OCZ 2GB DDR3-1600 Titanium – Intel XMP Edition

[PAGETITLE=Introduction] If you happen to be on the lookout for some great DDR3, you are in a good position. Although the latest memory architecture launched less than a year ago, the choices today are better than ever. Not to mention that pricing has improved drastically as well. Indeed, things have improved quite a bit in the past year. When DDR3 first launched, alongside Intel's P35 chipset, times were exciting. Faster speeds, lower power consumption - it seemed like a win/win. All until the pricing information was revealed, which as expected, was high enough to the point of being asinine. At that time, most 2GB kits of DDR3-1333 speeds retailed for $500 on average. This was unacceptable given the fantastic DDR2 prices. But since DDR3 was fresh technology, no one expected the pricing to be any different. This of course wasn't the fault of the memory companies, as the chips themselves are what proved so expensive, and the fact that they were being produced in such small quantities. Most companies who launched DDR3 at the official launch date actually lost money on their offerings for months, and some even today, I'm sure. Huge prices and not enough demand... not a good situation. As mentioned though, things are improving rapidly, and adoption is picking up. We mentioned a $500 kit above... the same kit now retails for $235 on one popular e-tailer. Faster kits, such as the one we are taking a look at today, hover around the $350 mark. When compared to DDR2, it's still expensive, but for those who are looking to pick up a DDR3 motherboard, the prices are far easier to stomach than ever - and things will only continue to improve. Closer Look. OCZ are no strangers to memory. They hit the market in 2000 with their DDR1 offerings and have been building on a great foundation since. They started out small, but have since evolved into a rapid-growing company that caters to the enthusiast and novice alike. Part of their success can be accredited to a few people they employ, such as Dr. Michael Schuette, who heads up product development. Even CEO Ryan Petersen gets his hands dirty on a regular basis, and the dedication of the entire staff has made the company a serious threat to competitors who've been in the market longer. Today we have one of their latest kits on our test bench, a DDR3-1600 offering that includes Intel's XMP support. If you are unsure what XMP is, you may want to check out our interview with Chris Cox, Sr. Staff Engineer for Intel's Platform Memory Operations. Essentially, XMP is similar to NVIDIA's EPP technology with DDR2, where special overclocked profiles are included in the SPD. This allows even the novice overclocker to go into the BIOS, set the active profile and reboot. Simple. As hard as it is to believe (at least for me), it's been well over two years since OCZ first launched their XTC heat-spreaders. These honeycomb-designed spreaders were designed to be light-weight, while offering plenty of room for airflow and heat dissipation. It's a design that has grown on me over these few years, and I've come to appreciate their simple yet effective design. This particular kit uses the Titanium-colored heat-spreader, representing modules that are built with extreme stability in mind. In my past experience, this has also carried over into being a kit that's not ideal for overclocking, but we'll see how this one goes on the next page. In line with other performance DDR3-1600 kits on the market, this one utilizes stock timings of 8-8-8-28 at 1.8v. High timings in the memory scheme of things, but not bad for DDR3. But no worries, that's where XMP comes into play, and if you have an X38 or X48 board in hand, the modules have the ability to utilize a second profile that improves on these latencies. We'll talk more about that on the following page. [PAGEBREAK] [PAGETITLE=Stable Overclocks, Memory Bandwidth] We've covered the fact that these are 800MHz 8-8-8 modules, but we'll now touch on the XMP capabilities, which increase performance all around. First and foremost, it's important to note that for these to work with XMP, an Intel X38/X48-based motherboard must be used. The X38 chipset introduced support for XMP, as we discussed further in the interview mentioned earlier. That said, another important fact to note is that just because there is an XMP setting for improved timings, it in no way guarantees that such a setting will be stable on your particular setup. XMP profiles are designed to offer good overclocked settings without forcing the user to do all of the tweaking on their own. It goes without saying that manually overclocking would yield better results, but for those not familiar with overclocking, XMP is a fantastic solution. So, if an XMP profile doesn't seem to work for you, it's likely to be a limitation of the processor you are using. With today's CPUs and motherboards, however, it's very unlikely that anyone will run into a show-stopping error with these, since most, if not all, current motherboards should be able to handle the overclock. One of the reasons I like XMP is thanks to the fact that companies like OCZ don't include lackluster profiles. Instead, they are profiles that even a casual overclocker would be able to select and be pleased with. In the case of these modules, the second profile offers the same DDR3-1600 speeds, but with much better timings of 7-6-6-28. The tRAS can also be tightened further, but it's rare to see an actual performance increase there. Your guess is as good as mine regarding the second profile of DDR3-1714, but it's there for the taking if need be. However, the timings are a far more loose 9-9-9, so should be avoided. These are performance modules, after all. Chips used in DDR3 are certainly getting better, proven by the fact that such a nice combination of frequency and timings can be stable with only 1.9v. Granted, that's still not low, and a RAM fan is always recommended, but it's a lot better than the 2.0v+ that would have been required for that with kits last summer. Overall, I have to admit I am impressed by the overclocking ability of this kit. It's changed my thinking on what the Titanium-colored heat-spreaders represent, because they are indeed fairly overclockable, as you can see below. Stable Settings 800MHz, 8-8-8-28-1T 1.8v. 800MHz, 7-6-6-28-1T 1.9v. 850MHz, 7-6-6-28-1T 2.0v. 900MHz, 8-8-8-28-1T 2.16v. I should note that for a setting to be considered stable, it needs to run through MemTest for Windows for up to at least 500% for total memory installed. Once completed, a half-hour loop of 3DMark 06 is run to assure absolute gaming stability. Lastly, stock settings are restored, and then the overclocked setting is chosen once again to make sure that it was not a lucky one-time affair. That said, the kit did prove a little tricky to overclock at the top-end setting. The first three settings mentioned above were an absolute breeze, but at DDR3-1800 speeds, more tweaking was required. In the end, most of the problem surrounded the motherboard Northbridge voltage, which had to be raised to support the higher FSB. Aside from that though, even the DRAM voltage had to be applied a small step at a time. At 2.10v, it was deemed unstable. At 2.2v, it was also deemed unstable. It was with a fined-tuned 2.16v that the kit became completely stable. Finicky, but I was glad to see that such speeds were possible. Testing Methodology. For our memory reviews, we use Windows XP as our OS choice due to it's lightweight (when compared to Vista) stature. Before benchmarking, the hard drive is first defragged with Diskeeper 2008. Before testing begins, the Diskeeper service is disabled and the computer sits idle for five minutes. Our testing machine is as follows: Processor: Intel Core 2 Extreme QX9650 Quad-Core 3.0GHz Motherboard: ASUS Maximus Extreme. Memory: OCZ 2GB Titanium DDR3-1600 Intel XMP Edition. Sound: Onboard HD Audio. Storage: Seagate 7200.10 320GB x 2. Power Supply: Corsair HX620W Chassis: Antec P182 Mid-Tower Etcetera: Windows XP Professional SP2 Cooling: Corsair Nautilus 500 Each benchmark is run three times, with the results averaged. For our memory testing, we use both SiSoftware's Sandra XII and Lavalys' Everest Ultimate Edition. Memory Bandwidth. One area where DDR3 excels when compared to DDR2 is raw memory bandwidth. There is lots, and on a kit such as this, there could be as much as 100% more when compared to a standard DDR2 offering. Having a huge amount of bandwidth will aide with processing large projects, although the real-world benefits of seeing a number higher than 8,000MB/s is debatable. Our top overclock, though not much greater than stock, made quite a difference on the average bandwidth output here. Despite our 900MHz setting having to suffer with tighter timings, the extra bandwidth should outweigh the loss in latency. First-generation DDR3 kits were unable to break through the 10,000 mark in Everest, but stock speed takes care of that here. Similar to our Everest test, the top overclock reigns supreme, even though it's timings are more loose. On the next page, we will take a look at the latency these modules offer and wrap up with our final thoughts. [PAGEBREAK] [PAGETITLE=Memory Latency, Final Thoughts] Although high bandwidth is nice to have, it can be even more important to have a lower latency depending on the type of application you are running. After all, the quicker a byte can be retrieved, the faster the process should finish. For our tests here, we once again use both Everest and Sandra, with the latter being used for more in-depth latency analysis. Not surprisingly, both Everest and Sandra scaled well with each other, although their latency algorithm differs, quite noticeably. Interestingly though, even though our 900MHz overclock had 8-8-8 timings, it's overall frequency helped it prove faster than even our 850MHz setting with 7-6-6. I didn't expect to see that. In fact, our 800MHz 7-6-6 proved to be the second-best...

OCZ 2GB DDR3-1600 Titanium – Intel XMP Edition

Page 210 –

Rob Williams