Intel + HP 's data center push - saving power and saving labor
By Tom Foremski - June 21, 2006
I met with Hewlett-Packard and Intel (Intel is a sponsor of SVW) recently to talk about the new blade servers coming from HP including Itanium based servers. Scott Stallard, senior vp and GM for the enterprise group says the company has a lot of pent up demand for its Itanium systems which is good news for the Intel/HP designed 64-bit microprocessor.
Itanium has had a long and rocky road from development to production systems and now the more advanced generations of Itanium based systems. That's to be expected in creating the large ecosystem that's needed for a new microprocessor architecture. You need the compilers, the development tools, and the system design is more challenging than for PC clients. And the road has been longer than some industry analysts expected, and a lot harder to travel than Intel and HP expected.
The new Itanium systems will be up against the best POWER and SPARC based systems so it will be interesting to see how they perform in the market for heavy-duty scientific number crunching. Itanium is designed to process big computational problems such as predicting global warming effects, protein folding, drug discovery and many other grand challenges. We are going to need all the help we can get to deal with some very challenging environmental and healthcare problems.
Lisa Graff, general manager of server platforms at Intel says that Itanium's prior problems with speed and power usage have been solved and the latest systems provide benchmarks that exceed IBM's POWER and Sun's SPARC based systems.
Benchmarking such large systems is notoriously difficult because of the different types of applications that this type of "big iron" runs, but Intel has stayed the course with Itanium to make it into a competitive product against long established architectures and it now this seems to be paying off. However, it will be a considerable while longer before Intel will have its ROI.
Mr Stallard gave the familiar data center pitch that Sun, IBM and others give: data center constraints are electric power, therefore to pack in more computing power you have to have low power consuming systems that are built as blades--easily slipped into data center racks. Then you need good management technologies to automate as much of the admin as possible; also you need virtualization technologies so that you can improve server efficiency from abut 30 per cent to as much as 70 per cent.
It's a familar pitch but HP does have some interesting data center technologies that could set it apart in the market. One of these is its Virtual Connect Architecture which automates the many different connections between blade servers and all the other equipment that they communicate with. This means equipment can be deployed more quickly and redeployed according to different business requirements.
All this greater efficiency in server utilization could lead to customers buying/needing fewer servers. But Ms Graff points out that there is no such thing as needing less computing, organizations will find plenty of uses for the freed up computing cycles, and that will drive more sales. She also said that the new Intel server microprocessors coming out this summer will exceed the performance, and have lower power consumption than Advanced Micro Devices' Opteron server chips.
Intel has been late coming into this market and challenging Opteron. This opened up an opportunity for AMD to get into the lucrative data center server business, where it previously had no presence. But Intel is like a very large supertanker and course corrections take time , and once they happen Intel can bring its massive manufacturing prowess to bear and win back market share. Its ability to be among the first to integrate leading edge chip technologies and quickly ramp up chip production in its fabs is formidable. AMD knows this and is rapidly expanding its chip fabs in Dresden, Germany - it will be an interesting race to follow.
« UPDATE: Gmail outage... GOOG's sagging infrastructure breaks down | Main | Horn Group's salon on top of the Clift »
June 21, 2006 | Permalink | Comment | Category: | Subscribe to SVW
- Top Stories:
- Tech Awards For Benefiting Humanity
- The Death Of The Search Algorithm? Techmeme Has Six Editors
- TEDxSF - Little TED Just Like The Big TED
- SNCR Research: Social Media IS Influencing Business Decisions
- What's Next? Beyond Real-Time...
- PearlTrees: A Novel Approach To Human Mapping Of The Internet
- MediaWatch Analysis Part II: Google Has More To Lose Than Murdoch
- MediaWatch Analysis: Murdoch Will Negotiate Payment For Access To Basket Of Content With GOOG et al
- WeekendWatcher: The Sheer Number Of Things Will Devalue Them
- ChipWatch - Where Will The Next Generation Of Engineers Come From?
- Public Healthcare Could Cut Startup Costs And Help Spur Innovation
- Is GOOG's $750m AdMob Buy Strategic Or Dumb? An alternate view...
Comments (2)
Tom, these guys are dreaming. Itanium is so dead. Itanium is the Alamo and high-volume consumer technologies are Santa Anna. Itanium is a tragic case of rear-view mirror thinking. AMD's market share gains prove it. Barrett was conned and Otellini should kill it.
It isn't about the hardware - its the software. And there is very little software for Itanium. When cheap 64bit chips have enough mojo to support dozens of virtual servers, what is the point of the big iron strategy? I've been writing about Intel lately over at http://storagemojo.com/.
Posted: June 22, 2006 6:45 PM
I think Itanium has gotten interesting again like enterprise software. I had written off both of them, a couple or three years ago, but now I'm not sure, I'm interested in seeing if Intel's determined persistence with Itanium might be paying off. And that would be great comeback story if they can pull it off.
Posted: June 23, 2006 2:50 AM