by Anton Shilov
05/03/2011 | 12:55 PM
The decision of Oracle to stop developing software for Intel Corp.'s Itanium microprocessors clearly put the future of the chip as well as HP-UX operating system under a huge question mark. While Intel officially claims that it has code-named Poulson and Kittson chips in the development, it has been relocating Itanium developers to Xeon products. An industrial source revealed this week that Kittson may indeed be the last major Itanium update.
"Intel Itanium 'Kittson', which is scheduled to come out after code-named Poulson will probably be the final Itanium processor delivered to the marketplace. [...] There is a slight possibility that there could be a post-Kittson processor, but no other significant performance or feature enhancements would be delivered," a source, which wished to remain anonymous, with knowledge of the situation told X-bit labs.
An Intel official said that he would not comment on rumours claiming that Kittson is the last and final Itanium processor. Still, he noted that Kittson was in the development.
"Itanium's public roadmap already stretches further into the future than our competitors' project[s]. We have not announced a launch date for Kittson but new Itanium processors are on an approximately two-year cycle," said Patrick Ward, a spokesman for Intel.
Earlier this year Intel officially unveiled its next-generation Itanium processor code-named Poulson, which is due in 2012, and officially confirmed that it would be followed by Kittson central processing unit (CPU). Given the fact that Intel plans to release new Itanium chips with two-years cadence, the Kittson would be released in 2014.
Chief executive of Intel also indicated that the company had "multiple generations of chips currently in development and on schedule", nonetheless, the world's largest chipmaker did not publicly named the successor for Kittson.
HP, Intel's main partner on the Itanium project, said that it would "continue the development and innovation of Itanium-based Integrity server platforms with its HP-UX operating system using a roadmap that extends more than 10 years" and would also "continue to support customers running existing versions of Oracle software on Itanium-based Integrity servers, both existing and future platforms, during the same timeframe". Ten years, however, does not mean that HP does not have plans to include Intel Xeon E-series support into HP-UX 11i v4 or v5 operating systems after end of sales of the v3 in 2014.
HP will hold Discover 2011 forum in early June in Las Vegas, Nevada, where it may update the roadmap for the HP-UX as well as Intel Itanium.
Despite of rather optimistic promises by Intel and HP, it should be noted that industrial support for Itanium is getting low. Both Microsoft and Red Hat have already stopped developing software for Itanium. Large makers of servers, Dell and IBM, dropped Itanium back in 2005. In early 2011 Intel discontinued support for Itanium in its C/C++ and Fortran compilers.
For HP, Itanium is important as it allows it to sell its own HP-UX operating system. This allows the company to continue serving it clients who use mission-critical systems that it has been serving with HP-UX and PA-RISC or FOCUS processor-based systems since the mid-eighties. A natural thing for HP to keep those customers is to redesign HP-UX for x86 and offer a clear path for transition. HP clearly has plans to release HP-UX 11i v4 and v5 OSes in the next several years. Perhaps, it will have to add x86 support into the next-gen operating systems. Of course, HP will still depend on Oracle when it comes to software: the latter does not seem to want to support HP-UX and even non-Oracle Linux operating systems.
Organizations using business-critical machines demand clear hardware and software product roadmap as well as stable support. Those large customers often run software designed by Oracle and therefore the lack of Oracle DB for HP-UX and other software will either catalyze them to migrate to IBM DB2/IBM AIX-based machines or try to migrate to SAP and Linux. In any case, for many Oracle's decision to drop support of Itanium and uncertainties of the longer-term plans of HP and Intel will cause headaches.
For Intel, maintaining support of Itanium is largely a question of reputation on the market of ultra high-end servers. The company sells a lot more Xeon processors than Itanium chips. Moreover, top-of-the-range Xeon chips (E7-4870, 10 cores/20 threads, 2.40GHz, 30MB cache 130W) are now more expensive ($4394) than highest-performing (9350, 4 cores/8threads, 1.73GHz, 24MB cache, 185W) Itanium microprocessor ($3838). Intel earns more with Xeon platforms and also probably has better margins with its x86 mission-critical chips. Moreover, without necessity to develop Itanium, Intel will be able to streamline its research and development operations.
Server makers like Dell, IBM and others as well as software developers like Red Hat are no longer interested in Itanium, but are more interested in getting lucrative customers away from HP. As a result they will do everything possible to compete exactly against HP-UX-based machines, which should not be that hard, given the fact that Xeon processors already offer better performance and RAS features than IA64, according to Intel itself.
Since both Intel and HP promise to support Itanium for a decade, which should ensure smooth transition to any kind of servers for their customers, there should logically be a successor for Kittson. But it is unclear what that successor will be like and whether it will materialize at all.
Perhaps, Intel and HP will introduce "Kittson+" chip in 2016 with certain enhancements to support current deployments (some servers will go out of order, some companies will find ways to squeeze more servers inside their datacenters and will need similar machines with IA64 CPUs). In that case, the new chips will be able to get additional cores and also gain clock-speeds, which will not require major redesigns compared to Kittson.
Still, some platform-related enhancements may be needed in order to balance performance of microprocessors as well as memory/system bandwidth. The next-generation Itanium code-named Poulson will be compatible with current-generation platform that supports Tukwila processors introduced in 2010 and the Kittson is also projected to be compatible with the same socket. However, nothing is clear about hypohetical Kittson successor will also be compatible with the same platform.
Post-Kittson chip due in 2016 or 2017 should allow Intel and HP to just support the new platforms for three or four years without substantially boosting their performance in order to fulfill the promise to do so till the year 2021.