New HP, IBM and SGI hardware

Barcelona, 24-10-1996 New machines from HP/Convex , IBM and SGI/Cray, are expected to find their way to European customers. A short update on these systems.

HP has introduced two scalable computing systems, named HP-9000 S-class and X-class, which are based on the new PA-8000 processors and are a continuation of the Convex Exemplar SPP systems. They are designed for data-intensive jobs requiring extremely fast time-to-completion combined with very high throughout capabilities. The S-class is used to streamline mechanical design, electronic design and petroleum exploration. As for the X-class, it is targeted at solving technical and scientific challenges for leading-edge technology. Both use a memory crossbar which provides high bandwidth, low-latency non-blocking access from CPUs and I/O channels to local memory. HP asserts that the exemplar system design can stream data to processors at rates up to 10 times that of competitive systems.

Additional information can be found through http://wsg.hp.com and http://www.convex.hp.com

SGI and Cray Research have released, in turn , a new family of high performance servers: Origin 2000, which are binary-compatible with SGI Challenge and Power Challenge systems, and drive computational world to the 21st century. Origin system offer many competitive advantages for WWW applications, Data Warehousing, decision support systems and is targeted at solving technical, commercial and scientific challenges.

In the table which follows we detail the main characteristics an performance of both companies products:

	HP		SGI
	S-class	X-class	Origin 2000	Cray Origin 2000
# CPU	4-16	16-64	1-64	65-128
Clock Frecuency (MHz)	180	180	180-195	195
Peak performance (Gflop/s)	11.5	46	25	50
Memory (GB)	16	64	128	256
Cache Size (MB)	4	4	1-4	4
Max. Memory Bandwidth (MB/s)	15	60	20	40
Max. I/O Bandwidth (GB/s)	1.9	7.6	40	80

Like IBM, the company has taken a major step in the technology evolution of SP nodes incorporating the next generation of Power 2 Super Chip. The new nodes are ideal for technical applications in areas such as engineering analysis, scientific research, seismic processing, chemical molecular modelling, and electronic design automation. These nodes are also optimal for commercial applications requiring fast uniprocessor performance because they offer comparable price/performance to 604 High Node. The new nodes can be added to existing frames. In addition, upgrades from existing nodes are also available. The 66 MHz wide Node, and the 77 MHz Wide node can be upgraded to the 135 MHz Wide Node , and the Thin Node and the Thin Node-2 can be upgraded to the 120 MHz Thin Node, which deliver 540 Mflop/s and 480 Mflop/s, respectively.

Anna Solana