It doesn't seem like it was that long ago when DDR 1 was just coming out and only a small handful of people had systems running DDR memory. Move it forward a couple years, and we've got DDR 2 memory available, but not in too much of a demand. Unlike DDR 1, however, upgrading to a system running DDR 2 will most likely require the replacement of nearly every other component in your computer, so it stands to reason that DDR 2 may take a little while to catch on.
Never the less, however, we have a nice kit of here for review. I would have to say that when I sat down to start this review, I really had no idea of exactly what was different between DDR 1 and DDR 2, and just how exactly I should go about testing these DDR 2 modules. After a bit of research this was quickly put aside, and for anyone that would like to know a little more about the particulars of DDR 2 memory, you can check out this review which goes into a little detail.
There is one other issue that currently plagues reviewers of DDR 2 memory, and that is the fact that there are only a small number of motherboards that would allow for a direct DDR 1 to DDR 2 comparison. Unfortunately, I do not have on of those boards, so I can do no such comparison.
Mushkin PC2-5300 Dual Channel DDR 2 Memory
The package that I received from Mushkin contained 2x512 MB PC2-5300 DDR 2 modules, 1 metal Mushkin case badge, and 1 ink pen. I have traditionally been used to reviewing Corsair memory, in which the more recent modules have elaborate heat spreaders with LED's and all kinds of other crazy gadgetry. These Mushkin modules, however, appear to be all about business as they sport the more traditional plain heat spreaders. I see absolutely nothing wrong with this, as flashy LED's on your memory modules are only of a small amount of use. There was also no fancy packaging, just sealed anti-static packaging housing the individual memory modules.
One thing I find of particular interest is the speeds that DDR 2 memory runs at. These particular modules are rated for 667 MHz DDR 2 speed (which equates to 1.334 MHz). Most of the Intel LGA775 CPU's that I know of currently run with a FSB of 800 MHz, which leaves quite a healthy margin for overclocking the CPU before even reaching the rated speed of the modules. This healthy margin of overclocking, however, proves to be a bit of an issue in trying to review these modules, as it is quite difficult to actually reach the rated speed of the module. I was using an Intel P4 3.0 GHz LGA775 CPU to review these modules, and I was only able to overclock the CPU to a total of 3.75 GHz @ a 250 MHz (1000 MHz) FSB, which is 334 MHz short of the rated speed of these modules.
With the CPU stock, the board auto set the memory speed to PC2-4300. The best timings I could get at PC2-4300 were 3-4-4-4, which is quite a jump from the rated 4-4-4-12.
Test Setup and Benchmarks (Stock and Overclocked)
The following is the hardware that will be used in the test sysem:
Asus P5AD2 Premium Motherboard
2x512 Mushkin PC2-5300 DDR 2
80 Gig Western Digital Special Edition Hard Drive
MSI GeForce PCX 5750 GPU
I tested these modules with the CPU at stock speed of 3 GHz, as well as my maximum overclock of 3.75 GHz (15x Multiplier, 250 MHz FSB). The following software was used to test the modules:
SiSoft Sandra 2004
TMPG Video Encoding
CDEX Audio Encoding
Unreal Tournament 2004
SiSoft Sandra 2004
This is only a synthetic benchmark, but it can be used as a basis of comparison between other similar test setups. These results are in megabytes per second, and higher is better.
This is quite an improvement in memory bandwidth, both being just over 1000 MB/s faster than at stock speeds. This shows that the memory bandwidth that these modules are capable of is quite high when we get CPU's that run at the same speed as the memory.
PiFast is a program that allows you to compute Pi out to a specified number of digits. For our testing we compute 10,000,000 digits of Pi using the Chudnovsky method, no disk memory, and a 1024 K FFT size. Results are in seconds and lower is better.
As the table above illustrates, our overclocked processor, which is running much closer to the rated speed of the memory, was able to shave off almost 12 seconds from the computation speed. I don't recall getting that low of a computation speed with any of the DDR 1 modules I ever tested with this CPU overclocked. This result is also obviously not all a result of the memory, as I did gain 750 MHz of computing power over the stock speed. However, there is still room to go higher than this even before reaching the rated speed of the memory.
TMPG Video Encoding
Video encoding is one of the most stressful tasks that can be done with a computer. The speed of the CPU as well as the amount and speed of the system memory all play a large part in the time it takes to encode a movie. For our testing, we use a 150 MB AVI and encode it to Mpeg 2. The bitrate was set at a constant 5000k/Sec. The video frame size was set to 720x480, NTSC 16:9. The results are encoding time, and lower is better.
Just as with the PiFast results, the closer the CPU gets to the rated memory speed, the better the results. Our overclocked results shaved nearly 4 minutes off of the encoding time. Our test video is only 9 minutes long, so it is quite obvious how much of an impact the faster speeds would have encoding a 2 hour long movie.
CDEX Audio Encoding
I took a CD (Bad Company - 10 from 9), and ripped it to one long 414MB .wav file. I then turned around and encoded that wave file to a 320Kb/s sample rate MP3. I am sure there is a negligible HD performance influence in this test but with the amount of memory we are testing, it shouldn't be more than a fraction of a second. The time noted below is only the result of encoding the wav file to MP3. The result is the time taken to encode the file, and lower is better.
While not quite as significant as the video encoding results, there is still a drop in encoding time between stock and overclocked speeds, loosing 22 seconds.
Unreal Tournament 2004
And of course, what good is a fast computer if you don't play the occasional game on it. For our game testing, we decided to use Unreal Tournament 2004. To avoid testing the video card, and rather focus the majority of the computing tasks onto the CPU and memory, we run the benchmark in 640x480 resolution. This is obviously not an indication of the actual performance that would be achieved while playing the game, as I highly doubt anyone would play at 648x480. However, this allows us to gain a rough comparison of how performance will scale based on CPU and Memory choices. The results below are the average frames per second for each benchmark, and higher is better.
Just as with the other benchmarks, the closer the CPU FSB gets to the rated memory speed, the more noticeable the difference.
It will probably take some time yet before the majority of computer enthusiast migrate to DDR 2, especially considering the fact that they will have to replace, at minimum, their motherboard, CPU, Memory, and Graphics card.
However, having the number of different speeds of DDR 2 coming from as many of the memory manufacturers as they are, DDR 2 should be quite refined by the time most people migrate to it. During my testing, I noticed absolutely no problems with this memory. I am quite impressed with the results that I was able to achieve when overclocking the memory as well.
While I can not make any assumptions about the overclocking potential of this memory, I can quite confidently say that there is plenty of headroom for any current CPU, which should allow you to hold onto this memory for some time after purchase.
Pros: Excellent performance, plenty of overclocking headroom.
Cons: No CPU's even run close to rated speed yet, requires a whole new DDR 2 machine.
Bottom Line: If you're looking to upgrade to a PCI-Express system, the Mushkin PC2-5300 memory is definitely a good option for your memory.
If you have any comments, be sure to hit us up in our forums.