"That AMD is going to have a tough time making compelling proof points for Barcelona, other than some strange metric of performance per clock per dollar per watt in SPECfp_rate. But once reviewers get a hold of some production parts, the reviews will be scathing. Then will come the downgrades, etc."
Making compelling proof to who? I say this stepping is meant to satisfy the socket compatible upgrade customers. Are they going to be disappointed?
You maybe right about the reviews. You may be right about the analysts reaction. The truth is this is the first monolithic x86 quad core processor known to mankind, and something Intel will not be able to achieve until Nehalem and CSI.
Re: No need for any excuse. There is always some obscure measure by which an AMD product comes out ahead. Since AMD's benefactors seem happy to finance continuing losses, the performance/$$ metrics will be the new banner for AMD.
Don't get me wrong, Sarmad. I'm quite sure AMD will find a way to declare victory, and I've already said so, too. My interest is what the analysts and investors will say about it, which will largely be influenced by the reviews and technical analyses. These guys and gals are somewhat banking on a profitable 4th quarter, or at the very least a path to profitability in 2008.
What happens when they suddenly realize that Barcelona is a second rate processor launch, when AMD promised them an Intel killer? And that AMD's next credible promise won't be until Bulldozer in 2009...?
Re: I say this stepping is meant to satisfy the socket compatible upgrade customers. Are they going to be disappointed?
Perhaps. The upgrade proposition itself is pretty underwhelming. Think about it:
"Sorry we promised you an Intel killer and couldn't deliver, but if you pay $370 per processor, you can upgrade those dual core Opterons to something that offers slightly incremental performance. Just don't bother if you already have a 2.6GHz Opteron or above, since upgrading will cause you to lose performance in some apps...."
And by the way, this assumes these customers who have been waiting patiently haven't already gone back to Intel. Somehow, the whole upgradability promise seems a bit of a wash to me. Intel's customers had forwards compatibility ever since Dempsey, and all the way through Harpertown. Barcelona doesn't promise anything new.
Re: You maybe right about the reviews. You may be right about the analysts reaction. The truth is this is the first monolithic x86 quad core processor known to mankind, and something Intel will not be able to achieve until Nehalem and CSI.
Ok, but here's a practical question: "So what?"
Why would people care if Barcelona is native quad core or not. Are you the kind of person who prefers a $100 bill instead of two $50s? Or how about this: Woodcrest was the first quad-issue x86 processor known to mankind, something AMD won't have until... Bulldozer at the *earliest*, I guess. But that and Native Quad Core are specifications that are best relegated as interesting trivia. Since this is an investment forum, I'll focus on whether these features can contribute to having the best processor performance at a given power level. That's what essentially governs the ASP that the product will receive, which contributes to earnings, which contributes to the stock price. Anything else is irrelevant.
We already have a pretty good idea of what to expect from Barcelona.
We've seen slides claiming 15% over Opteron for general integer stuff. Hence at 2Ghz, we might expect it to equal a 2.3Ghz Opteron. That would mean spec_int and anything similar is likely to look *VERY* bad. At spec_int, Barc will be left in the dust by a 3Ghz DC Opteron, let alone Intel's best.
So I'm with wbmw on this one. I think a lot of apps will fall in this category. So is there any good news?
We know about the extra SSE units, which should make Barcelona look very good for video encoding software that's suitably optimized. We also expect it will look very good on spec_fp_rate. Those will be the main high points.
Another high point will be on truly multi-threaded software, where a variety of threads are all touching the same cache lines. This is particularly bad on the Intel fsb architecture because moving cache lines between processors requires flushing them to main memory in between. Barcelona will handle this much better -- and the "true" quad core feature will turn into a real benefit for this class of apps.
But will the high points be enough? I think not. As wbmw says, at 2.6Ghz released 6 months earlier, it would have been an exciting part. At 2Ghz, I expect reviewers will be pretty disappointed.
Are you saying you predict a 2.6Ghz standard power Barcelona within 2 months of the intro?
That would be an impressive ramp. 30% increase in speed at the same power in just 2 months.
I hope you forgive wbmw if he's a bit skeptical. As you point out it's not on any reported roadmap, and I can't think of any instance where AMD published a roadmap, and things turned out even better. But maybe I'm forgetting. If it ever happened before, I'm sure you can point us to it.
wbmw - Right. Human beings are so predictable and rational in their behavior that you can conclude that if Barcelona isn't stellar, that means that's why they left. Please. All you're doing is setting up a false correlation as fact, and expect us to accept your hypothesis as law, if and when the 2 incidents occur in a similar timeframe. No sale here. And I won't downplay your hypothesis any more then than now if what you hope actually happens; I find it baseless now, and I will then, unless some actual evidence becomes available. 'nuff said from me on the matter.