|OCZ Reaper HPC PC2-9200 DDR2 1150MHz RAM Kit|
|Reviews - Featured Reviews: Memory|
|Written by Olin Coles|
|Saturday, 13 October 2007|
Page 4 of 5
OCZ2RPR11502GK Test Results
I am going to begin this section by stating that the OCZ Reaper HPC OCZ2RPR11502GK memory kit is not going to be something system builders and casual users should purchase for vanilla systems. Not to sound like an OCZ salesman, but one visit to their website and you will discover that they make a memory kit for every single application imaginable. Seriously, with well over fifty different SKU's I think there's a perfect match for every system. The Reaper series however is for the overclocker, enthusiast, and anyone else who wants to modify settings. I say this because of one simple fact: you cannot reach 1150MHz without changing something in your BIOS.
As it turned out, this Reaper HPC DDR2 1150MHz RAM kit could overclock to 1178MHz without adding additional voltage. This isn't much in the big scheme of things, but it's worth noting that I spent a few days trying to reach farther but nothing ever tested completely stable regardless of voltage. Keep in mind that the default memory voltage for this kit is already 2.3V, and OCZ maintains that the warranty is safe up to 2.35V. So basically, you can add a whopping 0.05V and be safe. Unfortunately, none of the DDR-2 motherboards I have available to me offer RAM voltage increments in less than 0.1V steps. After checking with OCZ, there are a few in existence: the abit IP35 Pro, EVGA 680i, ASUS P5B-Deluxe and P5K-Deluxe, none of which I have on hand or will ever own since DDR-3 is the new standard. What I do have is the abit AB9 Pro (i965) and the Gigabyte GA-P35C-DS3R (P35), and they both work in 0.1V increments. So anyways...
Our first results were recorded from Lavalys EVEREST using the cache and memory benchmark tools. The results shown below represent the average measurement obtained from the OCZ PC2-9200 CL 5-5-5-18 Reaper HPC OCZ2RPR11502GK DDR2 1150MHz RAM Kit while overclocked to 1178MHz.
With both CPU-Z and EVEREST reporting memory clocks at 5-8-8-27, it's a safe bet they are correct. While OCZ rates the Reaper with 5-5-5-18, I have discovered that my Gigabyte GA-P35C-DS3R motherboard is not the best tool for uber-tweakers because it lacks the ability to manually adjust memory strobe timing. Too bad, because the ASUS P5K3 DELUXE I am prepping to replace it does. I also noticed that the Gigabyte GA-P35C-DS3R BIOS did not pick up an Enhanced Performance Profile (EPP) for some reason, as it had done with the Aeneon Xtune PC2-8500 CL5 1066MHz DDR2 RAM Kit I just reviewed. I suppose the serious hardware enthusiast would prefer to climb around the BIOS and custom set these parameters, but nothing makes me happier than hardware that does exactly what it's supposed to do.
Next up was the PassMark PerformanceTest benchmark which runs several different system memory tests in a row. Although some of the tests are specific to the performance of the RAM, others take the CPU clock speed and front side bus into account when developing a score. Most important are the memory read and write tests. In all honesty, I dislike this test because of the external bias it gives attached hardware, but the memory results are still worth noting.
It's also very important that I note how the scores will always be relative to the system components used. For example, PerformanceTest recorded 680.2 Memory Mark points to the OCZ Reaper running at 1278MHz and the Intel E6550 operating at 2576MHz. Conversely, this same program gave the Aeneon Xtune PC2-8500 a score of 670.5 Memory Mark points with the Gigabyte GA-P35C-DS3R motherboard and E6600 @ 2.7GHz., and yet just a few months ago I reviewed the Crucial PC2-8500 CL5 Ballistix DDR2 RAM and it scored 837 Memory Mark points with the Intel P865-based abit AB9-Pro with E6600 operating at 3.2GHz. Obviously the CPU and system board have a lot more to do with Memory Mark points than the actual memory, which is why "score" based tests are often less useful than benchmark readings. Eventually I will probably dump this program, or at least remove everything but the memory specific readings from the image.
During my bandwidth tests with Lavalys EVEREST v4.00.976 I discovered that each round of test results were not very consistent. For practical purposes, I ran the test series three times and averaged the results in the chart. The results for the average read, write, and copy bandwidth from EVEREST are displayed below.
The average read test showed a 2.8% improvement of 226MB/s when the OCZ Reaper HPC was overclocked from 1150MHz to 1178MHz. In the average write bandwidth tests, the overclock yielded a 2.5% improvement of 162MB/s. The 28MHz (2.4%) overclock doesn't seem like much of a difference, and the copy bandwidth test seemed to agree.
SiSoftware Sandra Lite XIIc saw things a little differently; or at least the chart below really makes it look that way. Although a 2.3% benefit given to the overclocked Reaper RAM with 159MB/s in the integer bandwidth test, the float buffered test gives a 3.5% benefit of 234MB/s.
RightMark Memory Analyzer was removed from our results, since the tests were more than a little redundant. I think that in future memory reviews, namely our upcoming DDR3 mega-roundup, I will try to replace one of the other tests with the results of Memory Analyzer. It hasn't been updated in over a year and seems out of development, but it's still not a bad tool since it gives consistent results every single run (unlike EVEREST and Sandra). It offers an average "real" RAM read and write bandwidth result which most enthusiasts don't appreciate because it combines the results of dozens of tests. This is the layman's alternative to the Sciencemark v2.0 test suite, which is also gone but not forgotten.
We also tested the stock settings against the overclocked settings in the game World in Conflict, but realizing that games can be both CPU and GPU bound made it difficult to compare the different memory sets since the clock speed of the processor would change with the front side bus. For what it's worth, World in Conflict did register a full 1 FPS difference in the minimum and maximum frame rates. Obviously, if you are using 1150/1178MHz DDR2 RAM, you aren't going to keep the CPU at the stock speed, and you probably use a video card that more than makes up for the small gains of overclocked system memory. But in the world of benchmarks, it would make comparing the difference in a game somewhat pointless.