discussion OCZ Flex EX PC3-17000 8GB CL10 Memory Review

[3oh6]

<p style="text-align: justify;">You have an Intel P55 based computer. You are looking for memory and slot yourself into the "high volume" sector of the market. There are a few different routes you can go. The standard enthusiast machine is likely to have a 2 x 2GB kit of memory providing a total of 4GB's of memory. For those in high demand of memory to run virtual machines—for example—you could double that up and be running 8GB's of memory.

The typical response to running multiple sticks of memory in the past has been that of pain and suffering with compatibility issues, timing problems, and overall discontent amongst users. With the Intel P55 platform, most of those growing pains from Socket 775 days are gone, but the latest craze on the memory stage has been higher density memory modules that offer another route to the 8GB system memory destination.

Today we break out of the 2x2GB mold and look at what is "next" in the memory world.

OCZ Technology announced a couple months ago of its intentions to step away from the mainstream—value—memory segment and focus more on flash based hard drives and the enthusiast lines of memory. The move is to shift its focus to what makes the company the most money going forward based on projections. That is SSD hard drive production and development. What it has also done is afforded OCZ to "grow the high-end memory offering".

Today we look at just that; a very high-end OCZ Technology memory offering. This kit of memory is high-end in every sense of the word. From the density (2x4GB), to the frequency (rated for operation at 1066MHz), to the newer slim designed Flex-EX water cooled heat sink. The new Flex-EX is designed to allow for 4 module use in most motherboards. That means that if you so desired, you could stick four of these new modules in a machine and have 16GB of water cooled memory on tap. Let's get started with a closer look at the specifications outlined for this OCZ Flex EX PC3-17000 8GB kit.</p><center><img src="http://images.hardwarecanucks.com/image/3oh6/ocz/ocz3fxe2133lv8gk/index-1.jpg" alt="OCZ Flex EX"></center>

 

[3oh6]

Memory Specifications

Memory Specifications

<p style="text-align: justify;">At this moment in time, the PC3-17000 Flex EX modules we are looking at today are the 'fastest' 8GB DDR3 kit in the OCZ lineup. They also happen to be one of the fastest rated 2 x 4GB kits available anywhere. Here are the complete specifications as listed on the OCZ web site:</p><center><img src="http://images.hardwarecanucks.com/image/3oh6/ocz/ocz3fxe2133lv8gk/specs-1.jpg" alt="OCZ Flex EX"></center><p style="text-align: justify;">The only numbers of interest are the operating frequency of DDR3-2133 or 1066MHz at timings of 10-10-10-30 running at the DDR3 standard 1.65v on the Intel P55 (LGA1156) platform. The obvious standout of these specs are the relatively high timings of 10-10-10, but you have to keep in mind that the clocks this memory is designed to run at are quite high and the timings are proportional. In our overclocking and results section we will be testing the memory at a few different frequency/timing combinations so we'll get a better idea how the timings affect potential bandwidth capabilities of the memory.

The primary aesthetic focus of these modules is the new slim Flex-EX heat sinks. As mentioned in the opening, the new heat sinks should offer the ability to run in four module configuration on most motherboards. Here is a diagram of the new heat sink that shows how the Flex-EX functions.</p><center><img src="http://images.hardwarecanucks.com/image/3oh6/ocz/ocz3fxe2133lv8gk/specs-2.jpg" alt="OCZ Flex EX"></center><p style="text-align: justify;">The Flex-EX heat sinks provide a cooling channel from one end of the module along the top to the other end via 1/8" barbs. The diagram above illustrates this idea. Depending on the finish inside, the cooling capacity should be pretty decent based on surface area of a relatively long channel alone. Keeping in mind that memory doesn't need active cooling and well, the setup is more than adequate. With that said, many people will be asking, is the channel aluminum as they are concerned with mixed metals in their loop. Here is the best evidence that I can find to say that the inside of the cooling channel is the same anodized aluminum as the exterior.</p><center>

</center><p style="text-align: justify;">Based on that photo, I am saying it is anodized at the ends, assuming it is anodized right through. The surface area also looks rather rough which is a good thing, not that it does any good really.

If anyone is familiar with my memory reviews, you are likely aware that I tend to scoff at elaborate cooling solutions for memory. From a performance stand point, not since the Y2K bug was on everyone's mind did cooling actually offer performance benefits for memory. With that said, I can understand an aesthetic aspect of a cooling solution such as the Flex-EX modules. For a high-end build with presentation as high up on the list of requirements, the Flex-EX heat sinks do have plenty to offer.</p>

 

[3oh6]

A Look at the OCZ Flex-EX Modules

A Look at the OCZ Flex-EX Modules

<p style="text-align: justify;">This is the portion of the program that I chastise the manufacturer for putting huge—heavy—modules in a plastic package that cannot contain them properly during transport.</p><center>

</center><p style="text-align: justify;">As noted, I have historically not been a fan of heavy modules coming in the flimsy plastic blister package that has been used by most manufacturers for a lot of years now. Fortunately, OCZ pulled a little rabbit out of their hat for the Flex-EX equipped memory, and possibly other kits as well. The key difference is the little ridge around the outside of the plastic housing. This ridge seems to add strength to the package. It is still easy enough to shake the modules loose within the package, and as soon as you open the package both modules freely fall out—which is my beef with this setup—but at least the package is improving marginally. My suggestion to OCZ will still be to package premium memory in a secure premium package.</p>

</center><p style="text-align: justify;">The package comes with a cardboard insert to identify the modules, but there are no installation instructions on the insert like some manufacturers supply. Let's be honest, if you can't install memory without instructions, you shouldn't be installing memory. I know, how elitist of me, but reality sometimes isn't kind.

This being my first look at these new heat sinks has me somewhat impressed. The finish on the heat sinks appears to be very durable, and the construction appears to be well thought out. As is typical, a sticker on each module outlines the part number, and major specifics of the memory.</p>

</center><p style="text-align: justify;">The heat sinks are comprised of two parts. The front piece which includes the top portion, or water cooling portion, of the heat sink as well as the cooling fins to aid in heat dissipation in air cooling environments. As the image on the left above shows, the back plate mounts to the front section via two screws and simply connects the rear IC's to the cooling apparatus on the front piece.

I have to admit, the heat sinks look quite attractive and I can definitely see the aesthetic value in the design. The integration of the cooling fins on top in-between the 1/8" water cooling barbs is well done and really makes the modules look quite professional. From an engineering perspective based on function, water cooling is overkill, and unnecessary in my opinion. From a design perspective, OCZ has done quite well in my eyes. The heat sinks have little touches that look really good. The only complaint from me would be why they continue to use green PCB's on the modules. If the only purpose of the heat sinks is aesthetic, why not make the entire module look good with a dark PCB?</p>

 

[3oh6]

A Closer Look at the OCZ Flex-EX Modules

A Closer Look at the OCZ Flex-EX Modules

<p style="text-align: justify;">You know how I have been talking about the virtual uselessness of heat sinks on memory, let alone water cooled heat sinks on memory? Well, the next two photos will emphasize my affliction with regards to water cooled heat sinks even more.</p><center>

</center><p style="text-align: justify;">Each image above is of an individual module. In the image to the left, we can see that the thermal material is not even touching the bottom 1/5th of the IC's on the back side. If one is to go to the length of water cooling memory, wouldn't it be beneficial to have contact over the entire IC? You wouldn't mount a water block over only 4/5th's of a CPU, would you?

The photo on the right shows an even further misuse of such cooling methods. On this module at one end of the memory, the back side heat sink is making little to no contact with the IC, on top of the thermal adhesive being in the wrong position. I suppose this could be the only kit off the production line like this, but it isn't the first time I have seen a kit of memory come to the lab without making proper contact with the IC's it is designed to cool. As good looking as the modules are, and as well thought out the design is, if implementation is this bad, what is the point? I digress. Let's take a look at the one item on memory that does affect performance, the IC's.</p>

</center><p style="text-align: justify;">Bare IC's and the marking <strong>E004V8A0941X</strong> doesn't give us a whole lot to work with, especially when we can't find anything via Google. There is one other kit we found with the same IC's, a Mach Xtreme kit tested by James over at <a href="http://www.xtremesystems.org/forums/showthread.php?p=4493234&postcount=1">Xtreme Systems</a>.

The speculation is that they are PSC, but they behave so differently from other typical Powerchip IC's and they don't have the same naming convention. Another suggestion is Elpida C-Die. At this point, it still isn't clear what IC's these are. So much for the IC's helping us out.</p>

</center><p style="text-align: justify;">As mentioned, taking these modules apart was easy. All we had to do was remove the two screws holding the back plate on. With the limited contact on our modules the back plates pulled off easily and from there it was just a matter of leveraging the modules off the thermal pad on the front side heat sink.

With the heat sinks removed, we can see that the design is solid and well thought out, but obviously there is an issue somewhere as our poor contact highlights. I think we have looked at the heat sinks enough, let's move on to a look at installation, the test setup, and our testing methodology.</p>

 

[3oh6]

Install & Test Setup

Install & Test Setup

<p style="text-align: justify;">We will just take a couple quick photos of the modules being installed as it is already quite clear that the memory has height issues. We are mostly interested in ensuring the modules can fit in side by side slots.</p><center>

</center><p style="text-align: justify;">As mentioned, we know these modules are tall. They will interfere with a TRUE mounted in the orientation that has the fins hanging out over top of the memory DIMM slots. This is not a surprise and in no way anticipated otherwise. You can still run a TRUE with almost any motherboard and these modules, you just have to run the TRUE in an East/West orientation, not North/South.

We were also expecting to be able to mount the memory side by side in adjacent DIMM slots without issue based on the OCZ web site. The DIMM slots on the P55A UD7 are as close as they will be on any other motherboard, and as we can see in the photo above, a pair of OCZ Flex-EX modules fit without compromise. This simply means that we can run 4 sticks of Flex-EX for a total of 16GB on pretty much any motherboard that can handle capacity.</p><center><table cellpadding="0" cellspacing="0" width="735px"><tr><td align="left">

</td><td align="left">

</td><td align="right">

</a></td><td align="right">

</a></td></tr></table><br /><table border="0" bgcolor="#666666" cellpadding="5" cellspacing="1" width="735px"><tr><td colspan="4"><b><font color="#ffffff">P55 Test Platform:</font></b></td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>Memory:</b></td><td align="left" bgcolor="#ececec" width="75%">OCZ Flex EX 2x4GB PC3-17000 10-10-10-30 (OCZ3FXE2133LV8GK)</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>Motherboard:</b></td><td align="left" bgcolor="#ececec" width="75%">Gigabyte P55a-UD7 (BIOS F7)</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>Processor:</b></td><td align="left" bgcolor="#ececec" width="75%">Intel i7 860 (925B478)</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>Processor Cooling:</b></td><td align="left" bgcolor="#ececec" width="75%">Chilly1 SS (Tuned by Ruffus)</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>Thermal Paste:</b></td><td align="left" bgcolor="#ececec" width="75%">Arctic Silver Ceramique</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>PCH Cooling:</b></td><td align="left" bgcolor="#ececec" width="75%">Stock</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>PWM Cooling:</b></td><td align="left" bgcolor="#ececec" width="75%">Stock</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>Power Supply:</b></td><td align="left" bgcolor="#ececec" width="75%">Ultra X-Pro 750W</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>Video Card:</b></td><td align="left" bgcolor="#ececec" width="75%">Biostar 8600GTS</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>Additional Fans:</b></td><td align="left" bgcolor="#ececec" width="75%">Scythe Ultra Kaze 120MM 2000RPM 87.6CFM (DFS123812L-2000)</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>Hard Drives:</b></td><td align="left" bgcolor="#ececec" width="75%">Seagate 7200.9 80GB SATAII 8MB cache</td></tr><tr><td align="center" bgcolor="#cc9999" width="25%"><b>OS:</b></td><td align="left" bgcolor="#ececec" width="75%">Windows 7 Ultimate x64</td></tr></table></center><p style="text-align: justify;">Our setup in use today is very straight forward, aside from the CPU cooling. I happen to run a single stage phase change unit on this setup all the time. Some may argue that it will skew memory overclocking results because of the colder IMC on the CPU. To that I say, I am testing memory not the CPU's IMC; if this were a processor review, then absolutely, colder than average temperatures will skew results. What we have done here—or at least to the best of our ability—is eliminated the limitation of the CPU in memory clocking. This will allow us to fully test what the memory is capable of. Let's quickly go over our overclocking methodology now.</p>

Methodology

<p style="text-align: justify;">A debate as to what is and isn't stable for memory and CPU's will go on forever, so we are not here to perpetuate that discussion. Instead, we simply want to outline what we feel is adequate testing in order to provide accurate overclock results to you the reader.

• 8 threads of 32M SuperPi Mod 1.5 (ran at the same time using Hyper Pi 0.99b)
• 20 passes of LinX using all available memory LinX 0.6.3
• 2 hours of HCI MemTest Pro in Windows using all available memory
• 3 loops of 3DMark Vantage (over 30 minutes of looping the full test suite)

<p style="text-align: justify;">All of the overclocks listed on the next page went through the stress testing outlined above. This may or may not equate to 100% stability on your system, but it is a very accurate gauge of what we could expect out of this particular kit of memory in a 24/7 machine.

Keep in mind, however, that the overclocking results of our kit being tested today relies on a number of variables including the memory itself, the setup, and the person behind the controls. Just because we reach certain overclocks with this kit, doesn't mean every pair of modules will achieve the same overclocks. Every stick of memory is different, we are presenting a single set of results from a single kit of memory.</p>

 

[3oh6]

Overclocking Results

Overclocking Results

<p style="text-align: justify;">The heart of any memory review is the overclocking section. There isn't really anything ground breaking going to come from a kit of memory, and the specifications—performance—can be easily summed up in a short break down of the rated clocks. The real information you are looking for comes from the overclocking capabilities. We have already emphasized that overclocking results here are from a single kit, but it still should provide some details of what these IC's are capable of. After all, our random sample might not be the best out there...not by a long shot. Let's look at the results, then break down the numbers.</p><center><img src="http://images.hardwarecanucks.com/image/3oh6/ocz/ocz3fxe2133lv8gk/overclock-2.png" alt="OCZ Flex EX"></center><p style="text-align: justify;">First up, <strong>voltage</strong>. For 24/7 clocks, we quickly found out that anything over 1.72v would end in failure. Some tests like single 32M SPi can run at 1.74v or even 1.76v, but as soon as HCI is fired up, errors would start cascading down rather quickly with those kinds of voltages. So for all our results posted, the voltage is <= 1.72v.

Next, <strong>timings</strong>. At first glance it might look like we got lazy with this kit and only tested a narrow range of timings; that isn't exactly the case. Notice the huge gap between 8-8-8 and 8-9-8? Yeah, dropping tRCD down to 8 resulted in a huge drop off in memory clocks to 850MHz or DDR3-1700. Going down to CAS 7 decimated this kit not allowing us to boot over 750MHz or DDR3-1500, let alone run stable. Needless to say, these IC's are built to run loose timings...and do it quite well. 10-10-10 tapped our memory controller—or simply exceeded the limits that the motherboard could handle. Even with high VTT(1.45v) nothing over 1164MHz could handle HCI Memtest, even at 11-11-11.

Overall, the results are quite solid. With a small bump in voltage to 1.70v vDIMM, we were able to run the rated clocks of 1067MHz or DDR3-2133 at 9-9-9 over the stock 10-10-10. At 10-10-10, this kit seemed to want to clock forever. We were able to run 4x32M SPI @ 1200MHz no problem, but required high VTT, and HCI MemTest would error out rather quickly at those clocks. Here are the screen shots of the runs in the chart above...</p><center><table align="center" bgcolor="#666666" cellpadding="5" cellspacing="1" width="90%"><tr><td align="center" valign="top" bgcolor="#ececec" width="25%">8-8-8
click for full size...

<br>850MHz 8-8-8-28<br>@ 1.72v</td><td align="center" valign="top" bgcolor="#ececec" width="25%">8-9-8
click for full size...

<br>1050MHz 8-9-8-28<br>@ 1.70v</td><td align="center" valign="top" bgcolor="#ececec" width="25%">9-9-9
click for full size...

<br>1068MHz 9-9-9-30<br>@ 1.70v</td><td align="center" valign="top" bgcolor="#ececec" width="25%">10-10-10
click for full size...

<br>1164MHz 10-10-10-30<br>@ 1.72v</td></tr></table></center>

 

[3oh6]

Temperature Analysis

Temperature Analysis

<p style="text-align: justify;">This section is simple. Mount thermal probe to the side of an IC and measure temperature both with and without the Flex-EX heat sinks, as well as with and without active cooling by way of a 2000RPM 120mm fan. HCI MemTest will be used loading up all available memory on our highest overclock of 1164MHz at 10-10-10 with 1.72v vDIMM. The temperatures in the chart and photos below were taken after 15 minutes of HCI MemTest doing what it does...make memory hot. Here is a nice macro of how we mounted the thermal probe to the side of the IC.</p><center>

</center><p style="text-align: justify;">We mounted to the side of the IC so that we could take the same reading with and without the heat sinks attached. Ideally we would mount the thermal probe on the face of an IC, but then that would hinder cooling with the heat sink on.

We end up with 4 different setups. Here are the photos of those 4 setups, and the temperatures that each garnered. In the photos, the top temperature on the digital thermometer is ambient temperature as measured by the probe hanging above the setup in the photos. The bottom temp or <strong>T2</strong> is the temperature of the probe mounted to the IC on our module.</p><center>

</center><p style="text-align: justify;">The first two photos are with active cooling, a 2000RPM 120mm fan sitting over the modules is my standard cooling method for memory. I do not run a system without it...ever. Here we can see a 2 degree difference (33.6C VS 31.6C) in temperature under full HCI MemTest load going from a naked module to a module with the Flex-EX heat sink mounted. Not significant, but definitely a difference showing a cooler IC with the Flex-EX heat sink.

Once we removed the fan, things got interesting.</p><center>

</center><p style="text-align: justify;">This is where we found out that these IC's have a temperature threshold. Basically, at or around 37C on our temp probe, HCI MemTest would start to throw errors...lots of them. I am talking instant, cascading row after row of errors. This is when it reality kicked in, overclocking was likely limited to such a low voltage, because of the sharp stability loss at a certain temperature. As we can see in both images above, the temperature of our module breaks that 37C threshold with no active cooling, and both runs above ended in a run of errors, followed promptly by a blue screen.

The naked module did get hotter quicker before the system crashed, but even with the Flex-EX heat sink, the module would have likely gotten hotter if the system ran longer. What we have here is a bit of an interesting situation. Yes, in an air cooling setup, the Flex-EX did help keep temperatures slightly cooler. With the sharp edge these ICs seem to have as far as temps go, this should help with being able to handle more vDIMM.

Now, I know what you are thinking...where are the water cooling results? The answer is, a lot harder to come by than first expected. Getting down to 1/8" tubing from a 1/2" loop proved more difficult than expected with the lack of a local water cooling shop. For now, we have to leave the conclusion on water cooling down to a theoretical one as far as offering better overclocks.</p>

 

[3oh6]

Conclusion

Conclusion

<p style="text-align: justify;">I guess now is the best time to admit that I am glad I washed my foot this morning because it doesn't taste half bad. I was wrong — there I said it. I mean, water cooling on memory is still overkill in my opinion, but the numbers don't lie. This particular kit of ours was temperature limited which limited the amount of volts we could run through it for 24/7 use. Water—in theory—can help with that allowing the memory to run more volts stable and therefore higher clocks.

The fact of the matter is, higher volts and the associated gains in memory clocks won't equate to 50MHz of CPU clocks in performance. The effort to go to water cool memory just doesn't seem like it is worth it to this reviewer for daily use. Now, for the overclocker, that is a different story. With that said, this kit isn't exactly a "benchmarker's" kit of memory. Either way you look at it, the Flex-EX heat sinks are great to look at, and do quantitatively provide better cooling allowing for better overclocks.</p><center><img src="http://images.hardwarecanucks.com/image/3oh6/ocz/ocz3fxe2133lv8gk/conclusion-1.jpg" alt="OCZ Flex EX"></center><p style="text-align: justify;">Naturally, we have to re-iterate the poor heat sink application though. In light of the temperature limitations of this memory, proper heat sink application is crucial, along with active cooling to get the most out of these modules. It was a bit disappointing to see poor contact on the ICs but this could vary from one sample to another. We suggest you check your modules nonetheless and call OCZ's excellent support department if you have issues with heatspreader fitment. Another point of contention is the 1/8" barbs. Most enthusiasts aren't likely to have 1/8" cooling loops, so concessions are going to have to be made. But as far as we can tell, the interior of the heat sinks are anodized and shouldn't cause any mixed metal issues. Also, to allow for 4 slot use, the barbs can only be 1/8" so it is a design element we can live with.

With all this talk of heat sinks and cooling, we have completely overlooked one thing. We have 2x4GB kits of memory available that can easily run 1000MHz+. Heck, running 1100MHz or DDR3-2200 at 10-10-10 timings was a piece of cake for our kit. Times certainly have changed and it is good to see the early 2x4GB kits being so strong. Sure the timings are loose, but it is 8GB of memory that can easily be 16GB on any enthusiast motherboard. That in itself makes up for any kind of miniscule performance loss from loose timings. Now we just need to find a reason to have this much memory on tap.</p>
<b>Pros:</b>

• Flex-Ex heat sinks look great.
• Form does not limit function. 4 modules can be used, and cooling is solid, even on air.
• There is definitely head room with these modules, 1150MHz+ is great for 2x4GB.

<b>Cons:</b>

• Such elaborate heat sinks require quality application, that wasn't the case with our sample.
• Whether the Flex-EX heat sinks help or hinder performance, they will cause prices to be higher for these modules than they would without.
• These ICs do not like tight timings, these aren't going to be Elpida Hyper killers.
• Knowing that you can fit 4 modules in a standard motherboard, makes you want to go out and buy a second kit.
• Having 16GB will force you to spend hours on end trying to find a use for all the memory

<p style="text-align: justify;"><center><b><i>Hardware Canucks would like to thank OCZ Technology for making this review possible and supplying the memory used today.</i></b></center></p>