A quad-core system will not perform that well in gaming.
As an owner of an AMD Opteron 270 for some time I can already confirm this (2 sockets both populated with AMD64 dual-core processors, with a fast interconnect and the ability to aggregate memory sub-system performance on top of that using NUMA).
A dual-core (3 issue per core) 65nm AMD64 will not perform all that much better (per clock it should be very similar) than the existing dual-core (same 3 issue per core) 90nm AMD64 processors.
The 65nm ones are likely to be clocked around 3.5 GHz (vs the 'apx' 2.4 GHz of cost effective 90nm ones of today) and use similar power (also because of the 65nm transistion), but even at 3.5 GHz they won't be able to compete with a 4-issue dual-core processor .... well not in gaming at least.
A quad-core 65nm would not help in gamers either .... but would in server apps.
ie: A 'quad-core on a (single) chip' 65nm AMD64 (Athlon 64 / Opteron) would be great for typical server style roles though, just not in gaming.
Great news is I can get a very nice gaming machine cheaply (US$530 for Intel 'Core' Conroe CPU around October 2006), and can upgrade servers to quad-core (per socket) at the same time if need be
It is win-freaking-win baby 8)
It'll take gamers quite awhile to accept the above, but that is going to be their loss, not mine / ours.
Anyone familar with both the Intel and AMD roadmaps & their CPU manufacturing (mostly die shrinks from 90nm to 65nm) already knows what is going on and why AMD may trade their 'small gamers market' for a 'larger server market' then try to break into the business PC line. (Sales are better than just selling a 'few' high end parts in smaller quantities).
With HD Video (editing, etc) taking off, a quad-core machine will be easier to sell to Joe Consumer, but most games will not be taking real advantage of such processors for 5 years or so. (eg: It is better to have a wide-issue single core at the highest possible clock speed for gaming, instead of multiple thin-issue cores at not quite the same clock speed - We already know this today without doubt thanks to various skilled coders).
If AMD can not get a new microprocessor architecture out within 9 months (and I am not talking just more cores on a die shrink, I mean something like a AMD K10 release) their best bet would be just to make dual-core at 65nm (vs quad-core at 65nm) and sell the CPUs for 75% what Intel are asking to give them a price/performance advantage (putting their mainstream CPU at the time around US$400 by October 2006).
If someone can get an up-to-date AMD roadmap and post a link here it might shed some light on the situation.
The best seaons to buy PCs are March - April, or October - November, most very large megacorperations already know this.... and during the next 2 large sale periods (end of 2006, and early 2007) AMD are going to be losing the market share they just gained.
No doubt Intel planned this to reclaim the market share they just lost, and then some.