Intel should have made a stronger statement than it did about its Itanium processor at this week's launch of the "Westmere-EX" Xeon E7 high-end server processors, considering the bashing that Oracle has given the chip in recent weeks.
However – for whatever reason – Kirk Skaugen, general manager of Intel's Data Center Group, pretty much stuck to the party line. Except at one point where he hinted that some sort of Itanium news was on the horizon.
"If you like HP-UX, OpenVMS, and NonStop from Hewlett-Packard, or GCOS and ACOS from NEC, we continue to invest in Itanium," Skaugen said.
Skaugen added the mantra that Intel and HP have been repeating for the past several weeks: that Intel was on track to "double performance" with the future "Poulson" Itanium processors, and that the "Kittson" kicker was in development. "Itanium is on track for a two-year cadence," Skaugen said, confirming what most observers had been speculating was the case.
This means we should expect the eight-core Poulson chip sometime in March or April 2012. Skaugen wouldn't say exactly what the future Kittson chip will be, but he did say that it would come out approximately two years after Poulson. So we're into early 2014 for that chip, with at least a two-year life.
That puts us into 2016 for Itanium chips on the current roadmap, and before another Itanium chip would be necessary. And since HP has committed to at least ten more years of Itanium-based Integrity server sales running its operating systems, Kittson might have a very, very long life, much like the PA-8900s did – the last RISC chips made by HP. Or it might mean that another one or two Itanium processors will be available out beyond Kittson.
Kittson's scheduled 2014 arrival, by the way, reaches almost as far as Oracle's Sparc roadmap, which runs to 2015. That arrival date also provides a hell of a lot more detail than Intel is giving out on its Xeon chips (well, publicly anyway), and Advanced Micro Devices is revealing about its future Opterons, which we can see out through 2012 on the public roadmaps).
Unfortunately, we're somewhat in the dark about IBM's plans, as well, since Big Blue doesn't really provided public roadmaps any more. We know Power8 is in development, but we don't even know if there will be a Power7+ next year or if Power8 is coming on the three-year cadence IBM has set for major Power chip generations. So Power8 – whatever it is – should appear in 2013. But don't expect Big Blue to confirm that, or to commit to Power9 by 2015, either.
Oracle has been very clear about what it thinks the future of Itanium is. Two weeks ago, just ahead of HP CEO Leo Apotheker's first meeting with HP shareholders, the software giant announced that it had stopped development of future database, middleware, and application softwarefor Itanium.
Oracle added that it would continue to support existing software that it had created for Itanium platforms, and after some back and forth, Oracle accused HP of "knowingly withholding" information from their joint customers about the future of Itanium. "HP is well aware that Intel's future direction is focused on X86 and that plans to replace Itanium with X86 are already in place," Oracle said in a statement.
Red Hat pulled the plug on Itanium development back in December 2009, not adding support for Itanium-based servers with its Enterprise Linux 6 distro that came out last year. Microsoft said back in April 2010 that Windows Server 2008 R2 and SQL Server 2008 R2 are the last key back-end operating systems and databases that it will support on Itanium. Novell still supports its SUSE Linux Enterprise Server 11 on Itanium chips, but once Attachmate closes the acquisition of the company, it seems unlikely that SLES 12 will run on Itanium – unless HP and Intel shell out some dough to make it happen.
HP is the volume leader now among the few remaining Itanium system builders. Unisys, Silicon Graphics, NEC, and Bull will still sell you an Itanium 2–based server or an upgrade to an existing machine. But none of these vendors launched machines last year that support the quad-core "Tukwila" Itanium 9300 processors supporting Windows or Linux workloads. And all of them have launched high-end Xeon 7500 and now Xeon E7 machines that they're pitching as perfect boxes for supporting big back-end databases – the job that Itanium was designed to do and do well.
NEC and Bull are interesting cases. The Japanese server maker is still using its existing Itanium-based platforms to sell ACOS mainframe platforms, which were ported over to Itanium years ago, just like HP/Compaq ported HP-UX, OpenVMS, and NonStop to Itanium. But back home in Japan, NEC is not quite done with Itanium. Bull did upgrade its NovaScale 9010 GCOS mainframes last May with Tukwila machines, and will continue on with a Poulson box at some time in the future.
"NEC Japan will do future Itanium machines based on the QuickPath Interconnect and using NEC chipsets," Mike Mitsch, general manager for the IT Platform Group at NEC America, tells El Reg. "We're fully committed to that product and we will roll it out."
Mitsch was not at liberty to say if future NEC Itanium servers would be based on the current Tukwila Itaniums or the future eight-core Poulson chips. He did add that the machine was really intended for Japanese customers using NEC's mainframe software, and would probably only be sold outside of Japan in those instances where Japanese-based multinationals required it. The company has no plans to support Windows or Linux on these machines, even though it is technically possible.
With the Intel Developer Forum being hosted in Beijing on April 12 and 13, perhaps Intel will say a bit more about the future of the chip and its platform.
Skaugen gives that impression. "We'll have announcements in the near term that will ratchet up the commonality between Xeon and Itanium," he said at the Xeon E7 launch earlier this week.
That is sure to set the tongues a-wagging. Starting with the Itanium 9300 and Xeon 7500, many of the machine check architecture (MCA) high-availability features that were previously only in the Itanium chips were ported to the Xeon chips. With the E7, more were etched into the Xeon silicon. The Itanium 9300 and Xeon 7500 processors already use a common "Boxboro" Xeon 7500 chipset and a common "Millbrook" memory buffer and DDR3 memory board setup.
So what's left? The socket, perhaps?
Perhaps not. Intel delayed the Tukwila chips so that the Tukwila, Poulson, and Kittson chips could plug into the same LGA 1248 socket. So maybe the then-current top-end Xeon socket when Poulson or Kittson comes out - Ivy Bridge-EX, perhaps? - will have a variant that will plug into the Itanium socket. Or, perhaps Intel is changing the plan and porting Kittson to the future Xeon socket.
Whatever Intel is up to, we'll let you know as soon as we find out. ®