Bottom line: Only about 20% of PIIIs, Durons are reaching 1Ghz using air cooling. Even fewer Celerons are reaching that level. Many if not most people reporting 1Ghz+ from Intel chips are using watercooling and/or Peltiers. 1Ghz is also rare for aluminum-based TBirds. Dresden-made copper interconnect TBirds are the only fairly safe bet at this point. They are
rare at 750Mhz, but AMD’s next round of price cuts should put 950Mhz CPUs (which are often Dresden copper chips) down to around $250.
You want 1Ghz. You don’t want 933, or 950 or 980. You want four digits, and you don’t want to hear anything different, and you want it now.
What should you do?
I analyzed our CPU database, and if you’re looking at air cooling, I can’t say prospects are too good at the moment.
Don’t Just Look At What, Look At How
When you look at our database, you need to look at more than the top numbers. You need to look at what people have been able to do with the kind of cooling setup you’re going to have.
If you’re going to use the stock Intel heatsink/fan, for instance, you can’t look at the results from people using multiple Peltiers and say “That’s what I’m going to get.”
What I did was exclude anybody using Peltiers and watercooling from my estimates. If you use these, you will increase your likelihood of success, but it will not guarantee 1Ghz; some use them and still don’t get there.
Take them out and just leave aircooling, and you get the following:
Likelihood of success at 1Ghz roughly 20-25%
Right now, less than 10%, but I expect the percentage to move up some as Celeron processors running at over 600Mhz become affordable, and you can attempt 1Ghz without a big overclock of the PCI bus. However, I wouldn’t expect the numbers to get any better than for the PIII.
What about cC0?
I expected cC0 stepping chips to make 1Ghz a reasonably safe bet.
However, Intel recalled the 1.13Ghz cC0 chip because it would not run reliably at speeds above 1Ghz. The cC0 chips starting to come out are just the same as that recalled chip.
Intel did not recall/pull back the other cC0 chips because it said the problems that forced the 1.13Ghz recall didn’t occur running below 1Ghz. You won’t be running them below 1Ghz; you’ll be running them above 1Ghz; just at the speeds where the 1.13Ghz chip became unreliable. Odds are you’re going to run into the same problems.
So don’t buy it.
Intel is supposed to have a revamp in a few months. If you have to have 1Ghz now, AMD is your only choice. If you have to have 1Ghz and Intel now, find a nice, friendly store that will let you try out a bunch of cB0 chips, or wait a few months until Intel fixes the cC0 problem.
Per our database, about 20% of Durons are hitting 1Ghz. There’s some anecdotal evidence the percentage may be rising, but that remains to be seen.
Looking at the database for TBirds is complicated because there’s no less than four types: slot A aluminum, slot A copper, socket A aluminum, and socket A copper. It looks like the aluminum TBirds rarely hit 1Ghz; while the copper ones usually do.
Clearly, you want a copper TBird. They have been tough to find at a reasonable price, though, just about all 750 TBirds I’ve seen are aluminum. It looks like copper 750s were made in week 24, and not much longer than that, presumably to cover aluminum production shortfalls, or just to get some test processors out there.
However, the percentage of copper 950s and 1Ghz are much higher. Up to now, so have the prices, but if AMD cuts its prices next week, a 950Mhz TBird will only cost about $250, and should have a very good chance of hitting 1Ghz (around 1100Mhz is probably a reasonable expectation). Not much of an overclock, granted, but if you want a pretty sure bet at 1Ghz really soon at a fairly reasonable price, that’s the best out there now.
As I’ve said before, you should visually inspect the CPU to make sure that it is copper. It should look fairly obvious next to a number of green aluminum ones, and codes on the second line starting with AEEA, AHEA, and AJEA verify their copperness (hopefully, we’ll have more in a few days).