest and the Sisoft Sandra tests were expected. The modules performed well nonetheless, accomplishing this speed with tight timings does not happen often.



The Crystalmark results were nothing to sneeze at either, these modules accomplished 13,000 points with relative ease. The Winrar score is decent as well, dashing past 1,100 KB/S without a problem. The results do not seem too significant at the moment, because they are not compared to any other CL3 results. It was unfortunate that these modules could not be pushed any further.



Next up is Cinebench, it is somewhat based on the frequency of the ram modules and the timings. Again, the chipset issue makes the CL3 results almost equivalent to CL4, so you will not see too much of a performance increase here.

These modules are definitely something to consider for overclockers. They have the ability to run CL3 under warrantied voltages, and they only get warm to touch at 2.35V. Any higher voltages would not yield any significant frequency.

Let's move onto the juicy parts of this review, where we compare the XP2-8500 results to the http://www.hardwarecanucks.com/hardw...-2-review.html results. It is time to display some CL4 results, will the high performance Mushkin kit stomp all over the OCZ budget kit in Bandwidth tests? Let's see! (The brighter coloured bars are the results for the XP2-8500's and the translucent bars are the results for the previously tested OCZ Platinums.)


Here are some unexpected results. Both tests were done with the same hardware and software, yet the OCZ kit of ram exceeds the Mushkins in these benchmarks. At 29MHz and .05V less than the Mushkins, the Platinums beats the XP2-8500's by 39MB/S read and an amazing 609MB/S write! The Sisoft Sandra results were equally as shocking, do not be alarmed, testing is not over yet, do the XP2-8500s have what it takes to get earn a good score? Let's move on!



Crystalmark tests the ram modules and it calculates points based on how well your memory performs. This kit has to be the fastest I have ever tested in this category. It reached an astonishing 15,843 points, chugging along at DDR2-1070. The Winrar benchmark also shed some light on these modules, because they were able to top out at 1,242KB/S, which is an achievement in my books.



The Cinebench scores were a big improvement over the CL3 testing, due to the frequency advantage. The higher memory frequency also enabled the CPU to run at a greater speed, which also had its affect on this benchmark.

The CL4 frequency results were nothing short of fabulous. The Muskins sadly could not keep up with the frequency:results ratio of the OCZ Platinums as far as Everest and Sisoft Sandra benchmarks go.

These modules proved to be quite impressive throughout their CL5 runs. The highest frequency attained was DDR2-1121 with only 2.3V. Here are some results for the Everest and Sisoft Sandra bandwidth tests:




The Platinums have a better frequency:results ratio, but, the Mushkins still perform better with sheer speed. The timings were quite loose, but the speed that these modules were running at overruled the ill effects. The Mushkins are definitely the way to go, they achieved the best frequency, enabling the end-users to run a higher FSB speed, and they can hold their ground with the bandwidth benchmarks. It was unfortunate that the XP2-8500s did not respond to higher voltages; minimal overclocks with instability were achieved with anything higher than 2.3V for CL5.



The Crystalmark scores were quite good. A top score of 16,187 points was recorded, which towers over all the other results in this review. This goes to show that sometimes it is best to run higher frequencies rather than tighter timings. On another note, timings did come into play. At 36MHz less, the DDR2-1070 CL4 beat the DDR2-1106 CL5 results by 150 points. This clearly displays the effectiveness of running these modules with tighter timings. The Winrar CL5 results dominate in comparison with the CL4 result, beating it in every stable frequency.



The performance increase of the Cinebench benchmark is almost linear. As the memory frequency increases, the CPU frequency increases, therefore increasing the final score. Overall, the Cinebench results were quite impressive; high scores would not be possible without these modules, because they have the ability to deliver huge performance while staying within the lifetime warranty.

The CL5 results were absolutely the best. Running at a high frequency was no challenge for these modules, and the voltage scaling was also impressive up to 2.3V. Reaching DDR2-1121 is quite an achievement; it enables the end-user to run their FSB speed of their CPUs at an astonishing 560MHz! This means that the end-user is capable of reaching his/her CPU limit without any restrictions.