AMD is supposed to demonstrate a four-way Sledgehammer Opteron at Computex in Taiwan. There’s a bit of a hullabaloo about what will be demonstrated.
These won’t be introduced for close to a year. Running a four-CPU cluster is a good deal more complicated than what AMD’s done in the past.
There’s nothing wrong with such a product not being close to being finished at this point. Actually, there’d be something wrong if it were. If that were the case, AMD should bring it to market a lot quicker.
The core problem is that’s there’s a conspiracy going on.
The conspiracy is between the PR types, who live and die for dog-and-pony shows, and the media, who mostly won’t pay attention/comprehend anything more complicated than a dog-and-pony show.
So to get some early attention, you end up with demonstrations that are more like dog-and-pony fetus shows. Not surprisingly, the fetuses don’t do the tricks too well, so what the trainers do is scale back the act.
Intel often does that when they demonstrate the processors they plan to have out in a year or so. You’ve had media see a superduper, supercooled processor do nothing more than say how fast it was running. Wow.
The same purpose would be served if Intel just said, “We’re testing a ___ speed processor now, which we expect to have out in a year,” but I guess that’s too much for the press to remember without visual reinforcement.
If I sound really negative, it’s because I find these kinds of demonstrations completely useless. Unless you’re a cyborg, and very heavy on the cyb part at that, how are you going to be able to gauge how much faster a machine really is solely by seeing the typical demo?
I remember a couple years back, Intel was “demonstrating” a 1GHz processor at a computer show. Somebody rebooted the machine when the Intel folks weren’t looking, and found out the “1GHz” processor was actually a 667Mhz processor.
The matter becomes even more ludicrous when you’re talking about a product that won’t be out for quite a while. All such a demo tells you is that they got it to work. Frankly, I really don’t care whether or not a product works nine months before I get it. It does me no good. If they can’t get it to work two months before it’s due out, then I start caring.
I care even less how fast it is nine months before it’s due. When you’re developing a processor, first you get it to work, period. Then you worry about getting it fast.
It’s like trying to judge a future Broadway show by the first rehearsal. I don’t care if every cast member is a theatrical legend; the first rehearsal is going to be lousy compared to the finished product. That’s why you have rehearsals, to develop the product, see what works and what doesn’t, and make changes to improve it.
Dumb and Dumber
The AMD PR firm in Taiwan apparently drafted some talking points on what to say and not say, probably to those representing AMD at Computex.
“The demos we’re currently showing are anything but high-performance. They’re pretty low-performance right now, and we don’t want to invite questions about just how fast they’re running.”
This is neither shocking nor significant. AMD did pretty much the same thing when they initially debuted Clawhammer late last February. Apparently, they were running about half-speed processors bouncing balls around.
It wasn’t treated as a big deal, and it wasn’t. Again, get it to work, first; get it fast later.
These talking points got accidentally sent out to the media, though, and somebody who either doesn’t know better or was desperate for a story decided to make an issue out of this.
Now it was AMD’s turn to be stupid. Rather than to simply explain the facts of life to this ignorant and/or dumb reporter like I just did, this came out instead:
An AMD spokesperson in Hong Kong confirmed that the company does plan to demonstrate a four-way Opteron server at Computex, but denied that the Opteron chips it plans to use in the demo are low-performance processors. “This is wrong,” says Carol Mui, a company spokesperson. “It’s an eighth-generation processor. How can it be low performance'”
When you run them at low speeds like the statement indicated, dodo.
Of course, a high-performance processor running slowly is different than a slow processor, but when you leave out the “running slowly” part, it doesn’t make a whole lot of sense to the average person, now does it? Sounds more like lying by omission to the average person, doesn’t it?
Yet another case of technical accuracy. For those new to the concept, technical accuracy is what I call a statement that is factually accurate, but misleading.
For instance, Bryant Gumbel is a long-time television personality. Did you know that Bryant Gumbel once posed for a pornographic magazine? This is a technically accurate statement.
Now I did leave out two little details, namely, the Penthouse feature he was in was about men’s fashions, and he posed in suits. Paints quite a different picture, now doesn’t it?
See what I mean?
The reporter may have just been ignorant, but the AMD response is flat-out stupid because now one of the following will happen:
Rather than “AMD has four-way servers functional,” the story now becomes, “Just how low-performance are these machines?” and reporters who otherwise wouldn’t have pursued it now will.
If AMD tells them, the headline is “AMD Opteron servers are slow.” If AMD doesn’t, the headline is “AMD refuses to tell the truth about its slow Opteron servers.” Meanwhile, the Intel folks are sitting back quietly loving it.
What would be even worse is if AMD panics and rushes to get the latest, fastest Hammer silicon out to Taipei. If it crashes, the headline reads, “AMD Opteron servers crash.” Remember the media headlines when a machine crashed on Bill Gates?
If that happens, the Intel people will have problems sitting back quietly because they’ll be near- if not fully orgasmic. Just try to persuade conservative server stakeholders to abandon Intel when the media is telling the world that your products crash.
Dumb, dumb, dumb.
With relatively few exceptions, superficial, shallow media is like the weather. You adapt to it; it doesn’t adapt to you. Maybe not too many members of the press have a deep understanding of what they’re covering, and what is or is not important.
But all of them understand concepts like not telling the truth, or covering up, or machines failing, and they’ll all write about what they do know.
All of this has zero significant meaning. But perception is often more important than reality.