Initial bandwidth benchmarks were obtained using Sandra 2003. All results were run Buffered. I began running CAS-2-3-3-6, at 200FSB (1:1) PAT in Turbo mode;
Although 4634MB/s isn't the largest amount of bandwidth attained at 400MHz, it's not bad either. Next I decided to push the memory a bit, and found the ceiling to be 230FSB (460MHz) while running tighter latencies, e.g. CAS-2-3-3-6. In both cases I simply left the VDIMM at 2.85, since the memory is warrantied to 2.9V. I may have been able to run the 400MHz bench at 2.75V, perhaps 2.65V, and I'd recommend finding the lowest voltage for the speed you choose, as it will most certainly prolong the life of the memory, and of course reduce heat. Voltage = heat, and in some cases pushing VDIMM can be defeatist. If the core voltage is 1.85V, then you’re not simply taxing the I/O buffers, your generating heat. As with anything, finding the lowest yet, adequate voltage will increase the life of your silicon.
I wasn't able to squeeze much more out of the modules at these Latencies. This is where 3.0V would have been useful. I may have been able to reach 240FSB at these setting, and perhaps seen higher bandwidth.
Next I pushed the memory to 265FSB looking to find the ideal frequency for maximum bandwidth. I've discovered running high-speed/high-latency DDR intended for Canterwood/Springdale chipsets, usually results in better stability, and larger bandwidth. Each memory I've tried, albeit Corsair PRO4000, or OCZ Gold4000, when running the memory at SPD (which is usually 3-4-4-8) results in scores equal to those run under CAS-2.5. In the benchmark below I first ran the memory on SPD (CAS-2.5-4-4-8) placing PAT on Standard;
Next benchmarks were run setting PAT in Turbo mode, which does tighten the latencies between NB and CPU; Only a 148MB/s per second difference, offset somewhat by the 1MHz speed increase. None-the-less, this does indicate it's possible to run this memory using PAT-Turbo settings, and there are a few PC4000 memory kits out there that simply can't do this, albeit their default speeds or not different. Turbo seems to over-tax many DDR500 modules, possibly because it's in conflict with the manufacturer's SPD. It's almost impossible to run Turbo at the memory's max overclocking speeds. In designing their EL4200 as a "universal" DDR, OCZ seems to have found a conducive SPD for PAT's more aggressive modes. In fact I was able to run the memory in Turbo, up to 280FSB!
I didn't so much reach a ceiling, as I did a plateau determined by voltage. Had I more VDIMM, I certainly could have taken this memory much further. I have the tools for a voltage mod, but since damaging my Corsair Twin-X4000 PRO, I don't dare try it until I purchase a digi-volt meter.
You can get pretty decent Bandwidth, as it surpasses Intel's theoretical maximum (all things being perfect). I did try to tighten the latencies, at various speeds, above 230FSB, but was unable to do so. Still, I'm excited to find memory capable of 560MHz under Turbo-PAT settings. I was unable to do this with OCZ's Premier 4200, although the memory did overclock almost as well. The most frustrating aspect of this review is my anaemic VDIMM. Although adequate, given the current state of DDR, I feel every motherboard should have 3.0V available to the DIMM's. I pushed the memory just about to its highest point running PAT in Standard, and Timings on SPD at 284FSB 1:1;