On the one hand, AMD has been repeatedly saying that they are within weeks of beginning commercial production of 90nm Hammers, which means actual chips sometime in the third quarter, as this EE Times article reiterates.
On the other hand, XBit Labs reports (no doubt after hearing from someone working in a mobo or OEM company) that engineering samples of 2.6GHz 90nm chips will only become available in October, which means they won’t likely become available until the very end of 2004/early 2005.
Which should you believe? You can believe both. There’s a difference between getting out a 90nm chip and getting out a fast 90nm chip.
Just a few months ago, Intel was saying they were doing fine making 90nm Prescotts. Then we found out that that meant they were doing fine making 2.8GHz Prescotts.
One should also recall that AMD did not introduce desktop Hammers first. They introduced Opterons with a maximum speed of 1.8GHz first, and only six months later did desktop chips show up.
In all likelihood, we’ll probably see this repeated. AMD will first produce slower server (and perhaps notebook) chips first, faster desktop chips later.
Does that mean AMD is having the same level of difficulty making 90nm Hammers fast the way they did with 130nm Hammers? That’s hard to say.
On the one hand, AMD has started off new process generations with non-desktop chips even before Opteron (though then, those were notebook chips)
On the other hand, it would only be to AMD’s benefit to get as much of an advantage they can over Intel while that company is having difficulties, so it’s probably safe to say that they’re not quite ready to put these chips out.
No, it isn’t some clever strategy on the part of AMD to hold back. That would be like David hitting Goliath with one rock from his sling and saying to himself, “I’ll wait until he recovers before shooting another one.”
The best way to judge how much of a problem (if any) AMD is having is to watch for the occasional leak of an advance Opteron processor and see how fast it is running. If they’re running at 2.4GHz or 2.6GHz, there’s little to no problem.
However, if all the 90nm Opterons we see run slower than, say, 2.4GHz, there’s a problem, and the slower it is, the more of a problem there is.
In any event, for the average person reading this, the wait for 90nm will now likely be a bit longer.
Given PressHot’s problems, is anybody else thinking “missed opportunity?”