Opteron jumps to HP clusters and beyond

Palo Alto (USA) – Following the strategy that recently saw it alongside the CPU line of its partner, Intel, the 64-bit processors of AMD, HP has launched a new line of servers, the ProLiant DL585, on the market. capable of supporting up to four Opterons and 64GB of memory.

Compared to the Opteron-based line of dual-processor systems introduced by HP in February, the new servers push AMD’s chips into the family of high-performance computers that HP dedicates to technical computing and server clusters.

The base model of ProLiant DL585 costs $ 8,299 and adopts two 1.6 GHz processors and 2 GB of memory; the top model costs $ 22,396 and includes four CPUs and 4GB of memory.

The Opteron processor, which in these days is celebrating its first birthday, has become one of the protagonists of the challenge between HP, IBM and Sun – three of the largest server manufacturers in the world – to conquer the market for high-end 64-bit systems. medium and low. Its biggest opponent, in the low-end range, will be Intel’s upcoming 64-bit Xeon, Nocona, expected on the market towards the end of the current quarter.

AMD prepares to respond to Intel’s move with the debut, in the third quarter, of the first versions of Opteron with 90 nanometer circuitry. Production of these chips began just this week at the Fab30 in Dresden, Germany.

Opterons manufactured with a 90nm silicon-on-insulator (SOI) process will have an area 40% smaller than current 130nm chips and will be able to support higher clock rates while reducing the consumption. The new manufacturing process is also expected to lead to cost cuts, especially when combined with the use of 300mm silicon wafers in early 2006.

According to some sources, AMD is expected to release three new 2.4 GHz Opteron models with support for 1, 2 and 8 processors next June.

For the time being, AMD has stated that it has no plans to migrate the Athlon XP to 90nm technology.

In recent days, the Sunnyvale-based chipmaker announced the opening of two new “Automated Precision Manufacturing (APM) Innovation Centers”, located in Austin, Texas and Dresden, respectively.

APM is AMD’s suite of over 250 automation and optimization technologies used to reduce microprocessor production times and lower costs.

The two new centers will be used by AMD’s engineers and software designers to integrate the new APM generation, version 3.0, into the Fab 36, the new 300mm wafer manufacturing facility currently under construction in Dresden.

“This first APM solution integration effort will allow us to make manufacturing processes within the Fab 36 fully automated, quickly and efficiently,” said Gary Heerssen, group vice president, AMD’s Corporate Manufacturing Group.

The current version of APM, 2.0, was created for the production of 200mm wafers and is currently operational in the AMD Fab 30 and FASL LLC Fab 25.

AMD explained that currently, within the Fab 30, APM 2.0 operates as a sort of “central nervous system”, forming a connection and control structure for the hundreds of tools located inside the factory. This sophisticated infrastructure constantly monitors the state of microprocessors throughout the manufacturing process, collecting and analyzing data from tool sets once the wafers enter and leave them. Thanks to these real-time data analyzes, APM automatically suggests the movements that the wafers will have to make inside the Fab and indicates any changes to be made by the different tools, in order to optimize the performance of the chips.

APM 3.0 will play a similar role within the Fab 36 but, according to AMD, will feature greater precision, integration and levels of automation.