Print this article

OCZ PC2-9200 FlexXLC


Date
: 01/15/07 – 12:35:47 AM

Author
:

Category
: Memory

Page 1 : Index

Manufacturer:
OCZ Technology

Price:
$599 CND (Ramstore.ca) / $448.99 USD (Newegg.com)

OCZ Technology is one of the best companies in the industry when it comes to satisfying enthusiasts, gamers, and overclockers alike in their memory needs. OCZ was founded upon the idea of providing top of the line memory solutions for those that demanded the best, the enthusiast. This mantra has been in effect from day one back in the late summer of 2000 when OCZ was founded and still holds true today.

Aside from being able to develop ground breaking new technology, OCZ has also gained a reputation for being the best in the entire computer hardware business at taking care of its customers. Technical support and warranty processes are raved about all over internet forums with nothing but praise for the OCZ support experience. I have personally dealt with an OCZ memory RMA as an average joe consumer and I can vouch for these claims. We are not here today to talk about support though, the memory manufacturing industry has become increasingly competitive over recent years and OCZ has fought to retain their position as a leader of that charge. Innovations as of late have pushed the envelope in DDRII memory and that is why I am writing this review.

On November 20, 2006, OCZ Technology unveiled the next step in memory cooling with the introduction of the FlexXLC heatsink on their PC2-9200 2x1GB memory kit. Not only was this memory the fastest factory spec'd memory at the time of announcement but also the first to incorporate a water cooled memory heatsink from the factory. The FlexXLC heatsink is a dual purpose heatsink meaning it can act passively through large aluminum cooling fins or actively with use of water cooling. These modules are certainly one of a kind and this review should be quite exciting, let's get started.

Page 2 : Package & Contents

The OCZ PC2-9200 FlexXLC modules will definitely be the most interesting kit of memory that I have photographed. We start off with a look at the memory in the molded plastic blister pack that it ships in.

Obviously the package is a little bit different from the blister pack that regular OCZ memory comes in but the size of the FlexXLC heatsinks forces a change. The modules are seated on top of each other instead of side by side but despite this layout difference, the blister pack is only marginally taller and wider than the standard memory package we are all use to from OCZ.

At the bottom of the package we find the specifications sticker that is normally on the front side of the modules. I would imagine OCZ has decided to place the same sticker on the rear of the modules instead of defacing the front. With such incredible looking aluminum heatsinks, why would you want cover that with a sticker?

The top of the package is home to the usual OCZ logo and the new FlexXLC logo on the right hand side. So far, the PC2-9200 modules are the only offering with FlexXLC heatsinks but there have been rumors of a PC2-6400 part coming out sporting FlexXLC heatsinks. Keep your eyes peeled to see if that happens or if I am simply propagating mere speculation.

On the rear of the cardboard insert, OCZ has some information regarding the FlexXLC heatsink. This diagram shown here is also on the OCZ web site and we will be looking at it a little bit later in the specifications section.

We will now get a nice close look at these modules outside of the package. I have recently taken a liking to photographing memory so buckle up and strap yourself in as there are plenty of photos to go through. These modules are certainly something to look at, so I will try to cover every conceivable angle.

As suspected, the module has the same sticker outlining the memory specifications on the rear of the heatsink. This is the same sticker that appears on the bottom of the insert seen in an earlier photo.

The sticker on the rear covers up one of the screws and states that warranty is void if removed. These screws are what hold the FlexXLC heatsink on the memory and with the sticker placed where it is, it prevents taking the heatsink off without voiding warranty.

I am still not 100% sure exactly how these heatsinks are put together but if I work up enough nerve, I might just try my hand at taking them apart to have a look. The large piece on the front and back clearly clamp onto the IC's but I don't think it is a one piece design. There might be as many as 4 pieces from what I can see.

This side view of the heatsink gives us a great view of how the ICs come into contact with the aluminum. The front is one piece and the rear is another and they both appear to connect above the water passage to the large aluminum fins at the top. If heatsinks do actually facilitate in cooling memory modules, these heatsinks are going to do it the best of any I have seen. The aluminum heatsinks are really quite beefy and feel very well made.

The biggest asset of these FlexXLC heatsinks also happens to be a bit of a detriment. The aluminum cooling fins are sure to provide excellent air cooling but also make these the largest modules I have used. I am sure this will cause some headaches when it comes to mounting large aftermarket CPU coolers.

Page 3 : Specifications

OCZ has always pushed the envelope with their performance memory. They have always had some of the fastest rated memory and continually look for ways to improve memory cooling with heatspreader design. It seems like only yesterday that I received one of the first sets of DDR memory that came equipped with the XTC honeycomb heatspreaders and now today I look at the first modules offered with the FlexXLC heatsinks which put these modules into a whole new league for memory cooling. Here is a brief excerpt from the OCZ web site describing the FlexXLC heatsinks:

So the heatsink is made of copper and aluminum, the copper being the interior for the water path I hope, and aluminum which would be the exterior that we have already seen. The design of the FlexXLC heatsink is flexible because it can be used with a water cooling setup for active cooling or simply run passively thanks to the large aluminum cooling fins on top and beefy aluminum body. Here is a diagram of how the heatsink in question works:

The way the heatsink works is pretty straight forward. There is a passage for the water to flow from one barb to the other. As the water flows through the passage inside the heatsink, it absorbs heat that is being passed up to the cooling fins from the ICs. If you are not running the modules on a water-loop, then the heat keeps rising to the top of the module where the large aluminum fins get cooled by ambient air. There is more to these modules than the heatsinks however, underneath those new heatsinks are some pretty fancy sticks of memory as well.

The number that should stand out is the massive 1150MHz DDR2. Memory running at that speed is one thing, memory rated at that speed is another. Right now there is a "fastest rated memory" contest going on between manufacturers and PC2-9200 isn't even at the top of the heap. With that said, there are still a few treats tucked into this specifications list. Rated voltage of 2.35v and their EVP slapped on at 2.4v means these modules are warranted up to about 2.5v. The last notable point is the use of an 8-layer PCB as opposed to the 6-layer PCBs that were used previously on high-end DDRII modules. OCZ says this about the 8-layer PCB:

The 8-layer PCB plays quite an important role with performance memory and we will be seeing many manufacturers switch their top lines over to an 8-layer design.

The OCZ web site also has a pair of detailed settings listed for 965 and i680 chipset based motherboards. These timing sets will assist users with these motherboards in getting the PC9200 FlexXLC modules running at their rated speeds. This is your basic plug and play memory, you will have to go into the BIOS in 9 out of 10 cases with these modules to maximize their performance and overclocking of the system is a necessity on AMD system in order to run at DDR1150.

Page 4 : Installation

Normally memory installation is just a matter of dropping the sticks in place and locking the tabs. The taller heatsinks and the fact that you can use water cooling with these modules means more install photos that usual.

The biggest CPU cooler I have that mounts on AM2 is my Thermalright Ultra-120. As you can see, in this orientation, there is no way the Ultra-120 is going to be mounted. Even with the modules in the outside slots, there just isn't near enough room. This cooler is designed to sit high enough to fit over-top of standard height memory modules but the FlexXLC heatsinks are anything but standard.

The Ultra-120 would fit with ease in this orientation and the fan on the CPU cooler could be setup to work very well at cooling the modules but as you can see from the mounting hardware in the foreground, the Ultra-120 only mounts in the other orientation on AM2. This is a drawback of the AM2 rectangle mount. The Socket 775 mount for Intel allows both orientations to mount so using this cooler on my P5B-Dlx would definitely be possible with the OCZ PC2-9200 FlexXLCs.

The next CPU cooler I grabbed was the Cooler Master Hyper TX. This cooler is much smaller in footprint than the Ultra-120 but still is unable to mount with the modules in the Orange slots. Turning the Hyper TX around would not help the situation.

Moving the modules back into the yellow slots does, however, solve the problem. The CM Hyper TX now has no problem mounting the CPU and with the fan sitting where it is, the modules are getting some free cooling. Of course that warm air is going to be now used to cool the CPU but it likely wouldn't hinder CPU temperatures.

I was then going to move onto setting up the FlexXLC modules with my Asetek WaterChill water cooling loop but ran into a problem. For some reason, I completely missed the part in the specifications section where the barbs are listed as 1/4". My Asetek WaterChill setup runs 3/8" and are obviously too large for the FlexXLC heatsinks. This is rather disappointing because 3/8" is about as small as tubing diameter that enthusiasts run. 1/4" tubing diameter is usually only found on pre-made, packaged cooling systems sold by major manufacturers and that type of setup is not likely going to take kindly to an added restriction in the loop.

I understand that there is only so much you can do for barb size connecting to the heatsink and the FlexXLC is already wide enough, but the true enthusiasts that are going to be paying this much for memory are not likely going to be running a water loop of this kind. Of course you could always reduce the tubing diameter for incorporating the FlexXLC heatsink but the flow restriction may not be worth the cooling benefits. Either way, I will only be doing testing of these modules with air cooling because of this compatibility problem.

Because I cannot do any water cooled testing, I will simply be testing as I usually do with a 120mm fan over the memory modules. Even with the added width of the light switch hold down plate for the CPU waterblock, the modules have plenty of room in the orange slots closest to the CPU socket.

As always, the modules will be cooled with a 120mm fan that is positioned directly overhead. This is what the setup looked like throughout the overclocking and benchmarking process. Now we will have a look at the OCZ PC2-9200 FlexXLC modules on my Intel C2D platform where they will also be overclocked and benchmarked.

The first step was to see if I was going to have to move the Maximizer to allow for the PCB to fit in the middle slot and much to my surprise, the PCB fit…barely. Other than that, installation went rather smoothly with these modules in this setup. I already mentioned that the Thermalright Ultra-120 has the ability to mount in both orientations on the P5B-Dlx so there was no test-fit necessary as I knew there wouldn't be a problem.

So in effect, the added height of the FlexXLC heatsinks really didn't alter my usual setup in any way. The 120mm fan that I hang over memory modules was still able to hang off of the CPU mounting bolt and cool the modules and Noctua north bridge cooler. This may not be the case with all users though because large CPU coolers may still interfere. Be sure to check for things like this when considering the FlexXLCs.

The above image is of the complete setup and how it looked for the testing of these modules on my P5B-Dlx. With both platforms being tested and these modules having such a warranted voltage range, there is going to be plenty of stability testing and benchmarks to cover so let's get started.

Page 5 : AMD Overclocking

In the photos on the previous pages we saw that these models literally glisten but the overclocking sections are where this memory gets to really shine. The OCZ PC9200 FlexXLC models are rated for operation at DDR1150 at 2.35v with 5-5-5 timings but that doesn't mean they only run that frequency with those timings. I will be exploring the full gamut of voltage, timings, and frequency combinations that these modules are capable of. We start with a look at the overclocking on my AM2 setup, here is a list of hardware used for testing.

Overclocking memory is an intricate process that requires intimate knowledge of the hardware being used and how each component interacts with each other. The three main players in this process are the motherboard, the memory, and in the case of AMD systems; the CPU with the Integrated Memory Controller. Please keep in mind that I have spent more time overclocking memory on this hardware than most gamers have spent looking through the crosshairs in BF2. Because of experience and knowledge, my results are going to be the absolute best that this memory can produce and every memory kit will have different results. This section should be used more as a reference, combined with other users results with this memory, and not the end all say all that every kit of PC9200 FlexXLC memory will do.

Here is what I have used as a process for stability testing each and every overclock listed on this page:

• Dual 32M runs of Super Pi Mod 1.5 (ran at the same time)
• 2 hours of Prime 95 using Orthos Beta on blend mode
• 1 hour of dual MemTest in Windows using 750MB/instance
• Multiple loops of each 3DMark 01 / 03 / 05
  

  As always, I have broken the overclocking down into the various timing sets that I will be clocking at. I will start with the tightest at 3-3-3-8 and work my way up from there through 4-4-4-8 and finally with 5-5-4-5 which is basically the equivalent to the stock timings of 5-5-5-18. In conjunction with the chart outlining the results of the overclocking, I have included one screenshot of each timing set to illustrate the testing that was completed on all the overclocks.

  
  3-3-3-8
  
  

  

  I was a bit surprised with my 3-3-3-8 results with this kit of memory because normally my AM2 setup doesn't do to well with these timings. These OCZ FlexXLCs really did well at these tight timings and managed to climb up to DDR800 with only 2.29v. For a high frequency kit like these PC9200s to run PC6400 CL3 with relatively low voltage is an impressive accomplishment and should indicate excellent overclocking results at 4-4-4-8. Here is a screenshot of the maximum I could stabilize at 2.50v:

  DDR860 3-3-3-8 @ 2.50v

  

  
  4-4-4-8
  
  

  

  I just mentioned that the tight timing performance of these modules should produce some excellent frequencies at 4-4-4-8 and I certainly wasn't disappointed. The big number that was accomplished was DDR1100 which is simply incredible at only 2.39v. Being able to run DDR1100 at 4-4-4-8 is not something many memory modules will be able to do so these FlexXLCs have really impressed me today.

  DDR1100 4-4-4-8 @ 2.39v

  

  
  5-5-4-5
  
  

  

  With the tighter timing results looked at, we move on to the timings and frequencies that this memory is specified to do and as I expected, the OCZ PC9200 FlexXLCs continue to impress. I honestly wasn't expecting to be able to push this memory much past the DDR1150 that it is rated for because that is already a pretty crazy frequency but of course, going to DDR1200 at the rated operating voltage wasn't a problem. I would have loved to be able to push this memory further because it is clearly capable of more, but my X2 4000+ CPU runs out of gas for anything over DDR1200. Hopefully I can get my hands on an AM2 CPU with a better memory controller that can handle higher memory clocks so I can update these results with some DDR1250 action.

  DDR1200 5-5-4-5 @ 2.33v

  

  We will now switch gears and see how this memory clocks on my Intel C2D platform.

  
  Page 6 : Intel Overclocking
  

  These OCZ PC2-9200 FlexXLC modules are off to a great start having easily maxed out my AM2 system with voltage to spare. It will be interesting to see how they overclock on the P5B-Dlx as the P965x chipset is known for being able to run relatively high memory frequencies when configured correctly. Here is the list of hardware used for the Intel overclocking and benchmarking sections:

  CPU: Intel C2D E6300 (L630A864)
  MB: Asus P5B-Dlx Wifi-AP Edition
  GPU: Biostar 7600GS 256MB DDRIII
  PSU: Silverstone Zeus 560W
  HD: Seagate SATAII 80GB 8MB NCQ
  OS: Windows XP SP2 (with all updates)
  Hipro-Tech.com DDR Maximizer v1.2

  Memory:

  OCZ FlexXLC Edition 2x1GB PC2-9200 5-5-5-18

  Nothing changes in the stability testing for the move to the P5B-Dlx so like I did for the AM2 overclocking section, every overclock listed on this page will have gone through the following battery of stability testing:

• Dual 32M runs of Super Pi Mod 1.5 (ran at the same time)
• 2 hours of Prime 95 using Orthos Beta on blend mode
• 1 hour of dual MemTest in Windows using 750MB/instance
• Multiple loops of each 3DMark 01 / 03 / 05
  

  Like the AM2 overclocking on the previous page, I won't be overclocking this memory at the specified timings only. Instead, I will be finding the stable overclocks at a variety of voltages with an array of timings starting with 3-3-3-8.

  
  3-3-3-8
  
  

  

  With such a wide range of voltages, we have quite a few results here. At the low end, this memory managed to pump out DDR720 with only 2.00v. The other side of the coin has the PC2-9200 FlexXLCs running a solid DDR860 with 2.50v. I was kind of hoping for DDR900 after getting DDR860 on AM2 but matching that on the P5B-Dlx is pretty decent. After you consider that this memory is binned for higher frequencies and looser timings, DDR860 at 3-3-3-8 is quite nice.

  DDR860 3-3-3-8 @ 2.50v

  

  
  4-4-4-8
  
  

  

  When I loosened the timings up to 4-4-4-8, I was really hoping this memory was going to knock me for a loop…but it didn't. Yes DDR1090 at just over stock voltage is mighty impressive and I have no reason to complain but you can't stop a guy from wishing big. My ridicules expectations aside, this memory is really something else and we still haven't seen what it will do at the rated timings of 5-5-5.

  DDR1080 4-4-4-8 @ 2.40v

  

  
  5-5-5-8
  
  

  

  I don't feel I need to comment on the 5-5-5 overclocking because the chart right above this speaks volumes. What tiny bit of disappointment I felt with the 4-4-4-8 results was washed away in a heart beat after making the jump to 5-5-5-8. Not only was I able to surpass the specified operating frequency at a mere 2.10v but just look at what happened at the higher voltages. I mean, come on…DDR1260 at 2.30v?!?! These modules are just amazing at 5-5-5 on the P5B-Dlx. They quickly stopped responding to voltage past DDR1260 but DDR1280 was made stable with 2.5v.

  DDR1280 5-5-5-8 @ 2.50v

  

  The smile on my face from these 5-5-5-8 overclocking results won't diminish for a while but I must continue with this review, ear-to-ear smile or not, so we will now look at the benchmark results on the next couple of pages.

  
  Page 7 : AMD Benchmarks
  

  It is now time to see how the various RAM frequencies and timings perform. Normally there is a bit more of a gap in the five different sets of results but with the rating on these modules at PC2-9200, that requires a hefty overclock on the CPU just to run "stock" speeds. Here is the list of settings that I will be running the benchmarks at.

  

  Like I said, in order to run the rated frequency of this memory on AM2, you need to have a CPU that can run at 2875MHz on the 10X multiplier. Let's see how the benchmarks shake down.

  Lavalys Everest Ultimate 06 – Memory Bandwidth

  

  I'll be honest, I don't have an explanation for these results. Apparently 5-5-5-18 provides better bandwidth than 5-5-4-5, even at a lower CPU and memory operating frequency. I double checked results and every time I run these various settings, I get the same numbers. In order to run 5-5-4-5 I did have to change drive strengths slightly and the secondary timings are a bit different so that may be the reason but the difference in bandwidth shown in Everest is bigger than I would expect from the changes.

  SiSoft Sandra 07 SP1 – Memory Bandwidth

  

  The Sandra bandwidth numbers are closer to what I was expecting but still not showing a big advantage to 5-5-4-5. This just goes to show that tighter timings are not always better in every setup.

  Lavalys Everest Ultimate 06 – Memory Latency

  

  The Everest latency results show the same sort of pattern as the bandwidth numbers do. The decided advantage goes to DDR1000 at 4-4-4-8. This is not the first set of memory that shows an advantage to the 4-4-4-8 timing set on this motherboard with this CPU. DDR1000 at CL4 seems to be the sweet spot for memory in my setup.

  FutureMark PCMark 05 & 3DMark 01 SE

  
  

  With the bandwidth numbers being different than I expected, it was going to be interesting to see how the rest of the benchmarks ended up. PCMark 05 doesn't seem to favor the 5-5-5-18 that much but that could be more because of the CPU speed being slightly lower at 5-5-5-8 compared to DDR1200 at 5-5-4-5. 3DMark 01, on the other hand, does seem to like 5-5-5-18 giving up a few more points to that setup.

  Super Pi Mod v/1.5 – 1M

  

  Super Pi Mod v/1.5 – 8M

  

  Super Pi Mod v/1.5 – 32M

  

  SuperPi isn't much of a good overall system test like 3DMark 01 is but it does a good job of giving users a platform to compare CPU and memory power at various performance levels. All three, 1M, 8M, and 32M calculations scale equally a show advantage to higher memory frequencies despite timings. DDR1200 is king in SuperPi, even at the same frequency.

  Gaming Benchmarks – Far Cry / Half-Life 2 / UT 2004

  

  The verdict is in and the games that rely on the system heavily show the advantage going to the 5-5-5-18 results. You will notice in the overclocking section, I did my overclocking at 5-5-4-5 timings and this was because they seemed to overclock slightly better than 5-5-5-8 or 5-5-5-18. After going through these benchmark results, the overclocking difference almost makes sense. If 5-5-4-5 is slightly slower on this system then perhaps that is why it seemed to overclock a bit better. Either way you look at it, the difference is minimal and this memory is just amazing regardless. Let's now look at some Intel benchmarks results.

  
  Page 8 : Intel Benchmarks
  

  Unlike the AM2 platform, the P965x chipset that runs my P5B-Dlx has memory dividers that allow users to run memory at higher ratios that the FSB such as 2:1 for instance. This will allow basically anyone to run this memory at the rated speed without having to have a very big overclock on the CPU.

  

  The above chart shows that I will be running the rated frequency and timings of this memory at 2.1GHz and then giving it a slight overclock from there going up to 2.2GHz with the memory at DDR1260. I will also be running the memory on three different dividers with the CPU speed remaining constant, much like the AM2 testing.

  SiSoft Sandra 07 SP1 – Memory Bandwidth

  

  SiSoft Sandra shows natural progression for the bandwidth because it clearly doesn't measure bandwidth, Sandra simply calculates it based on variables such as RAM and CPU frequency as well as RAM timings. The neat, evenly escalating graph for Sandra is completely different from that of Everest which actually measures the memory bandwidth.

  Lavalys Everest Ultimate 06 – Memory Bandwidth

  

  In Everest, we get an actual reproduction of memory bandwidth and the DDR1260 result (blue), is almost equal with the DDR1281 settings (white) despite being 700MHz light on CPU speed and slightly slower on RAM speed. The reason for this is due to the North Bridge latencies and the fact that the NB operates quicker at 400FSB or less than at 401FSB or higher.

  Lavalys Everest Ultimate 06 – Memory Latency

  

  The latency results mimic the bandwidth numbers but both the yellow and blue settings run quicker than the rest, again, due to the NB and its internal timings that are much tighter below 400FSB. The interesting result is the DDR1068 4-4-4-8 (grey). It shows a memory latency that is almost the same as DDR1280 5-5-5-8. This is similar to what we saw on the AM2 testing where CL4 seemed to be the sweet spot. Keep an eye on this in the rest of the results.

  FutureMark PCMark 05 & 3DMark 01 SE

  
  

  Unlike the bandwidth or the latency results, PCMark 05 and 3DMark 01 show a consistent edge to CPU frequency and then memory frequency, regardless of timings or NB latencies. The two programs are a fairly close representation to how real world applications react to the different overclocks. The difference may not be huge but it is consistent and higher frequency is out in front.

  Super Pi Mod v/1.5 – 1M

  

  Super Pi Mod v/1.5 – 8M

  

  Super Pi Mod v/1.5 – 32M

  

  The SuperPi results are almost a mirror image of the above FutureMark benchmarks. Higher CPU frequency is the biggest influence on performance with memory frequency coming a distant second. Again, higher RAM clocks are faster, but only marginally.

  Gaming Benchmarks – Far Cry / Half-Life 2 / UT 2004

  

  The gaming results are the last set of graphs we will look at here and these look a lot like the SuperPi and FutureMark results. These last three groups of results really go to show that perhaps memory speed isn't as big a deal as long as you can run lower frequencies with considerably tighter timings. Of course to run DDR1068 at 4-4-4-8 for example, a pretty high-end kit of memory is going to be required. Buying a kit of memory like these OCZ PC9200 FlexXLCs really gives you the ultimate in flexibility for matching up your memory to whatever FSB you are going to be overclocking your CPU to, whether it be at 3-3-3, 4-4-4, or 5-5-5.

  
  Page 9 : Conclusion
  

  I don't think there is much for me to say here that I haven't already said and showed in the previous pages. OCZ has really pushed the envelope with the FlexXLC heatsinks on these modules, it is just too bad my water cooling wasn't compatible with the smaller barbs because I would really liked to have seen if water cooling added any overclocking headroom. Even if it didn't, the massive aluminum fins did a fine job letting these modules stretch their legs.

  

  As mentioned, these modules certainly showed that they are not just all show and no go. The overclocking results on both my AM2 and C2D systems really impressed. The DDR1280 achieved on my P5B-Dlx is by far the fastest that motherboard has ever run memory and I don't expect many kits to match that frequency any time soon. For high frequency operation, this is definitely one of the kits to have. If you are in the market for memory that runs tighter timings at lower frequencies, don't run the other way, the OCZ PC2-9200 FlexXLCs held their own at 3-3-3-8 and 4-4-4-8 respectively.

  Despite the alarmingly high price, all modules in this performance group are priced equally high anyway, the OCZ PC2-9200 FlexXLC deservingly take home the O2 Editor's Choice award for their incredible overclocking performance and innovative cooling solution.

  

  Advantages

  Incredible stock frequency
  Passive or active cooling through FlexXLC heatsinks
  Plenty of overclocking headroom with the right motherboard

  Disadvantages

  FlexXLC heatsinks will cause interference issues with some CPU coolers
  You can't ignore the price tag
  I would like to see at least 3/8" barbs

  Overclockers Online would like to thank OCZ Technology for making this review possible.