Again remembering my days when overclocking four sticks of 1 Gb memory, my hopes for overclocking four modules of 2Gb were not very high, to say the least. Looks like I was wrong though, but I didn't find out just by myself. Actually one of the reasons that I accepted to test this review is the fact that Tony was very enthusiastic about these modules, boasting he could get them past 1000Mhz (PC8000) without difficulty. This I had to see for myself, but to be honest, in order to get there I needed some help, which Tony gladly provided.
The specific difficulty in overclocking this memory is that these modules are quite picky about the Trfc (Row Refresh Cycle Time) setting. To be able to overclock to a higher level, these modules need a Trfc setting of 54. The problem with this is that most Asus boards do not offer any higher setting than 42 in the bios, which makes overclocking these modules very difficult. Fortunately, Tony (and OCZ) has come up with a solution by reprogramming the SPD set for these and some other modules, setting the Trfc in the SPD already at 54 cycles as standard.
Unfortunately, our test modules were still programmed with a lower Trfc setting, making our overclock look... well... futile to be honest. Not to be despaired however: Tony emailed me a more recent SPD file with the right Trfc setting, the only thing I had to do was reprogram the SPD of the modules
. For this I used a nifty little program called spdtool
, which unfortunately didn't work in my Vista 64 OS. Flashing the SPD in XP proved painless however and I was quickly on my way into more promising OC territory.
For a complete guide, please visit Tony's report
in the OCZ support forums
Dual Orthos Run - 1000 Mhz DDR
Without much problems we got our modules up to a 1000Mhz DDR2, with just a slight elevation in memory latencies. Instead of 5-4-4 we had to relax to 5-5-5, but with that setting and just 2.0 Volts on the memory, everything was fine. We could run Orthos for a couple of hours (screenshot above) and complete our full test suite. We pushed the memory a little harder still, and went up to 1050Mhz DDR, however, full stability in the Futuremark Vantage benchmark was no longer possible. Therefore this setting was not included in our test results. More voltage proved futile, our tested modules reached their peak performance between 2 and 2.1 volts. Other testers report even lower voltages, some reaching DDR1000 with only 1.8 Volts.Testing methodology and benchmark suite
For the speed tests of these modules we put together a simple test methodology. First of all, stock results were tested, with the timings of the modules put exactly according to the OCZ spec sheet. This would leave our QX9650 cpu run at the standard 333x9 setting, or 3Ghz of CPU speed at the standard front side bus of 333 Mhz. The memory was put at a 2:3 divider to the front side bus, resulting in exactly 400 Mhz (800 Mhz DDR) or PC6400 speeds, which is the stock speed for today's modules. The timings are the standard CL5-4-4-18-2T timings.
The second benchmark run was run using a simple overclocking scheme: the front side bus (FSB) was increased to 400 Mhz, while the CPU multiplier was reduced to 8, resulting in a CPU speed of 3.2 Ghz. For a first test, the memory was run synchronized with the front side bus, again at the speed of 400Mhz (800Mhz DDR), while for a second test the memory speed was increased by using a 4:5 memory divider, resulting in speeds of 500Mhz (1000Mhz DDR), something which this memory kit has proven to be capable of in the previous paragraph. The timings for the modules were now 5-5-5-18-2T. For the tests with 2 modules of 2Gb, we used the exact same settings, only for the 2x 1Gb modules, for which we used Corsair's Dominator PC8888C4F kit, we ran at the sharper timings of CL4-4-4-12-2T.
As soon as we ensure full stability of our memory modules on the applied settings, we can get to our test suite. We included a limited test suite more directed towards applications this time, instead of games or generic benchmarks:SiSoftware's S.A.N.D.R.A. 2007
Lavalys Everest 4.00
Futuremark 3D Mark '06 - total score and cpu score
Futuremark PCMark Vantage - memory suite
Superpi 1.5, Xtremesystems edition
Crytek/Electronic Art's Crysis, through Madboris' Benchmark Tool
Techarp's H.264 benchmark
Adobe's Photoshop CS3, benchmarked with Driverheaven's Photoshop benchmark
Maxon's Cinebench 10
Have a look at the test results on the next page >