For reasons unknown, a good portion of the enthusiast market still thinks that memory cooling is a fruitless effort. Many high-end users still believe that memory chips run plenty cool, and that putting heat spreaders and heatsinks on them is a superficial effort at best, and that they don't provide any major benefits. The fact that nearly all high-end memory module and graphics card memory manufacturers use some sort of additional cooling doesn't seem to matter.

For a long time, I was a member of this camp as well. The first memory modules with heat-spreaders showed themselves in 2000-2001, right about the time that the industry was shifting from SDRAM to DDR SDRAM, and Rambus RDRAM modules were hitting the market, all of which needed to be equipped with heat-spreaders due to their implementations. It was right about the time when DDR-400 modules were hitting the market that DDR heat spreaders became all the rage. As clock speeds increased, heat spreaders became more and more prevalent, even though proof that they actually helped was questionable at best. Nowadays, they are more or less standard on high-end memory modules.

At first, memory chip heat spreaders were more for show and less for actual functionality, but once clock speeds of 800MHz+ hit, DDR and DDR2 modules start putting out some serious heat. Not only do modules have to dissipate self-created heat, but they also are (in most cases) sandwiched next to hot-running processors, chipsets, and graphics cards, all of which are outputting tremendous amounts of heat nearby. I've personally burned the ends of my fingers while touching un-cooled memory chips in cramped multi-core, multi-GPU systems running benchmarks. From that point on, we've come to realize that cooling one's memory modules may not be absolutely necessary all the time, but anytime you're dealing with that kind of heat, moving it away from the PCB and memory chips can only be a good thing, both for short term stability and long-term module reliability.

Now, with heat-spreaders and heatsinks both widely used, OCZ is taking the next step in terms of memory module cooling, a step which some might questions and others will be impressed by. Water cooling. Once it's mentioned, some enthusiasts will chuckle or raise an eyebrow at the thought. However, OCZ has gone about adding water-cooling to their memory modules in an intelligent fashion, by making it optional. This is the core design element of their new Flex XLC Edition memory modules. 

Retail Packaging - Top

Retail Packaging - Bottom

OCZ Flex XLC Water-Cooled CAS-3 DDR2-800 2 GB Memory Kit Features and Specifications

This latest edition to the highly-awarded OCZ Flex XLC product family is engineered to produce significant performance gains on the latest Nvidia nForce SLI platforms by implementing Enhanced Performance Profiles (EPP) and allowing the memory to operate at a stable 800MHz at CL3 upon start up. The PC2-6400 CL 3 Flex modules are integrated with an EPP programmed SPD to immediately boot at the correct settings to produce 3-4-4 timings on nForce 680i motherboards. This feature provides a true "Plug and Play" overclocking feature that eliminates the need for manual configuration and makes memory optimizations a household item for the complete range of consumers looking to maximize system performance through overclocking.

Looking over the specifications, OCZ has concocted a very potent mix of features with the Flex XLC product lineup. While OCZ does produce similar modules using this same cooling system at ultra high clock speeds (up to 1150 MHz at the time of writing), these particular modules which we got our hands on are designed for efficiency at lower clock speeds rather than high clock speeds at higher latencies.

These modules run at low CAS 3-4-4 timings. The majority of enthusiast class DDR2-800 modules run at CAS 4-4-4, whereas mainstream modules run at CAS 5-5-5. Even better, these modules support EPP (or are "SLI-Ready", if you prefer), and will automatically clock themselves to CAS 3-4-4 timings if you have an EPP-ready motherboard.

The OCZ Flex XLC modules definitely stands out in a crowd due to their unique cooling system. While the actual memory chip PCB is standard sized, when the cooling system is added, the overall module height is roughly double a standard module. The height is similar to Corsair's "Dominator" modules, although OCZ's unique cooler design certainly separates themselves away from Corsair, and well, everyone else on the market as well.

The cooler works in multiple ways, which is the beauty of this particular design. The heatsink design is efficient enough to cool the module without specifically needing water cooling. OCZ combines a heat-spreader on one side of the module along with a larger heatsink on the other side, both of which connect to a series of fins at the top of the unit to help dissipate heat. Even the water-barbs help to dissipate heat when the water cooling system is not in use. Without water-cooling even being thrown into the mix, these modules can run at their intended clock speeds and latency levels just fine. 

While you don't have to use water cooling with these modules, it's hard to imagine most end users buying modules with integrated water cooling support and not using it. Each module has a 1/4" (inner diameter) water barb on both ends of the unit, which allows a simple path of water to stream through the inside of the heatsink, taking heat away along with it. Even while water is flowing through the module, the heatsinks on the top and side of the module also help dissipate heat, making for a very efficient overall design. Water cooling, again, is not required, but will certainly will help lower module and PCB temperatures while aiding in potential overclocking. 

We were able to run the modules at their intended CAS 3-4-4-15 timings at 800 MHz with straight air cooling, as expected, and showed absolutely no ill effects. However, in order to reach these timings, we had to bump up the voltage to 2.2V. The stock voltage which OCZ recommends is 2.1V, but at this level, our benchmarks would crash almost immediately at 3-4-4 timings. However, a slight bump up to 2.2V made our system rock solid stable, and we were able to run through all of our benchmarks without crashing, and we were also able to overclock quite a bit at this level as well. OCZ has EVP (Extended Voltage Protection) with these modules, which allows you to bump them up to 2.4V and maintain your lifetime warranty, certainly a nice feature to consider when overclocking.

We used 3/8 outside diameter - 1/4" inner diameter tubing for our tests, which were connected to a dedicated water cooling system. Other water blocks were taken out of the loop, although in most real world scenarios, the water-cooling system would be used to also cool the CPU and GPU along with the memory modules.

We wish OCZ would have included some sort of clamping system for when the tube is attached. Even though our tubing created a tight seal around the barb, it would be nice to have an additional protection method with water flowing next to expensive memory modules. For safety's sake, adding a small zip-tie helped our peace of mind. We did not experience any leaks with the OCZ memory kit, even without additional security measures in place. 

The modules are designed in such a way that they will not impede on the space of the module slot next to them. As you can see, water flows into the primary module, which is connected via a short tube to the next module, which flows water back out to the cooling system. You could certainly purchase two of these kits (four modules total) and water-cool all of them. It would likely look quite impressive, as well. 

For thermal testing, we used an infrared probe to measure the external heatsink temperature, PCB temperature of the memory module, and the PCB temperature of the surrounding motherboard. We wanted to see if the water cooling system helped to extract heat from the module itself, along with how this affected the overall motherboard temperatures.

As you can see, adding water cooling to the mix helps temperatures all around. With our water cooling system active, the memory modules were actually cool to the touch during intensive benchmarking and stress testing. Without the water cooling , the modules were warm (but not hot) to the touch. While the water cooling system is quite simplistic, it's also highly effective at removing heat generated by the memory.

We first tested module overclockability using air-cooling, just to see how far we could take them before adding the more efficient water cooling element into the mix. As we mentioned before, the modules run at 800 MHz at CAS 3-4-4-15 @ 2.2V, which was our starting point.

Our first step was to go down in terms of latency. Surprisingly, the modules were able to clock down to CAS 3-3-3-12 @ 800 MHz at 2.2V, which is the lowest latency we've seen to date on a DDR2-800 module. The module would not run at 1T timing, but this is somewhat standard with most larger capacity modules. After an impressive start, we got to pushing up clock speeds. 

Stock Speeds - 800 MHz @ CAS 3-4-4-15

800 MHz @ CAS 3-3-3-12

At the same 2.2V voltage level, we were able to push our DDR2-800 modules up to an impressive DDR2-1142 level at CAS 5-5-5-15 latencies, nearly the same timings of OCZ's far more expensive Flex XLC DDR2 modules.

With one final push, we cranked the voltage level up to 2.3V and got even further. At this level, we were able to clock our modules up to DDR2-1155 speeds at CAS 5-5-5-15 latencies. Perhaps even more impressive is that we were able to run these modules also at DDR2-1066 speeds at CAS 4-4-4-12 latencies when at this voltage level. Our screenshot shows 4-4-4-15, but we were able to push this further down to -12 in our next set of benchmarks. 

1066 MHz @ CAS 4-4-4-15

1155 MHz @ CAS 5-5-5-15

While water cooling did improve our module temperatures by quite a lot, it did not help the overclockability much. We were not able to achieve higher overclocks or lower latencies with water cooling over traditional air cooling. However, when running at these ultra high levels, we would feel much comfortable for long term operation with a water cooling system opposed to air due to the much lower temperatures all around.

We should also note that every pair of modules is different, and we're not assuming that every pair of Flex XLC modules will overclock as well as ours did. These modules overclocked very well, delivering the same clock speeds and latencies as much more expensive modules.

Test System Details Specifications and Revisions

• Intel Core 2 Duo E6600 (2.4 GHz) Dual-Core
• 1 x Nvidia nForce 680i Series Motherboard
• 1 x Maxtor DiamondMax 10 7,200 RPM Serial ATA Hard Disk
• 1 x Lite-On 18x DVD+/-RW Serial ATA Optical Drive
• Microsoft Windows XP Professional (32-bit)

  Test Candidates

• Modules at 1155 MHz (CAS 5-5-5-15)
• Modules at 1066 MHz (CAS 4-4-4-12)
• Modules at 1066 MHz (CAS 5-5-5-15)
• Modules at 800 MHz (CAS 3-3-3-12)
• Modules at 800 MHz (CAS 3-4-4-15)
• Modules at 800 MHz (CAS 4-4-4-15)
• Modules at 800 MHz (CAS 5-5-5-15)

SiSoft Sandra XI Synthetic Tests Higher levels are better for bandwidth, lower for latency.

At stock speeds (800 MHz @ CAS 3-4-4-15), these modules deliver a smidge better memory performance compared to most enthusiast class DDR2-800 modules (which typically run at CAS 4-4-4-15). We also documented the performance of the modules when overclocked. Cranking these modules up to much higher levels (without overclocking the CPU) didn't help memory bandwidth too much, but does improve latency by quite a bit.

Futuremark PCMark and 3DMark Synthetic Tests Higher numbers mean better system/gaming performance.

The additional memory speeds and lower latencies helped PCMark performance, but did nothing for 3DMark, which was limited by the graphics card. These tests were done on a board which can adjust memory frequencies independent of the front side bus, so it's interesting to see how little memory speeds affect real-world performance in a confined environment.  Please keep in mind, however, that increasing the memory speed while also overclocking the CPU and FSB would yield better results.

Half Life 2 and Prey Gaming Tests Higher numbers mean better gaming performance.

Half Life 2 shows a nice slope of performance gain as memory speeds increase. Prey shows smaller increases, but increases nonetheless. Gaming environments tend to show better performance with lower latency modules when CPU and FSB overclocking are not part of the equation, which make these OCZ low-latency modules a pretty nice fit for gaming boxes.

Our Conclusion: With the release of their Flex XLC modules, OCZ is offering a product which is substantially more flexible than most other enthusiast class memory modules on the market.  They can be air-cooled or water-cooled.  They'll run at lower clock speeds with tight timings, or at high clock speeds with more relaxed timings.  The Flex XLC modules are unique, efficient, and from what we've seen thus far, the pricing is in line with other high-end modules. 

While our initial reaction was to question the concept of water cooled memory modules, the final shipping product is actually quite impressive. We're quite impressed in the way that the modules don't HAVE to be water-cooled in order to get great overclocks - the air cooling system is highly effective in its own right. However, if you add water cooling to the mix, you'll drastically lower your module temperatures and simply have a safer and potentially more stable computing environment when overclocked.

We also like that OCZ is offering these modules both at high-end clock speeds and for those who appreciate low latencies without overclocking. However, even these lower-clocked 800MHz modules were able to heavily overclock, as shown by our tests, far further than we were expecting them to. These modules give you a lot of flexibility in if you want to run at ultra low latencies or ultra high clock speeds.

Our only major gripes can easily be fixed with a future revision of the cooling system as well. We would like to see OCZ move to larger 3/8 OD diameter water barbs, which are more common in today's water cooling systems. We also would like to see a clamp / security system so users know for certain the tube is in place and secured. Beyond that, the modules did everything we asked them to and well beyond.

Hybrid Air/Water Heatsink Design Impressive Temperature Drops When Water Cooling Is Used Low Stock Latencies Very Overclockable	Infrequently Used 1/4" Water Barbs No Water Tube Latch System Specific CAS-3 Modules Difficult To Find On Store Shelves

Discuss This Article In HotHardware's PC Hardware Forum!