Oracle's Ellison spins SPARC T4

Yesterday we reported how  Oracle has launched a new SPARC T4 chip  and shoved them under the bonnet of a line of servers.  Its Systems Executive Vice President John Fowler told the world+dog that they were five times better performers than anything on Dancing with Stars.

He said that the servers were being pre-ordered in droves to replace “outdated systems” from competitors (cough) – that is Intel’s Itanium.

What was the selling point was the fact that the T4 had finally caught up with the rest of the world and bought out-of-order execution (OOE) to the SPARC platform. This technology first appeared in the Power, x86 and Fujitsu SPARC64 processors back in the 1990s and meant that instructions run in a thread.

For some reason Sun did not develop the technology. Instead the T3’s architecture boosted multithreaded performance which made it OK for a Web server but rubbish when it came to databases. This meant that Ellison’s flagship software product would not run very well, and you could not have that.

Thanks to the “new tech”,  the T4 can run significantly faster than the T3 despite the fact that the T4 has half the processor cores per CPU.

However, the downside is that the T4 is slower than x86. This was surprisingly not mentioned by Ellision who appeared to pick his benchmarks with care when he announced the chip. Ellison repeatedly compared the performance of the T4-based Sparc SuperCluster to IBM’s Power line—and the Power 795. A one-rack T4 SuperCluster “is twice as fast as IBM’s fastest computer, at half the cost”.

The benchmarks appear to have mostly come from Oracle itself and only applied to software which had been tweaked for that processor.

Ars Technica noted that one of those third-party benchmarks was the TPC-H benchmark for a 1,000 GB load, in which the T4-4 beat the IBM Power 780 and Itanium-based HP Superdome 2 on price/performance, raw performance, and throughput.

However the same benchmark put the T4 at the bottom of the top ten for performance. It was bettered by x86 systems running SQL.

Writing in her bog, IBM Systems and Technology chief technical strategist Elisabeth Stalh pointed out that Oracle claimed nine T4 world records and yet seven of the nine are not industry standard benchmarks but Oracle’s own benchmarks, most based on internal testing.

“Oracle’s SPECjEnterprise2010 Java T4 benchmark result, which was highlighted, needed four times the number of app nodes, twice the number of cores, almost four times the amount of memory and significantly more storage than the IBM POWER7 result,” she wrote.


Still it is not all bad news. The smart money would have been on Larry Ellison getting bored with SPARC/ Solaris long before Intel lost interest in Itanium. Until now, if there is any migration taking place, it has been probably been from SPARC to Itanium. But those who love the SPARC, and cried buckets when Oracle bought Sun, should be jolly happy that he has not driven the technology to a deserted forest near Berlin and discreetely shot it in the back of the head.

Certainly the T4 will stop some defections, as companies seek to preserve their SPARC skill base, but we can’t see that the chip will win back customers lost from Intel, or convert IBM Power users.

Still it is proof that Oracle really does intend to invest in continuing Sun’s hardware and operating system business.