August 28, 2012, 3:00 PM — IBM's latest mainframe, the zEnterprise EC12, is big on data analytics and hybrid clouds.
This system, announced Tuesday, includes a new 5.5-GHz, six-core processor, versus the 5.2-GHz, quad-core processor that shipped with the zEnterprise 196, announced two years ago last month.
This may be the world's fastest commercial processor, say analysts. Despite this, it is not as big a jump in sheer clock rate as some of the earlier leaps in mainframe CPU speed.
Regardless, IBM says this system has 25% more performance per core and some workloads will see performance gains by as much as 45%.
The trend among system and chip makers has been to add cores and turn to software parallelism to improve performance. But Jeff Frey, the CTO of the System Z platform and an IBM Fellow, says they aren't trading off single-thread performance in the mainframe.
"We continue to press on single-thread performance," said Frey, who added that it remains important for customers who have applications that execute processes serially, such as batch applications.
This latest chip was produced at 32 nanometers, versus 45 nanometers in the earlier system. This smaller size allows more cache on the chip, in this case 33% more Level-2 cache. The system has doubled the L3 and L4 cache over the prior generation, said Frey. The system can support as many as 120 cores.
Joe Clabby of Clabby Analytics, who was briefed by IBM on the new system, said the increase in cache is particularly important for improving performance. "It's now gotten better at data intensive workloads," said Clabby. "The closer you can put the data to the processor, the faster it can be executed."
The zEC12 has 3TB of system memory, similar to z196, but also adds flash memory called Flash Express, with a maximum capacity of 6.4 TB, to improve system performance. The memory is easily configurable and protected with 128-bit encryption, said Frey.
Initially, the flash memory will be used internally for efficient paging of virtual memory, diagnostics and better performance of workloads, said Frey.
But he said that in the future, DB2 and Java will have direct exploitation of the flash memory, providing "huge improvements" in performance and scale of DB2, buffer pools and Java.
There's no timeframe for that capability, but once DB2 and Java are allowed to use flash memory directly, "you'll be able to have DB2 exploit very large in-memory databases with extremely good performance," Frey said.