Diamond Plus is the enhanced edition of process computers based on Intel technology.  It was first created mid last year in response to Intel introducing a new computing platform that sat between value systems (Diamond) and performance systems (Superhawk).  This new platform has integrated graphics and falls into the Diamond category.  We have conveniently called it Diamond Plus.  This situation is similar to Superhawk and Superhawk Plus for two renditions of performance platforms without integrated graphics. 

Why do we need two renditions of each platform?  The reasons are in performance, price, and architecture differences. 

Diamond is based on G41 chipset of Penryn descent with 775pin.  Diamond Plus was based on H55 of Westmere descent with 1156pin.  Intel changed its mind recently when releasing Sandy Bridge to succeed Westmere with 1155pin.  This is our latest Diamond Plus.  Intel further introduced more fragmentation of this platform with 2 chipsets- H61 & H67.  Whilst we have adopted H61 for Diamond Plus, should we bother to add another rendition called Diamond Plus Square?  This is a balancing act between fitness for purpose (fragmentation) and total cost of ownership (consistency).  We have identified the sweet spot and managed the spread and transition of system platforms well.  See this for the new Diamond Plus based on Sandy Bridge.       www.compucon.co.nz/component/option,com_wrapper/Itemid,245/

The supply of motherboards with a good chipset for supporting Intel Sandy Bridge processors has resumed.  Superhawk will land on this Bridge starting from 21 March 2011. 

Depending on our stock holding, we may have Superhawk 1156 and Superhawk 1155 offerings in parallel for a brief period.  The numbers refer to CPU pins- 1155 is Sandy Bridge and 1156 is Westmere.  The new model is about $60 more expensive than the old model and this is a reflection of more performance.   Do not compare the CPU frequency (GHz) or amount of L3 Cache if they have different pins.  Sandy Bridge has a better utilisation of processor resources than Westmere.  The top performer is still 1366 for Superhawk Plus (Nehalem) and the value segment is still 775 for Diamond (Penryn).  Find info on Sandy Bridge here: http://www.compucon.co.nz/content/view/874/57/

Diamond Plus on 1156 is the next system model to change to 1155.  It has integrated graphics whereas Superhawk has discrete graphics.  The evolution will take place as soon as the new chipset and motherboard are available.

Workgroup Server is also based on 1156 pin processors for single socket.  We estimate its evolution to take place in mid April timeframe. 

We have continued to do more benchmarking of Quadro cards since the last CPD disclosure and have continued to discover more open secrets.  Quadro Fermi is indeed a new technology that is not well understood by many people.  We will take a bit more time to organize our findings and to continue this discovery journey.  There is however some information that is stable for release.  For this purpose, we have created a new home page for Engineering Workstations on our public website:  http://www.compucon.co.nz/content/view/794/253/.

The information there is incomplete for now and more will progressively be posted.  We plan to release pricing info in April and will talk about the discoveries in the April CPD Seminar.

A major discovery is that not all CAD application packages are tuned for Quadro Fermi or CUDA 2.  Once we have installed the latest drivers, application performance goes up to 300%.  This discovery applies to the most popular CAD software in the world. Do the big boys know how fast the world is changing?   GeForce may have the same CUDA cores as Quadro, but the lack of tuned Fermi application drivers is a major deterrence to its use for professional applications. 

As disclosed in the CPD seminar on Wednesday, we have started a new journey into Quadro and have found many open secrets.  Quadro is a brand of GPU chip designed developed and produced by nVidia.  Fermi is the name of the latest nVidia GPU architecture based on CUDA version 2, DDR5 memory, Giga Thread algorithm etc.  It is faster than the previous generation in terms of GFLOP (computational power) and is up to 5 times faster in terms of Triangles per second (display power).  This technology is new and will propel real life applications by a quantum leap such as for visualization of scans of the human body or oil field, creation of digital contents, and simulation of engineering and building designs. 

Compucon has done several weeks of performance benchmarking as an effort to discover the sweet spots of system configuration to achieve fitness for purpose and total cost of ownership.  Our current effort is the culmination of industry development monitoring in the area of GPU performing CPU tasks.  With up to 512 cores, GPU is bound to perform much better than 4 or 8 cores CPU for scientific and graphical functions.  We will take a while to complete our due diligence process but feel comfortable to publish some preliminary information on our website to assist our channel partners and customers.  Our effort will continue for the next few weeks. Please see this: http://www.compucon.co.nz/content/view/794/29/