A few weeks ago AMD presented its updated server strategy to the financial analysts. The big message – it’s all about the volume market, and we plan to target the “sweet spot” – where the majority of the server sales take place.

In my 17 years in the server business, I’ve never been more excited about an upcoming year as I am for 2010 because of how we are targeting that volume market.  How serious are we about the volume market?  When you look at this volume market, you see that customers’ needs line up neatly along our product strategy. 

The top half of this market is driven by scalability and performance.  Virtualization, database, business and compute intensive applications- you know the drill.  The software that runs your business requires more performance and better efficiency.  More cores and more memory channels drive these applications and price/performance-per-watt is now king.

With the AMD Opteron™ 6000 Series platform we’re planning the introduction of aggressive and scalable 2P systems to tackle the top of the volume market, along with a new category of 4P servers.  Imagine a world where 4P servers in optimized configurations can finally start to challenge the top end of the 2P space, instead of the other way around.  What if you can get 4P performance with economics that more accurately reflect the 2P market?  Well, a new category evolves and allows our technology partners to actually grow the 4P market instead of watching it shrink as the 2P market nips at its toes. You won’t be talking in terms of sockets anymore – but rather in terms of core and memory capabilities.  Take a look at this excerpt from a recent AnandTech story by Johan De Gelas:

When it comes to expensive 2-socket servers, AMD positioning is cunning. In the midrange we will find servers with sixteen Opteron cores (2-socket x 2-quad-core die per socket) offering 8 memory channels and 24 DIMM slots. Performance will probably be “close enough” to the Westmere EP servers, which can only offer six memory channels and 18 DIMM slots. The extra amount of memory bandwidth might make a dual Opteron 6100 attractive to the HPC folks, while the higher amounts of DIMM slots together with a competitive price may very well convince the virtualization market.

Two socket AMD Opteron 6100 series platforms, with up to 12 cores and 4 memory channels per processor are poised to be an extremely strong competitor for the bulk of the 2P market workloads.

For the lower half of the volume market, we’ll have new AMD Opteron™ 4000 series platforms that are designed to enable low cost, low power systems that don’t compromise server functionality.  If you are looking at the emerging world of cloud computing, either internal or external, you know that a low cost platform with high power efficiency is the goal.  Obviously if it is your budget, you’ll want to make sure that you have AMD Opteron 4000 platforms, but if you are outsourcing your cloud services, make sure that your provider is using these platforms. Why does that matter? Because if you outsource, you’re costs are dependent on your partner’s cost structure.  And cost effective processors like these help drive that down for you.  It’s worth asking.

As a small or medium business, those AMD Opteron 4000 platforms enable low costs but without compromising the things that you really need in a server.  Like scalability with 2 processors, or ECC memory to help protect you from memory errors. We don’t believe you should be penalized because of your size.

A peek at 2011

2011 brings processors based on the “Bulldozer” core which will be drop-in compatible with the AMD Opteron 6000 and 4000 series platforms we plan to launch in 2010. This core is designed for a future loaded with concurrent applications.  Ask any software vendor today what the future looks like and you will probably hear a pretty consistent message – more concurrency, more simultaneous commands, and greater threading. The days of the single threaded application are nearing an end, if we haven’t gotten there already. This ties in perfectly with our strategy for delivering even more performance and scalability while also maintaining the power efficiency that people have come to expect from AMD.

Bulldozer is based on a modular architecture where two integer cores are teamed up with an extra-large FPU to create what we call a “Bulldozer” module today (however don’t get too used to that name as it will have a different one at launch time – planned for 2011.) Bulldozer modules are the basis of all of the designs that will be coming from this architecture, and it’s modular nature not only allows us to build processors with different sized core counts (like the 8-core Valencia and the 16-core Interlagos.) The Bulldozer module is a concept and part of an architectural design; it is not something that the user will come in contact with.  For instance, when an “Interlagos” system boots up, the hardware will see 16 integer cores, not 8 modules.  When the OS loads, it will see 16 integer cores, not 8 modules, and the applications will see 16 cores as well. 

The Bulldozer architecture can provide up to 80% greater expected throughput when running 2 threads simultaneously compared to a single thread running on a single integer core.  Our engineers estimate that the amount of discrete circuitry that is added to each Bulldozer module in order to allow for a second integer thread to run only adds ~12% additional circuitry to each module, which translates into only ~5% of circuitry to the total Bulldozer die.  We believe this is an excellent balance of greater performance with a very small silicon cost. The goal of the shared components is to help drive down power consumption.  When you consider that our 16-core Interlagos is being designed to fit in the same power/thermal environment as a 12-core Magny Cours, it is clear that we’ve made some good choices around the power optimization – without sacrificing performance or features.

Now that you have seen how the Bulldozer modules are built, how do those fit into a die?  Multiple dies are tied together, sharing a large level 3 cache, as well as other components. The diagram below is for an 8-core Valencia die.  

Just as multiple Bulldozer modules can be integrated into the die, because they are modular, in the future other modules, such as GPUs, could be integrated into the architecture.  This gives us additional flexibility to help meet customers’ needs.

So, as you can see, we feel pretty confident about how our offerings stack up in 2010 and beyond.  With two new flexible platforms that will do a great job of covering the market from low end 1P to high end 4P, we will have a compelling value proposition. And with a heavy focus on the “sweet spot”, we’ll be bringing new levels of value to a market that is poised for growth.

John Fruehe is the Director of Product Marketing 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.