Well, maybe not me, but perhaps my toys are. I travel a lot, and if I had to pick one toy out of the 3,457 things in my laptop bag that I could take on a trip, it would have to be my iPod Classic (that of course assumes that I get to take the Bose headphones with me as well because you have to be able to listen, right?)

The iPod Classic has a feature called “Genius” that takes any song and creates a playlist of 24 additional songs that are similar to the one that you started with.  It’s pretty scary how often it is dead on in selecting the best adjacent music.  Sometimes its downright creepy, choosing “That Was Another Country” by the Innocence Mission when I was strolling down the streets of Kowloon one evening or the time it chose something from “Cheap Trick, Live at Budokan” on a flight to Tokyo. (Yes, I grew up in Chicago in the late 70’s; I’m not apologizing for that.)   I am going to assume that it was pure coincidence and not some nefarious GPS-enabled trick being played on me.

When we bring out our Istanbul processor, expected for release in the second half of this year, we plan to introduce a new feature called “HT Assist” that can give you much better throughput over the HyperTransport^TM technology connections by reducing the amount of traffic generated by the processors in seeking the shortest path to data that they need.

One of the best analogies that I have heard about this was that it is like a rifle shot versus a shotgun blast.

Processors in a system are always looking for the shortest distance to the data that they need.  The fastest location is always their own cache. After that, the caches of the other processors is the next logical location, before moving on to the memory, or, hopefully not, the hard drives. Like a set of concentric circles, each data location is higher in latency as you move away from your own local data set.

Processors spend time sending out probes to ask if other processors have the data they need in their cache.  But you have to ask every processor and each processor has multiple cores and multiple caches.  Yep, you guessed it – shotgun blast.  It’s like shouting our “Hey, anybody have that file I need” in a crowded office.  When one person does it, it can be disturbing, but when everyone does it, it gets really annoying.

HT Assist tracks data in caches and helps direct the processor directly to the data in the other processors’ caches. The rifle shot instead of the shotgun blast.

So, how does this help you?  Well, by reducing the amount of bus traffic, we expect to see servers run more efficiently and scale more linearly as application growth increases.  And improving server efficiency can help in many ways, such as getting the job done sooner, which can allow those hardworking CPUs to relax and reduce power draw for a bit, helping to keep your data center cooler.

And THAT is genius.

John Fruehe is the Director of Business Development for Server/Workstation products at AMD.

His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.