hs2,
Since we're in the dark, it's pretty hard to pass judgement on AMD's management. In a year or two, we'll be able to assess how management is doing now.
There are some points that happened year or 2 ago, about which we already are a year or two in the future, and I think it is a good idea to reflect on them, evaluated what succeeded, what failed, and how the current plan is learning or not learning from past successes / failures.
I advocated early on a "Duron" even before the name existed, which would conist of Athlon core from Slot A minus L2, completely L2-less. There was a huge opportnity throughout 2000 for a chip like that, and AMD could have with a guerilla hit and run strategy walked away with 100s of millions of extra revenues with such a strategy. Basically, there would have been a single core, and external L2 would have been optional.
Once Thunderbird was in the full swing, the need for Duron disappeared, since the difference in die space (savings) vs. resources wasted, headaches, confusion resulting from 2 different cores produced a negative return. AMD management should have been quick to realize that, and not even start Duron, or kill it quickly, once it was obvious that the strategy is not working.
The problems has grown and grown, Austin became Duron only fab, running at 20 to 30% utilization, since demand for Duron just did not exist. Yet, AMD has this Appaloosa on the roadmap.
Maybe I am being presumptuous, but somehow I think that at least 1% of what is posted here finds its way up, maybe through AMD employes reading this thread. One way to look at this thread is that we are like gruppies, and AMD employees are like the rock stars. I believe some AMD employees are reading this thread, if for no other reason than for ego, since they are the center of all the attention here.
Anyway, back to my tirade. Clawhammer with a single memory channel is another huge mistake in the making. Maybe Dale is right about my original point that AMD should just make the Sledgehammer. His point is that Clawhammer should have dual channel DDR, but 1 MB L2 would be 2 much, while 512K would be appropriate. Maybe, maybe not. The trade-off is the same as with Duron. Does the die space savings warrant all the overhead associated with multiple different cores, multitude of products, while AMD is such a small company vs. Intel?
My point is that AMD should fight a guerilla warfare. For example, if it looks like a Sledgehammer is manufacturable, commanding high ASP in $200 up to possible $300, AMD should be nimble enough, drop everything else, drop Palomino, Thunderbird, Morgan, Appaloosa, mobile, SFF, Clawhammer, whatever, and produce a single thing, sledgehammer, make 6 million of them, make some $1.5 billion in CPU revenue alone, and while on this gravy train, think about where the next hole in Intel armor is, prepare for next hit and run.
But unfortunately, what I see is a paralysis. Going through the motions.
Joe |
