Why Not One Hammer Instead of Two?. . .

AMDZone got this comment the other day from someone in AMD about Clawhammer/Sledgehammer introductions.

“Right now, Opteron and DT Hammer have the same window of introduction, but it’s a 6 month window to operate within.”

What an interesting redefinition of time.

The next time you miss a date with your girlfriend, just tell her, “I didn’t miss the date, I just had a six-month window to operate within.”

In other words, they don’t have a clue right now when they can.

Playing It By Ear

Based on the information available, the explanation that best fits the facts is that AMD can’t make millions of high-speed Hammers right now, and they don’t know when they can.

When that’s the case, you go to your contingency plan, and in AMD’s case, that’s Barton. Under the circumstances, that’s a wise thing to do, and ought to let AMD muddle through and buy some more time (provided there aren’t any other skeletons in the closet).

The problem that it does create for AMD is what this does to Clawhammer.

Clawhammer really is the Celeron of the Hammer family. It has much less cache than Sledgehammer, and only has one memory channel, compared to Sledgehammer’s two.

Had it come out in October, or January, or even March, those shortcomings wouldn’t have looked so bad, but put it against Intel’s Prescott (which a delay until fall would mean), and it looks kind of pathetic.

Looks a good deal more even, doesn’t it?

Why Two Sockets?

For a company that’s complained about Intel spawning CPU sockets like sock-eyed salmon during mating season, it seems rather odd for AMD to come up with two sockets for one processor.

People would inevitably mismatch CPU and motherboard, and those who might want to eventually move to a multi-processor Sledgehammer system can’t start off cheaply with just a Clawhammer in the box.

Why Not Just One?

It seems to me it would make a lot more sense for AMD to forget about Clawhammer and just go with Sledgehammer as the next-generation premier desktop CPU.

Sledgehammer has the features needed to compete against and even beat Prescotts; Clawhammer doesn’t.

You can differentiate between the two easily enough; disable or don’t include the multiprocessor capabilities of the CPU.

Standardize the socket 940 design. Those mobo manufacturers who made socket 754 designs shouldn’t have to adjust too much; they don’t have to make the extra pins work in a socket 930 system. You’ll just have simple socket 930 boards, and multiprocessor socket 930 boards.

Yes, this will make the die size rather large, but what do you want, big CPUs that will sell at a decent price or little ones that won’t?

Save the little cheap critters for .09 micron, at worse, a shrunk Barton will fit the bill.

Yes, it’s probably true that getting the process down for Sledgehammer will be harder and take longer than with Clawhammer, but why be hellbent on making what will soon be a trailing-edge chip?

AMD needs to realize the value of “Keep It Simple, Stupid.” Otherwise, we’re likely to see the future spectacle of “Barton is our star. Oops, not any more, we just figured it out, now we’re going to make Clawhammers. Oops, Prescotts are better and nobody wants to sink their claws into our Claws, now what?”

AMD may be already thinking along these lines, at least as a possible option. The term “Clawhammer” has vanished from AMD’s vocabulary as of late. Instead, we hear about “Hammer DTs” and “64-bit” desktop processors. At least linguistically, they seem to be paving the path for that possibility.

Unless AMD figures its problems out real soon, it’s a possibility they ought to pursue.

Ed