With the maturity of a market space, comes the increasing desire to push the envelope. This is true no matter what sector of technology you focus on. Apparently, DDR2 (or PC2) memory has reached a maturity level that is allowing memory manufacturers to tweak more performance out of them, and then bring it to you at an attractive price point.

Case in point, not long ago, not only was a 2GB kit (2x1GB modules) of memory lacking in performance when compared to a 1GB kit (2x512MB modules) it was more then 2x the cost of said 1GB Kit. The timings have gotten better and prices have fallen, so lets take a look at a price conscious 2GB memory kit.

While SPD settings are fine as you can see it wants to time these modules at 5-6-6-18; seeing this, I thought I would run a quick benchmark (albeit a synthetic one) to show the difference (Manual was set to 4-4-4-12).

As you can see, timings do in fact make a difference on performance.

Testing

I will be testing these modules on an Intel based system, for comparison, I will be sampling the previously tested Super Talent 2GB kit (2x1GB PC2-6400) at stock speeds and at overclocked rate (maximum attainable by Patriot).

Test System: Asus P5WD2-E, HIS Radeon X850XT IceQ II, WD 250GB 7200RPM SATA-II, Intel 640 (P4 3.2 EM64T LGA775), Asus A55G 550W PSU

Memory:

Patriot DC PC2-5300 (1.8V) (4-4-4-?)

Super Talent PC2-6400 (1.8V) (4-3-4-8)

When you set them next to each other (the CL values) you can see a slight advantage going to Super Talent, we will see if this plays out through the testing phase.

Test suite

SiSoft Sandra 2005.SR2 - Although a synthetic benchmark, it's a popular one, freely available if you wish to make comparison benchmarks. We will be testing the memory speeds at stock 3.2GHz (667MHz) speeds as well as 4.0GHz (833MHz) OC'd

PiFast - A good indicator of CPU/Motherboard performance is version 4.3, by Xavier Gourdon. We used a computation of 10000000 digits of Pi, Chudnovsky method, 1024 K FFT, and no disk memory. Note that lower scores are better, and times are in seconds.

TMPGenc Plus 2.5 - Video encoding is a taxing chore, both on Memory and Processor, we will be encoding a 150MB AVI file to MPEG2. For the AVI to MPEG2 I used a bit rate of 5000k/Sec, as this is the midrange for a DVD, which is typically between 1000k/Sec to 10,000k/Sec. I used a frame size of 720x480 (DVD Std) and 16:9 NTSC. Note that lower scores are better

Quake 4 - Memory performance does play a role in gaming performance, especially a game that is as hard on an overall system as Quake 4 is. I wanted a “real world” test and as iD appears to have fixed their demo feature to show all effects, it had to be my game of choice for this measurement.

SiSoft Sandra 2005

 

Standard 667MHz above

Overclocked 833MHz above

The results are interesting in that as expected, Super Talent outperforms at stock speeds, however, when we overclock the memory to 833MHz, Patriot takes the lead.

PiFast

This is a dead heat across the board. Almost no change when both memory are set to 833MHz, showing a good performance for both.

TMPGEnc MPEG Encoding

In this test scenario the two pair of sticks are evenly matched, while once again, Patriot edges out Super Talent when set to 833MHz. An interesting trend for a slightly looser timings setup.

Quake 4

Interestingly, there was minimal improvement from stock to overclocked speeds, less then ¼ percent for both, however, the Patriot did have the larger increase (ever slight as it was) then the Super Talent.

Overclocking

While 2GB kits are typically harder to get good overclocks with, usually because their timings are already loosened up, I did not have much of a problem initially with the Patriot DC PC2-5300 kit getting to 833MHz. A simple juice to 2.0V got me there, with the same 4-4-4-12 timings mind you. However, no matter what I attempted to change or manipulate (beyond going above 2.2V, which I just can't get myself to do) I could go no farther, not even when loosening the timings to 6-6-6-18.

With that said, Patriot does a nice showing for modules rated at PC2-5300 (667MHz), climbing to 833MHz while maintaining timings is no small feat. Granted your mileage may vary depending on the system you put it in and of course the chips in your set.

Final Words

We've covered the 1GB to 2GB debate before and needless to say there is no application demanding reason at this time to dump your current kit for a 2GB one if you're satisfied with the performance. However, we here at VL are not the types to settle for a software manufacturer's recommended ram requirements. Our general rule is whatever they say you should have, double it.

has, however, made the decision between a 1GB or 2GB kit much tougher. As these modules are coming across as cost effective, with a little performance kick to boot. We always say, it is better to error on the side of, it might not be required today, but you know within 12 months there is going to be an advantage to having it, then to be caught short.

Although I have been running on 2GB for only a short time, I have yet to run across a situation where I noticed an improvement over 1GB. However, all it would take is that one program, that one thing you want to improve performance on, and then you are going to be hooked.

Overclocking was nothing short of impressive, tight timings and great speeds, gives you some headroom if needed.

It appears as though Patriot not only brings a cost effective 2GB kit to market, they bring one that can compete as well!

HOME