Barcelona and Floating Point
We have now seen the claim that AMD’s upcoming Barcelona chip will have beefed-up floating point operations. From the article:
Early testing on prototypes shows 42 percent greater performance than Intel’s Clovertown on some kinds of computer calculations known as floating-point operations that are used heavily in generating computer images and graphics. AMD is not releasing its other comparative test results. And full-scale independent testing probably won’t be available until July or so.
Of course, servers do not live on floating-point operations alone. Integer performance matters, too, and if AMD isn’t bragging about that while they’re bragging about FP, there’s probably nothing to brag about.
Of course, no one makes a living running floating-point or integer operations all by themselves. They run them to do something in applications.
So the question becomes, “Which applications mostly use integer operations, and which mostly use floating-point operations?
I think AMD’s description of the two types provides a good enough quick explanation:
“Applications such as databases, e-mail servers, Java application servers, and web servers all typically perform better when running on a processor with excellent integer performance.”
“Applications such as computational fluid dynamics, CAD/CAM, digital content creation (DCC), rendering, and financial modeling tools typically perform better when running on a processor with excellent floating-point performance.”
Seems to be Barcelona’s improvements are geared more towards workstations than servers.
Unless AMD decides to strip floating point units away from the junior partners in the next generation (which is unlikely but not inconceivable) what would beefed up floating point mean for us? Would this make K8L an Intel-killer for gaming?
Perhaps surprisingly, no, because not all that much programming in games use floating-point. As this article points out, floating point operations constitute only 10-30% of gaming instructions.
So a 42% max improvement on 10-30% of programming (and there are many other factors that influence real-world results) certainly will help, it’s just not an Intel-killer.
A Private Buyout?
It’s very safe to say AMD hasn’t done too well in the stock market lately:
A big reason for that, especially lately, is that stock market analysts increasingly feel that AMD’s eyes have been bigger than its wallet, and with price wars and much more debt due to the purchase of ATI, it may not be able to finance necessary capital improvements like building/redoing fabs out of its business earnings.
That’s why almost all analysts have been downgrading the stock.
In the latest downgrade, the analyst noted that there are rumors were going around that private equity firms might be interested in buying the company.
The analyst said, ‘While we do not doubt that private equity is sitting on cash it needs to put to work, we have a hard time seeing how it would get involved in AMD at the present valuation.”
I think that analyst is quite correct. Private equity firms generally like to buy companies that are financially healthy but lackluster with a little of their own and a lot of borrowed money. They then use the company’s cash flow to borrow a lot of money to pay back the money they borrowed, tighten up the corporate ship, boost profits, then sell it back to the stock market for a profit.
AMD is hardly a promising candidate for this kind of treatment. First, even today, buying the company would cost over eight billion dollars. Second, AMD is the opposite of a cash cow, it’s heavily in debt and needs a lot more money to finance their plans. A private equity firm couldn’t drain AMD of billions in cash flow, it would have to pump in additional billions instead.
In short, private equity companies like companies with more money than cut-throat business sense. In contrast, AMD’s core problem is that it has no money (and we won’t talk about the latter).
So we’re talking about spending twelve-thirteen billion dollars (which is a huge sum for one or even an alliance of private equity firms) on a decidedly number two company going against a much bigger, sometimes much nastier number one in Intel. Even worse, the more successful AMD is against Intel, the more likely the whole CPU industry becomes less profitable.
No, the sums are too great, and the risks too big for a private equity buyout. I think buyout rumors are just that, some wishful thinking on the part of some dismayed stockholders.
What is much more likely is the other major point the analyst made: ‘In the meantime we think management will be forced to come to the capital markets for operating cash before the end of the summer.’
In a sentence, you don’t want to do that when your stock price is really low.
AMD basically won’t have anything seriously new until the middle of the year. If K8L isn’t good enough to cause a major rally in the stock price, things could get ugly then.