You probably have keychain material CPUs lying around that could whip Itanium when it runs 32-bit. Hell, maybe even some of the ones you’ve already made into keychains, at least for the moment. You can see Tweakers.net’s benchmarks here. This is unbelievably ugly.
It is so ugly that even if McKinley (haven’t heard much about that lately) is ten times better than this, the Hammer will go through it faster than the Borg through a convention of quadriplegics.
If this is even vaguely representative of what we can expect Intel’s 64-bit processors to do in 32-bit, Intel is not serious about having transitional processors.
Maybe Itanium or (being serious) McKinley will be better in 64-bit than the Hammer, and so what if it costs ten or twenty times more? That approach should make about 2% of you happy, and leaves the remaining 98% to AMD. Intel didn’t become what it is today making 2% of the audience happy; Apple did.
You don’t have to be the absolute best to win in the CPU market; just be good enough, cheap enough, and readily available. That’s how Intel won in the past. Even its high-end products were pretty much the same as the mass-market processors.
People will point out the Pentium Pro as the model that Intel is following, and there are significant parallels.
Both were used as the initial step to move CPU bandwidth upward. The Pentium Pro was very expensive, but a small segment of the computing population was happy to pay the freight, and the technology was soon thereafter incorporated into later generations of Intel mass-market processors. Intel looks like it wants to do that with McKinley, too.
However, this is not 1995. There are several critical differences between then and now.
The Pentium Pro wasn’t this bad in 16-bit processing: Sure, it didn’t handle 16-bit code all too well, but it’s performance was lousy, not ludicrous.
Intel has real competition now: In 1995, AMD couldn’t have tossed out a competitor to the PPro if it had stood on its head. It had to buy NexGen to get up to speed with 32-bit computing.
The world has changed. Come early 2002, AMD should be ready with a line of 64-bit processors also. Processors that can run legacy 32-bit applications fast, unlike Intel’s. Processors that can affordably be put into desktops at around the same time in 2002, unlike Intel’s. Unlike 1995, people will be able to buy something else.
Cost Counts Now: In 1995, 1996, the guys who approved the budgets knew Intel, and only Intel. They didn’t know diddly-squat about computing, it was just something IT did. Computing was like having electricity or phone lines; you needed it, but it wasn’t an integral part of the business. Just a mystery you threw money at every once in a while, which led to a nice high-margin business.
Now, ecommerce is an integral, mission-critical part of most big businesses. Executives have to know more than they had to five years ago. Computing is a big and growing part of their budgets, and intense competition has cut their hardware bills down. They may not understand the differences between Linux kernels, but they do understand paying less money.
Times are getting rough. Even if we don’t go into a recession, the times are bound to get rough. The euphoria is over, and reality is setting in. The only thing the Internet guarantees for businesses is a far more competitive marketplace, and in the blindered world of most corporate beancounters, you face increased competition by cutting costs.
You’re Mr. Corporate Beancounter sometime in 2002. Your company’s finally got ecommerce down, but your margins are down, and you don’t see them getting back up again. You’ve cut your human resources and everything else you can down to the bone. When you’re not worried that you cut too much, you’re worried that if you don’t meet earnings expectations, your stock price goes into the toilet.
You didn’t even know what AMD was two years ago. Since then, you’ve heard little good about Intel in the business press but an occasional good word about AMD.
Your technical people say you need new servers. They recommend Intel McKinleys, and give you this humongous price tag. You decide that maybe it’s time to find out about these AMD people, just to see what they have to offer.
You have a few questions to ask the people selling the Intel stuff:
“Will it run my current software I’ve spend all kinds of money on?”
(Long spiels on the wonders of 64-bit software, a big chunk of that humongous bill which your tech staff didn’t quite mention in its proposal. Dire warnings about why you will doom your business if you don’t replace all your now functioning software that your company sweated blood to get to work right with new software so you can get to do that all over again)
“How will we benefit from this new software?”
(Incomprehensible answer hinting at incredible benefits sometime in the indefinite future, no hint as to immediate bottom-line improvement.)
No real answer.
“Will this new software work well with my old networked machines?”
More evasive manuevers.
You already know it’s going to cost $$$$$$$.
You bring the AMD guy in.
“Will it run my current software?”
“How much is it going to cost?”
A lot less than the Intel solution. You find out the AMD chip costs half or less as much, and you can’t figure out what the difference is. The real savings come from no need for new software now, and the option of what sounds like the same to you when you need it.
Which option do you think is going to sound better to Mr. Corporate Beancounter?
Don’t write me emails telling me why the Intel solution is better. Save your efforts for Mr. Beancounter; he’s the one you have to persuade, in terms he can understand and accept.
You tell him why spending a lot more money and disrupting his current computer infrastructure is going to help him a lot more NOW than the AMD alternative or with more of what he has now. With $$$ signs, not verbiage.
If you don’t believe this, go ask your friendly Microsoft Windows 2000 corporate salesperson. Ask him or her, “How’s business?”
“This is much better because I, Mr. Computer, say so” is no longer an acceptable answer.
And that last statement is the part that’s really the underlying problem.
To a very large degree, this is not driven by technical issues. They’re often just rationalizations, and to the degree they’re legitimate, that just makes them better rationalizations. Look underneath the words to the core motivations.
It’s Really About Power
I’m not saying this is going to be an slamdunk for AMD.
I’m sure many technical staffs will figuratively hold their breaths until they’re blue in the face for Intel, and sometimes they’ll win.
Mr. Corporate Beancounter might be persuaded that Intel systems would be more reliable, though every recall and RAMBUS problem in the media weakens that argument.
The tech staff may find to their horror that Mr. Beancounter turns their argument on them and says, “Since you’re so concerned about reliability, why are we buying new, unproven equipment? Why not buy more of the older and cheaper stuff we know works?”
Any new technology is going to be a much harder sell than in the past. A lot of people got burned with “flavor of the month” innovations during the time of euphoria. They aren’t going to be so gullible to a sales pitch next time.
More importantly, Intel’s 64-bit roadmap looks like a “Back to the Good Old Days” road, when computers cost a lot, companies made a lot of money selling them, and only the annointed understood why.
This is of course very appealing to the resellers and technical staffs; who’ve been wounded in their pocketbooks and egos lately. It’s not so appealing to the people who have to pay the bills.
It boils down to power, between the provider saying “We’ll tell you what you want” and the customer saying “No, I’ll tell you what I want.”
It’s about status, too. Let’s face it, at the same time computing has become more important, the status of the corporate hardware reseller or IT staff member has slipped since forty or thirty or twenty years ago. Back then, they were doctors, now, they’re closer to mechanics.
Why? Back then, nobody outside the Lord High Priesthood comprehended anything about how those mainframe temples worked.
Now, the average corporate executive probably suspects his thirteen-year-old kid could give his tech staff (which is often not much older) a run for their money. How hard could it really be if all these pimply-faced punks have it figured out?
(Of course, the corporate exec has maybe even more absurd illusions and delusions about his unique skills and abilities, but he’s the one with the money.)
The mystique of a generation ago is mostly gone. The blind faith is gone. The beancounters think they know enough to dare to lift the curtain, and they find Oz a lot. Or they think they do. The issue is not whether they know enough, they probably don’t most of the time. What has changed is that they now think they know enough to meddle when they didn’t before, and perception is more important than reality here.
This is terrible for the self-annointed Priesthood, how dare these lower species question their august judgment?
What to do? Why, get the mystery back! Find something new that these presumptuous ones won’t have the gall to think they comprehend. Like Linux. Or Itanium. Or anything else that will put them back in their place and us back in ours.
It boils down to the same for Intel. “It is us who determine what you need. We determine the pace. We determine the timetable of innovation. If we decide you need Rambus, then by God, you’re going to get Rambus, even if you can’t see why.”
“Oops, that didn’t work? If at first, you don’t succeed, try, try again, this time with Itanium and McKinley. If we decide you need 64-bit processing and have to dump your old stuff, then by God, you will. How else can we get our old domination back? We need a new standard only we can use.”
Insanity is doing the same thing again and again, and expecting different results. Intel has either learned nothing from the Rambus fiasco, doesn’t realize the Rambus lesson should apply here, or the bureaucratic dinosaur just keeps lumbering along on its own momentum no matter what the tiny brain on top says.
AMD seems to look upon this differently. Though they may be making a virtue out of necessity as a have-not, their attitude seems to be, “The good old days are over, and they weren’t very good for us anyway. We can still make good money in this business without being God. Let’s give the customer what he wants.”
AMD may back off after realizing the ultimate consequences of their actions (which is the commoditization of the CPU industry). They may not be able to follow through and deliver. But they have the right attitude for the times, and Intel doesn’t. That may prove to be their greatest weapon.