Thursday, October 13, 2011

AMD Bulldozer: AMD launches it's remake of the Pentium 4 story



Karl Marx is often paraphrased saying: "History repeates itself: the first time as a tragedy, the second time as a farce."  I reflected on this saying as I read the details of AMD's Bulldozer technology.

Once upon a time, Intel introduced a processor called the Pentium 4.  This chip played many of the same gambits AMD's Bulldozer architecture is predicated on.  Pentium 4 came out of the gates stumbling in the year 2000.  It cost more than its predecessor, consumed more energy, generated more heat, and it failed to out-perform its predecessor on a consistent basis.  A loud chorus of boos went up from the fan-base.  I was one of the guys booing.  On a few exotic media benchmarks, the P4 was better than P3, but these were not the applications that concerned most people at that time.

Eventually, the Pentium 4 had a brief day in the sun.  Let's mark that epoch as the years 2003-2004. By 2005, the Pentium 4 had failed.  In 2005, AMD shot-past Intel in all the real-world benchmarks, not to mention the esteem of all the hardware geeks on Earth.  It was a route from which no honor could be salvaged.  Even Dell, a once-proud Intel-only vendor, was forced to open up and use AMD processors.  Intel was forced to abandon the Pentium-4 Netburst architecture and pursue the 64 bit multi-core course AMD set for the entire industry.

Unfortunately, I think AMD apears doomed to repeat this history.  Bulldozer has come out of the gates stumbling.  If you expected AMD to decisively take back the performance throne from Intel in every phase of the game, you are going to be very sadly disappointed.  It's not clear that Bulldozer can take the performance throne in any phase of the game... at all.

It does not outperform its predecessor on a consistent basis.  It is faster at running a small number of exotic multi-threaded apps, but slower at most important tasks.  One of the non-exotics is 7-zip, and that is nice, but how much does that mean in the scheme of things?  Bulldozer consumes more energy--125 watts versus 95--and in consequently generates more heat.  It's not cheap either.  Intel and AMD both offer cheaper alternatives that perform a little bit better on average.

Shouldn't the Bulldozer trigger the same chorus of boos we heard at the Dawn of the Pentium-4?  I think so.  I am not happy about it either.

What does this mean for the future survival and success of AMD?  Nothing good.  The trouble is that AMD is absolutely not the performance leader of the industry in the year 2011.  They are marginally behind Intel in just about all performance categories.  AMD can ill-afford to fall further behind Intel on a failed strategic blunder.

There is one bright spot for AMD:  It looks like a pretty good server technology.  Bulldozer seems like a very good SQL/NoSQL processor.  It also looks like a good webserver chip.  This could be AMD's salvation.

As we all know, the industry is bifurcating into consumer and producer market segments.  This is the most important concept underlying what some people call the Post-PC era.  There are pure consumers, who produce no information or entertainment content.  These folks are moving more and more towards smart-phones and tablets.  There are also producers of good and services in the web world.  These people usually want big 'n bad desktops, especially if they do 3d or video work.  They also have needs for webservers and SQL Servers to deliver their content.

Eventually, Bulldozer could become a force on the production and delivery side of the fence.  Unfortunately, it doesn't look like that epoch will begin today or tomorrow.

Have a look at the following net video from TigerDirect.com, but be aware that they have an agenda to sell you hardware.