One of my blogs discussed the upcoming Bulldozer architecture, and the move of going from 3 ALUs and 3 AGUs per core to only 2 ALUs and 2 AGUs.

Someone commented that the ALUs and AGU's on the K7/K8/K10 architectures were shared, and therefore it is only 1.5 ALU and 1.5 AGU per core.
Well, that's wrong obviously. We've had some optimizations challenges right on this board, which did more than 1.5 ALU operations per cycle on K7/K8/K10. Even though the ALU and AGU may be shared units, it doesn't matter, since it can only dispatch 3 ops per cycle anyway. So what I said still holds: AMD's current architectures can do 3 ALU ops per cycle, or 3 AGU ops per cycle. Bulldozer may be able to peak at 4 ops per cycle, but only 2 of those can be ALU or AGU max. It can only do 3 or 4 ops per cycle if you have a blend of the two, where the current architectures can do 3 ops per cycle regardless.

Anyway, when I looked around, I found the source of this misinformation right here:
http://www.xtremesystems.org/forums/showthread.php?257927-AMD-s-Bobcat-and-Bulldozer&p=4523917&viewfull=1#post4523917

That's a post by JF-AMD: John Fruehe, Director of Product Marketing for Server, Embedded and FireStream products at AMD.
So either this guy doesn't know his own CPUs, or he's deliberately making their current CPUs look worse than they are, so Bulldozer looks more impressive. Which I think is rather weird.
Posted on 2011-08-13 16:19:11 by Scali
He's a marketing guy.
Posted on 2011-08-13 22:33:43 by Homer
Makes it all the worse, no?
Spreading false information about your company's products... I'm sure there are laws against that in many countries.
It's also strange... I mean, you work at the company, so you can easily double-check your information with the actual engineers.
Posted on 2011-08-14 05:55:52 by Scali

It's also strange... I mean, you work at the company, so you can easily double-check your information with the actual engineers.


If the person is a sales engineer, sure. If they area run-of-the-mill sales/marketing type, good luck with that.

People have a tendency to interpret and reiterate things at their own level of understanding. And if they happen to not understand something at all...
Posted on 2011-08-14 17:39:37 by SpooK


It's also strange... I mean, you work at the company, so you can easily double-check your information with the actual engineers.


If the person is a sales engineer, sure. If they area run-of-the-mill sales/marketing type, good luck with that.

People have a tendency to interpret and reiterate things at their own level of understanding. And if they happen to not understand something at all...


Did you bother to check the link in his sig?
http://blogs.amd.com/work/author/jfruehe/

He is basically the 'main man' when it comes to the server/workstation department of AMD (Opteron, director of marketing). He's the big guy who handles most of the press etc.
He writes long articles about Bulldozer for example: http://blogs.amd.com/work/2010/08/02/what-is-bulldozer/
He also does interviews, posts videos on YouTube etc.
So that is his job: he is the link between the press and the AMD engineers. That's why I find it so weird that he would make posts like that.
Posted on 2011-08-14 18:29:18 by Scali

(Opteron, director of marketing)


So he's the latter of the two groups. Mystery solved.
Posted on 2011-08-15 00:37:55 by SpooK


(Opteron, director of marketing)


So he's the latter of the two groups. Mystery solved.


Not sure if I understand your point (to me it was never a mystery that marketing people don't have the same level of technological understanding as engineers, hence the point of contacting engineers to verify their story).
My point is simple: if you're a marketing guy, it's your job to get your facts straight. In many countries there are laws against spreading false advertisements (AMD actually got sued my country: they advertised their NX bit in the Athlon64 to protect you from virii. They lost the case, had to take the commercial off the air, and pay a fine).
You rarely catch marketing guys on bold factual errors such as these. Certainly not the guys from Intel or nVidia, those are professionals.
Posted on 2011-08-15 06:18:14 by Scali
In today's news, we've discovered that all humans are capable of making mistakes. Film at 11.
Posted on 2011-08-15 18:46:08 by SpooK
Right... Not at all my point.
Posted on 2011-08-16 01:50:11 by Scali
At least now we know why Fruehe spread that information. It most likely was not a mistake. It seems I was right about thinking it was a way to cover up for Bulldozer's poor performance. They couldn't make Bulldozer look better than it was, so they tried to make the previous generation look worse.
Sadly for them benchmarks show that the previous generation looks better :P
Posted on 2011-10-15 16:00:53 by Scali



(Opteron, director of marketing)


So he's the latter of the two groups. Mystery solved.


Not sure if I understand your point (to me it was never a mystery that marketing people don't have the same level of technological understanding as engineers, hence the point of contacting engineers to verify their story).
My point is simple: if you're a marketing guy, it's your job to get your facts straight. In many countries there are laws against spreading false advertisements (AMD actually got sued my country: they advertised their NX bit in the Athlon64 to protect you from virii. They lost the case, had to take the commercial off the air, and pay a fine).
You rarely catch marketing guys on bold factual errors such as these. Certainly not the guys from Intel or nVidia, those are professionals.



Not fully true. Intel did have it's share in the U.S for false adverting. AMD was almost finishing up their first 8-core cpu. During this time frame. Intel releases their supply first 8-core cpu before AMD's.  They claimed it was faster then AMD's 8-core cpu's.

However, after the release about maybe 4 months after. AMD filed a lawsuit and won the suit. Claming Intel's false advertising. The CPU's actually was 2 quad-core cpu's smooshed together. So,  it's the same as if you bought 2 quad-core cpu's and put it on their own socket. They put this 2 on a single pcb board and have it connect to the motherboards main single cpu socket.  I am forgetting if the 2nd quad-core was a physical cpu or just software. I think it was on the software side but anyways the point was that it was pretty much a optimized quad-core cpu. 

So, Intel got fined. These 2 companies aren't perfect but they have both done some nasty things.  Like both were found guilty of monopolizing the market. They bought out Nvidia  and ATI. They then fixed the graphics cards prices to $500.

http://www.google.com/url?sa=t&rct=j&q=amd%20and%20intel%20found%20guilty&source=web&cd=3&sqi=2&ved=0CDgQFjAC&url=http%3A%2F%2Fwww.geek.com%2Fchips%2Famdintel-anti-trust-lawsuit-should-begin-to-move-561552%2F&ei=W9WEUaSKD-3h4AOy-oH4Bg&usg=AFQjCNGReusVBCZmbnlGg8kYn2SktUDUXg&bvm=bv.45960087,d.dmg
Posted on 2013-05-04 04:33:21 by hockey97
However, after the release about maybe 4 months after. AMD filed a lawsuit and won the suit. Claming Intel's false advertising. The CPU's actually was 2 quad-core cpu's smooshed together.


Do you have any sources for that? Because I don't know any Intel CPUs that put 2 quadcores together. I think you are either talking about two single-cores (Pentium D) or two dual-cores (Core2 Quad), or about quadcores with HyperThreading (4 physical cores, 8 logical cores).

So, Intel got fined. These 2 companies aren't perfect but they have both done some nasty things.  Like both were found guilty of monopolizing the market. They bought out Nvidia  and ATI. They then fixed the graphics cards prices to $500.


That has nothing to do with false advertising though.
Also, nVidia was never bought out.
Posted on 2013-05-04 06:18:08 by Scali