Underclocked Conroe puts cat amongst the pigeons

If FSB alone was problem, why not stick with at least 3.2GHz/1.06GHz FSB speed then? At the same time, we saw results from few online hacks who did overclock the early Conroe 2.33 and 2.66GHz parts with ease all the way up to 4.6GHz CPU with 1.7++GHz FSB on a current 975X mainboard from MSI - and these are early parts, keep in mind, no benefits yet of post-launch tune ups. Not to mention Conroe's twin, Woodcrest, which is seemingly launching fine in the coming month at 3 GHz CPU and 1.33GHz FSB speeds.

What's going on? Well, two possibilities - one is that, with Woodcrest being the "favoured" twin brother here, all the top Core 2 Duo bins are going to the server and workstation use in the two socket market, where Intel has more of a problem to beat the AMD HT-based Opteron solutions. The high-end desktop here can wait a while longer to really "unleash the secret speed beast" which some believe Core 2 Duo actually is, according to the early overclockers.

One of my contacts raised a possibility that, actually, the NetBurst clock speed tricks brought over into the Core 2 Duo design could, even with simple modular water cooling like Corsair Nautilus, enable consistent clock speeds pretty similar to this 4.26 GHz Presler XE in my house, but at the Conroe per-clock efficiency - right this year!

I'm not suprised - if the initial test midrange Conroes with fancy cooling can do 4.6GHz, well I'd supposed an optimised Conroe XE should easily reach 4GHz at 1.6GHz FSB with decent water cooler, too.

Another possibility is that Intel is leading AMD on. AMD is now feverishly trying to extract the last drop of juice from its aging 90nm process, and I do believe that 3.2GHz AM2 dual-core parts at near 100 W power envelope are their goal in the next two quarters, just in case 65nm doesn't roll out for them too well. Now, if Intel only announced the 2.66GHz Conroe and 2.93GHz Conroe XE in July, AMD might have less sense of urgency to fight, as the competitive benefit on Conroe side may not be that much then.

Then, after AMD has exhausted its 3Q / early-4Q fine-tuning potential, and depending on its degree of success in getting the first 65nm parts out of the door, Intel might try to deliver a big blow by announcing much faster Conroes (and possibly Woodcrests) by that time. Additional testing and open-market feedback Intel gets by that time might convince it to actually up the frequency quite a bit, well beyond 3.33GHz in that case.

But, there is one tiny bit of risk for Intel here: what if AMD does succeed with 65nm smooth migration earlier, and what if that migration has a shortened "initial fixing" time where usually there are no speed improvements over the previous process? If that does happen, AMD could shoot for 3.6GHz dual-core by Christmas, in the most optimistic scenario. Coupled with the excellent AM2 socket memory bandwidth (10 GB/s Sandra results on some sites were really music to my ears - see my own AM2 numbers here next week!), such chip would require Intel to push up Conroe speeds quite a bit, if they want that "unquestionable performance advantage" touted right now.

Of course, AMD is not exactly known for using all the chances it has. During the past 18 months of Intel's NetBurst misery, if it hadn't cancelled the K9 and/or K10 projects, and had one of them going ahead instead of K8L current core improvement, the firm could have delivered a true knockout blow to Intel - maybe even preventing the Apple-Intel "Macinteltosh" deal.

Well, Intel's Conroe underclocking at lauch, whether deliberate or not, might be AMD's last chance in a while to upset Intel with a tuned-up, top-bin announcement to keep the performance lead, if any. Knowing what comes from Intel in the next 12 months, the K8L next year might be the case of "too little, too late" - at least in desktop and workstation/small server space. In the four-socket arena, AMD still rules supreme - ask Mikey Dell, after all. µ