IBM Packs More Performance into its UNIX POWERparallel System

Lower entry price, more robust system

SOMERS, N.Y.--(BUSINESS WIRE)--February 7, 1995--IBM today announced hardware and software enhancements to the IBM Scalable POWERparallel Systems SP2(a) that can improve price/performance by more than 15 percent for both scientific/technical and commercial users and provide them what they have asked for -- an industrial-strength, high-end UNIX(b) operating system platform that provides higher reliability, availability and mainframe interoperability for use in business-critical environments.

The SP2 is IBM's general-purpose, high-performance parallel processing computer that runs AIX(a), IBM's UNIX operating system. Its price/performance is improved for numeric-intensive applications -- such as drug design, simulation of automobile and airplane acoustics and prediction of oil drilling sites -- as well as for commercial applications ranging from those normally associated with running a business to emerging new uses like decision support that analyzes customer buying patterns.

Today's announcements include:

o A new "Thin Node 2" processor the size of a breadbox that fits into an SP2 cabinet and provides the "number crunching" performance of bigger "Wide Node" processors at a lower price -- a boon for technical applications. Benchmark numbers indicate that the Thin Node 2 -- compared with its current, entry-level "Thin Node" counterpart -- performs an average of 40 percent better for most numeric-intensive tasks and 250 percent better on the popular LINPACK DP(c) (Double Precision) benchmark test.

o A reduced price for the entry-level SP2 processor -- the Thin Node -- to the lowest level ever, enabling customers eager to enter the world of UNIX parallel processing to purchase an SP2 for under $138,000. This new low system price provides over 10 percent improved price/performance for commercial applications such as transaction serving. In addition, IBM has made available an optional level 2 (L2) cache -- or temporary storage area -- for the Thin Node processor that improves performance over 15 percent.

o Benchmark numbers that indicate an SP2 with 32 Thin Node 2s substantially outperformed the Cray T3D(a) on the NAS benchmark suite(d).

o Software that provides higher availability and reliability for business-critical applications in such industries as banking and retail.

o Enhanced software for improved interoperability and high-speed data transfer between IBM and IBM-compatible mainframes and the SP2, as well as associated UNIX workstations. Jobs may be submitted and data easily accessed between platforms.

o Broad printing options providing attachment of high-speed IBM printers normally associated with IBM mainframes.

o Support of industry-leading IBM network management software products so the SP2 may be easily integrated into and managed as part of an open, heterogeneous network of IBM and non-IBM mainframes, workstations and PCs.

The new "Thin Node 2" processor is based on IBM's POWER2(a) architecture RISC microprocessor -- the "brain" of the system. With today's announcement, users can configure their SP2 systems with new Thin Node 2s instead of the current, larger Wide Nodes, if they don't need the Wide Node's extra memory sizes and input/output (I/O) slots, but want a similarly powerful parallel system in a more compact package at a lower price.

Parallel processing on the IBM SP2 links together from two to 512 IBM RISC System/6000(a) (RS/6000(a)) processors to perform numeric-intensive and data-intensive computations. The processors work on different parts of a problem at the same time, enabling the high-performance computer to support many users and analyze massive amounts of data at blazing speeds. The flexibility of the SP2 allows customers to run both parallel and non-parallel applications at the same time.

IBM also announced that initial test installations have begun for the recently announced POWERquery for the SP2(a), a cost-effective, large-scale decision support solution for business. The power and flexibility of POWERquery's parallel processing capabilities will enable customers to extract strategic information from vast amounts of accumulated data.

POWERquery is a set of highly scalable and flexible building blocks of IBM hardware, software -- including IBM's new parallel relational database, DATABASE 2(a) Parallel Edition for AIX(a) -- and consulting services that can be tailored to satisfy a customer's specific large database query and decision support needs. IBM intends to begin volume shipments of POWERquery this summer.

The use of POWERquery can result in competitive advantages including more successful marketing strategies and more efficient resource allocation. For example, POWERquery can help marketers to more effectively analyze the buying patterns of their customers. POWERquery's parallel processing power will allow these users to identify increasingly narrow market segments, down to the smallest segment of all, the individual.

IBM also announced that high-availability software developed for the RS/6000, and used successfully by RS/6000 customers, will become generally available in March on the SP2. The new software -- AIX High Availability Cluster Multi-Processing/6000(a) Version 3.1.1 (HACMP/6000 3.1.1) -- allows up to eight SP2 nodes to be configured into a single, highly available cluster. In addition, multiple independent eight-node clusters may be supported on a single SP2 system. HACMP clusters can be configured to eliminate single points of failure -- in other words, if one system component fails, its work is taken over by another system component. This is a necessity for customers like banks and retail stores that cannot afford "down-time" for their business-critical, on-line transaction processing applications.

This new version creates one of the richest high-availability offerings in the UNIX marketplace, without incurring the added expense of fully redundant hardware, and offers customers unmatched flexibility in designing systems that best meet their performance requirements.

IBM also announced SP2 support of additional system management products designed to enhance the ability to manage the SP2's powerful computing resources and allow it to be integrated into existing computing environments. NetView(a) for AIX, Systems Monitor for AIX and Trouble Ticket(a) for AIX are current AIX tools that have gained recognition as leading network management solutions. These tools may now be used to manage the SP2 seamlessly as part of a network of PCs, workstations and mainframes, or to allow an SP2 node to perform these management functions.

"We could not be more pleased with the positive reception that our Scalable POWERparallel Systems SP2 has received in the marketplace," said Dr. Irving Wladawsky-Berger, general manager, IBM POWER Parallel Division. "The most exciting thing of all, however, is the huge `crossover' that is taking place between technical and commercial computing. The technical world finds itself with increasing amounts of data to be managed, and the commercial world increasingly is applying sophisticated mathematical techniques to analyze vast amounts of information. We are in the fortunate position of having an excellent product family that makes these transformations possible."

IBM is a proven leader in the high-performance computing marketplace where commercial interest is rapidly growing. By the end of 1994, the first full year of IBM POWERparallel Systems shipments, more than 350 systems had been shipped worldwide. More than 120 systems are being used in commercial environments.

Benchmark Data

The Thin Node 2 performed 250 percent better than the current Thin Node on the LINPACK DP benchmark. An SP2 with 32 Thin Nodes outperformed the Cray T3D on both the LINPACK HPC(c) -- where the SP2 performed at 6.4 gigaflops (GFLOPS), or 6.4 billion calculations per second, and the T3D performed at 2.7 GFLOPS -- and the NAS benchmark tests.


Performance as Ratio to Cray C90/1 Test (no.)Nodes SP2 T3D EP 32 9.7 2.6 IS 32 2.0

0.5 FT 32 2.4 N/A CG 32

1.8 0.4 MG 32 4.4 1.2 BT

32 5.3 2.6 LU 32 4.6

1.4 SP 32 3.9 1.5

Note: Higher numbers indicate better performance. Numbers in tables represent the comparison to a Cray C90/1 processor. For example, in order to get the same performance on the Class B BT benchmark as a 32-node SP2, you would need over five Cray C90/1s.


The new Thin Node 2 processor becomes available to customers on March 3. Customers can order the enhanced Thin Node with L2 cache today.

IBM has once again lowered the cost of parallel computing by reducing the price of the current Thin Node. Customers can now order by special request, an entry-level SP2 system in a 49-inch compact frame. Its U.S. list price -- with two Thin Nodes, not including L2 cache; 128 MB of internal memory; two GB of internal disk; the low-cost High-Performance Switch; and basic systems software -- starts at $137,400.

The price of the SP2 scales, or rises, in line with performance. For example, an SP2 in a 49-inch compact frame, with eight Thin Node 2s, 512 MB of internal memory, eight GB of internal disk, the low-cost High-Performance Switch and basic systems software can be ordered for just under $500,000. This is compared to an SP2 in the full-sized frame, with 16 Thin Node 2s, one GB of internal memory, 16 GB of internal disk, the regular High-Performance Switch and basic systems software for just over one million dollars.

POWERquery offers a high level of configuration choice. Customers can tailor a selection of hardware, software and service options to meet their specific application requirements. As a result of this flexibility, there is not a specific price point that is representative of a "typical system" price. However, as an example, a base POWERquery configuration -- including a 10-node SP2, high-speed switch, 240 GB of disk storage and basic software; IBM's DB2 Parallel Edition; and related planning and installation services, would be priced just over $1.8 million.

IBM Credit offers an array of convenient and flexible financing options for IBM POWERparallel systems. Rates will vary by term and customer credit rating. IBM marketing representatives can provide further information.

(For more detailed technical information, contact Rob Anderson at 212/696-2000.)

IBM POWER Parallel Division

IBM produces world-class scalable parallel information and computing systems for commercial and scientific/technical customers. The IBM Scalable POWERparallel Systems SP2, an integral part of the IBM RS/6000 line, features design and performance leadership and offers exceptional reliability and versatility. IBM's recently announced POWER Parallel Division is headquartered in Somers, NY.

Editors' Note: IBM news releases are available on the Internet, via the IBM Home Page available through web browsers at

(a) Indicates trademark or registered trademark of International Business Machines Corp. Other product names may be trademarks of their respective companies.

(b) UNIX is a registered trademark licensed exclusively by X/Open Co. Ltd.

(c) The LINPACK benchmark suite compares the performance of different computers in solving dense systems of linear equations. LINPACK routines are compiled on a vendor's own FORTRAN compiler in an effort to approximate what a typical user would experience. The LINPACK Double Precision benchmark measures a system's double precision floating-point performance. The LINPACK HPC benchmark provides a way to compare massively parallel computers. Dr. Jack Dongarra at the University of Tennessee (and Oak Ridge National Laboratory) oversees the administration of the LINPACK benchmark.

(d) The NAS benchmark was developed through the Numerical Aerodynamic Simulation Program at the NASA Ames Research Center in Mountain View, Calif. The NAS Program is a large scale effort to advance the state of computational aerodynamics. The NAS Benchmark is a suite of benchmarks which consists of five kernel benchmarks and three simulated application benchmarks--LU Solver (LU), Scalar Pentadiagonal (SP) and Block Tridiagonal (BT). This suite mimics the computation and data movement characteristics of large-scale computational fluid dynamics (CFD) applications.

Performance measurements for the IBM POWERparallel system are the result of tests done in a laboratory environment at IBM Kingston using Message Passing Library (MPL). While these values should be indicative of machine performance, no warranties or guarantees are stated or implied by IBM. These measurements are offered only as an indicator of performance. Competitors' performance data are from David H. Bailey, Eric Barszcz, Leonardo Dagum and Horst D. Simon, "NAS Parallel Benchmark Results 3/94, RNR Technical Report RNR-94-006, March 21, 1994" and update of 10/94.