Final thoughts, HEXUS.awards, HEXUS.where2buy and HEXUS.right2reply
We've shown today that Corsair's new Dominator TWIN2X2048-8888C4DF is the fastest memory product available. At the time of writing we've seen nothing that can compete with it. However, the increased bandwidth and tight latencies don't seem to give much back in terms of a performance boost. It's measurable, but not noticeable.
Probably the best aspect of this RAM is its cooling solution. It's better than anything else and has helped Corsair to produce its fastest ever memory modules. Couple them with the Dominator Airflow and module temperatures can be kept lower still. In fact, on its own, the Dominator Airflow serves as a useful product for just about anyone looking to cool and overclock any RAM (on its own it should cost around £15-£20).
Speaking of overclocking, when these PC2-8888 modules aren't running at their tested speeds, they overclock higher than we've ever seen. That's got to appeal the extreme enthusiasts out there, right?
What lets down this product is its compatibility and the EPP issues we experienced. Right now, it looks like i975X simply won't play ball with these module at 1111MHz. That's a big blow to anyone who had their heart set on some 1111MHz RAM to run alongside their Core 2 Extreme. This problem is not Corsair's fault, however. It is a limitation of the chipset; unlucky for Corsair and many of its followers with Core 2 CPUs. On the bright side, the modules will enable tigther timings at, say, DDR2-900 on i975X; 100MHz faster than Corsair's current lowest latency DDR2.
Sadly, even running on the system used by Corsair for testing - the M2N32-SLI Deluxe - wasn't without its issues. EPP, in our opinion, doesn't do what we'd expect of it. It overclocks the CPU too far, doesn't alter the VDIMM voltage and ultimately doesn't provide the 1111MHz operating frequency without fuss. However, we did notice some peculiarities in the SPD entries for our modules (the basic SPD entries should be JEDEC spec DDR2-800, but weren't). Our problems could be down to the programming of the SPD/EPP chips on the modules and/or the M2N32-SLI not implementing the EPP settings correctly.
We hope Corsair will be able to shed some light on the problems we encountered, because we really think EPP should be doing a better job than it is with TWIN2X2048-8888C4DF.
Pricing looks to be around £400, although we've not seen any UK e-tailers actually ship any modules yet, so that may change. Still, that's £150 more than what we've seen the (also not yet shipping) DHX cooled Dominator PC2-8500s for. It's not the DHX cooling that's costly, it's the extremely tight screening and speed-binning that's necessary on the DDR2-1111 parts. So, unless you're filthy rich and have a burning desire for the absolute fastest, some more modest TWIN2X or similar will likely bring you the same amount of joy for a nice saving. In time, the appeal of these modules (or future DHX cooled modules) could increase, but right now, they're definitely a niche product.
For Corsair's new cooling solution, we give this product the HEXUS.extreme :: Innovation award, and for it being the fastest DDR2 RAM on the face of the earth, we give it the HEXUS.extreme :: Speed award.
At HEXUS.net, we invite the companies whose products we test to comment on our articles. If any of Corsair's representatives choose to do so, we'll publish their commentary here verbatim.
Paul Watkins of Corsair has kindly responded to HEXUS's .right2reply invitation, providing us with the following:
First of all, thanks very much for the awards. We have put a ton of effort into DOMINATOR and DHX, and we are glad to see that they are being well received.
Wanted to reply, as requested in the conclusion, on the EPP issue noted in the article. We tested on the bench to correlate with your findings, which by and large we were able to do. We believe the issue that HEXUS has seen in this review is related to the multiplier on the CPU. We believe this is a case (as was with the 8500) where a lower end CPU (anything lower than an FX-62 should do the trick) works best with EPP. This is due to the way Asus has the CPU multiplier tables set up. As you noted in your review, with an FX-62 the BIOS tries to set up the CPU to run at a 12 multiplier which gives a resulting frequency of something north of 3.2GHz; definitely too fast for that processor. However, any slower processor sets up with a 10 multiplier giving an end frequency of 2.78 GHz which is easily achieved.
As for the VDIMM, we found that Asus’ bundled AI overclocking application does, in fact, report the memory running at a 1.8 VDIMM but when you use a multi-meter to measure VDIMM it’s correctly running at the 2.4 VDIMM programmed in the module. So perhaps this may be the source of the voltage discrepancy seen by the reviewer.
Finally, I wanted to point out that you are absoutely correct that 975X-based motherboards are simply unable to take advantage of the 1111MHz bus speed. However, I did want to note that this performance can be achieved on the P965 chip set, on boards such as the Asus P5B Deluxe. So don't send back your Core 2 Duo just yet!
Thank you Corsair for shedding some light on some of the issues we had.