IBM says Intel eServers trounce Dell, challenge Sun
Pot-shots at dawn
IBM has announced a set of benchmarks for its Intel servers which it says show big performance gains in Web applications over competitive systems, from Dell and Sun Microsystems.
The eServer x440 and x255 hold first and second places, respectively, in TPC-W results for a 10,000-item database, outperforming Dell systems by as much as 84 per cent.
The x440 database server, together with x330 systems (functioning as application, image and Web cache servers), delivered a performance of 14,356.2 WIPS@10,000 and price/performance of $32.04/WIPS@10,000. The result is 84 per cent better than that recorded with a Dell PowerEdge 6400 system (7,783.3 WIPS@10,000; $24.50/WIPS@10,000), according to figures released by IBM yesterday.
In the tests, The x440 database server was configured with four Intel Xeon Processor MP 1.6GHz/1MB L3 cache processors and 4GB of memory, and ran Microsoft SQL Server 2000 and Windows 2000 Advanced Server.
A figure of 13,531.2 WIPS@10,000, or 74 percent better than the Dell system, was recorded by a x225 server system in comparable tests.
Separately, IBM announced that in NotesBench iNotes benchmark a 4-way eServer x360 delivered 91 percent of the performance of an 8-way Sun Fire V880 - at 40 per cent of the cost.
According to audited results, the IBM x360 supported 6,750 users accessing iNotes mail, while the Sun V880 supported 7,500 users. The Sun machine required eight 900MHz UltraSPARC III processors and 32GB of memory to achieve only 9 per cent better performance than the IBM server, which used four 1.6GHz Intel Xeon Processor MP and 3GB of memory.
The x360 system cost $8.01 per user and $15.48 per NotesMark (transactions per minute), compared to the Sun V880, which costs $18.06 per user and $34.97 per NotesMark.
The Sun Microsystems Sun Fire V880 configuration tested was priced $135,471.51, compared to $54,068 for the IBM eServer x360 rig.
The iNotes Web Access workload executes Lotus Notes transactions that model a server for users who access their mail via the Web. The resulting capacity metric for a server is the maximum number of users that can be supported before the average user response time becomes unacceptable. ®