I’ve looked at enough early Venice results to come to the conclusion that 3GHz isn’t a realistic expectation when you use air or even water cooling. The only way you can realistically expect to hit 3GHz or more is with a freeze unit.
A realistic goal appears to be 2.7-2.8GHz, and more than a few aren’t reaching that.
One characteristic of these chips is that they’ll often do fine with default voltage or a smidgen above up to 2.5-2.6GHz, but once you start having to overvolt 10%, you don’t have much more to go after that, no matter what you do.
What to do?
Well, if you really feel like you’re hurting, it is very unlikely that if you wait a couple months, these things are going to get a whole lot faster. I still think 3GHz will eventually be reached on air, but that could be quite some time.
The earliest one might expect to see 3GHz is when the 2.8GHz FX comes out. It’s usually a safe bet that AMD leaves themselves a cushion of 200-300MHz at their rated top end.
Then again, it’s quite possible that AMD may be binning at the very top end (or have to bin for 2.8GHz processors at default). Their roadmaps certainly don’t indicate they think they can go much more at 90nm.
Nor does it seem like 65nm is going to bring any big benefits in a year or so (this is just as true for Intel). Best guess right now is that 3GHz will be no problem, but don’t even think about 4GHz or maybe even 3.5GHz.
A Very Long Wait For Not Very Much
A lot of people have been waiting a long time.
Over the last eighteen-twenty four months, the air overclocking standard has gone from that of a 3.5GHz Northwood to that of a 4.2-4.5 (Northwood equivalent) Hammer.
To put it mildly, that’s not much. Not to blame AMD or Intel, they obviously can’t deliver the level of improvement what they used to anymore.
But what does that do to our hobby? The Athlon XP crew will eventually bump themselves up to Hammer, but then what? If somebody owning that 3.5 Northwood asks, “When can I expect double the performance of my eighteen-month unit?” the answer is “Outside of gaming (i.e. get a new card), forget about it anytime soon.”
One can say “dual-core” until one is blue in the face, but dual-core just isn’t going to be a speed demon kind of thing anytime soon. Being realistically optimistic, it probably won’t be, outside of a few exceptions, a generally speed-demon kind of thing until sometime in 2007.
One can make the most of what we have with a freeze unit, but most will balk at spending $500 or more for a 20-30% improvement (at least for Hammers, PIVs may be a bit more).
Anybody who has been around for a while has to start asking himself, “What’s going to happen to this hobby when the level of CPU improvement drops dramatically from 30-40% yearly to more like 10-15%, and the cost of serious improvement rises dramatically?”
Seems to me the hobby is going to end up shrinking. A lot.