Test Setup

When trying to boot both sticks with the 1001 BIOS the memory showed up incorrectly as 3348 MB, while the latest BIOS for this board works as it should. This goes to show that you should update your BIOS frequently if you tend to use the latest memory out there.

Performance & Overclocking

Taking a look at the memory SPD programming, it is quite interesting. All four SPD settings are meant for 1.5V operation, but with very specific speed - down to the MHz. Both memory DIMMs have the same production week, even though one stick was received at a later point. I actually won the first stick at an awesome afterparty at the Bellagio from OCZ during CES in Las Vegas and OCZ was so kind to supply the other module for this review at my request about half a year later. As a base for the benchmarks, the FSB was set at 400 MHz. This enables the setting of 1600 MHz for the installed memory. To keep the scaling of the CPU as constant as possible, the multiplier was dropped to 6x, making sure that the processor does not become the limiting factor during overclocking.

Booting the memory at CL7-6-6 and 1600 MHz did not pose a problem at 1.8 or 1.9V. Using the advertised voltage on the OCZ website, we managed a mere 1620 MHz before the memory called it quits. With 1.9V that limit was raised to a more comfortable 1678 MHz. Raising the operating voltage to 2.0V gives you a healthy bump to 867 MHz, which translates to 1734 MHz. So it seems that the memory scales well with voltage.
Next the DIMMs were set at CL5-5-5 and 1.5V. At this setting, they did not manage to boot at 1333 MHz, but worked fine with 1066 MHz. Thus that was used as a base to see how far we can get. At default DDR3 voltage, 1100MHz was the end of the line here. Raising the voltage to 2.0V, also raised the maximum overclockability well beyond the 1200 MHz mark up to 1296 MHz.

Going up the CL ladder, the next step was CL6-6-6, which is quite close to the intended 7-6-6 setting of the modules. We wanted to see what they manage at the various voltage levels. The highest at default voltage in this case was 1304 MHz, which is pretty close to the overall attainable speed at CL5. At a full 2.0V the memory barely managed to break the 1600 MHz barrier by 4 MHz. Running the memory at 1600 MHz means using a 480 MHz FSB. This is quite close to the maximum this board can attain, so just to make sure that the memory was at its limit, the divider was relaxed a bit to 1:2. The memory still did not manage any more with this setting either. While this is certainly nice, it is obvious why OCZ went for the CL7-6-6 rating, to be on the safe side of things and also offer a bit of headroom at the setting. Remember this is a 4 GB kit, which will not scale as well as a 2 GB variant.

The next step meant raising the CL to 7-7-7 and going through 1.5V to 2.0V in 0.1V intervals. With this setting we finally managed to boot at 1333 MHz at default DDR3 voltage and up to 1384 MHz. While it did manage to overclock higher, the benchmarks did not complete and the system was unstable. Cycling through the voltage settings, eventually meant moving the divider from 3:5 to 1:2, as the memory already manages 1708 MHz at 1.8V. Another interesting aspect is the fact that the memory only scales well with voltages beyond ~1.7V, as you will see very little gains when using less than that. In the end, the memory manages a very respectable 1742 MHz at 2.0V and CL7-7-7. This is slightly more than the pair of modules manages with CL7-6-6.

The next run was done with CL8-8-8. This time around, the overclocking was started with 2.0V, followed by 1.5V. Doing so, gives us an idea right away, if we are reaching the limit of the memory. If so, there is no reason to bench the memory at lower speeds in combination with higher latency settings. At this maximum setting, the memory managed 885 MHz, which translates into 1770 MHz. This is only 28 MHz more than with CL7-7-7. Running the pair at 1.5V was possible up 1456 MHz. A benchmarking run was done at default and maximum voltage, for your reference. Seems like we are slowly reaching the limit of the OCZ Platinum EB Edition memory.

As a last step, the modules were set at CL9-9-9 to see if they would go even further. At such high latency settings, DDR3 needs to reach speeds beyond 2000 MHz to spark any kind of interest. The reason for this is simple: DDR2 has matured enough, so that you can get 1066 MHz 4 GB kits running CL4. The OCZ Platinum Series did not manage to climb any higher than with CL8, which seems to be the limit of this pair of modules.




As you can see, the memory does scale with voltage. Raising it from 1.5V to 1.6V does not yield any noticable changes. It also reacts differently with different CL settings. The CL8 line has been included for reference, but the speed increase is so minimal, that CL7 will yield better performance. Remember your kit may perform different. The above results only reflect what is possible with our sample.