Ivy Bridge comes to UCS!
As Paul Perez reminds us in his evolutionary metaphor for our industry, time marches on, and today brings another tick of the clock. Many of you may be familiar with the development cadence that Intel maintains to carry forward the exponential burden of Moore’s law. By way of explanation for the uninitiated: the “tock” of the clock brings a new microarchitecture, the “tick” brings a new process technology, often referred to as a “die shrink,” which wrings out more efficiency and computing density from the platform. When we launched UCS M1 series servers back in 2009 we picked up with the “tock” of Nehalem processors. Over the past four years we “ticked” into Westmere, then a “tock” into Sandy Bridge, and now today, a “tick” into Ivy Bridge on our M3 systems. Ivy Bridge is the program name for the Intel® Xeon® Processor E5-2600 v2 product family, which ushers in an unfathomable 22 nanometer process (three dimensional transistors!) and as many as 12 cores per processor for workstations and mainstream servers. This week Cisco is introducing support for these new processors as options on our existing B200, C240 and C220 M3 servers and as upgrade kits for systems already in the field.
Cisco’s implementation of this technology is superior. There is ample evidence to support this assertion: this weeks news includes 7 record-breaking application benchmark wins for UCS, and we’ve seen 81 of those since the introduction of UCS just four years ago. Taking a look at the results posted today for the v2 family, Cisco is dominant, with more #1 results than any other server vendor. Why is the system so fast? It’s partly an outcome of the mechanical design advantages in compute density and airflow that come with a Unified server architecture. Further advantage comes from high performance physical and virtual I/O found in the ASIC-level innovation of Cisco SingleConnect Technology. Don’t settle for imitations.
We like to say that Cisco and Intel are “Joined at the Chip,” because the innovation each company brings is incredibly complimentary. Cisco’s innovation in the data center is an extension of the company’s historic focus: connecting things. Cisco Unified Data Center and products like UCS are the outcomes of our drive to connect the pieces. When we join forces with Intel’s leadership at the computing core, customers see an unbeatable combination.
There are immediate gains with this new processor family and the way it is implemented in UCS, both in terms of performance (as much as 48% faster) and efficiency (35% improvement.) These are crucial elements in reducing operating costs and supporting the new computing models that rely on powerful virtualization, encryption and security technology. But this foundation is just the beginning of the story.
The operational elements of IT are where customers face their biggest challenges. This is where the majority of innovation in UCS is focused. If record-breaking application performance is the icing, operational innovation for IT is the cake: with UCS customers reporting an 84% reduction in provisioning times and 61% reduction of ongoing administrative/management costs.
Introduction of this latest Intel technology on UCS also supports Cisco’s advancement of integrated infrastructure solutions with ecosystem partners: most recently, two new Cisco Validated Designs for high performance server virtualization for EMC VSPEX with Microsoft Hyper-V and VMware vSphere, which incorporate next generation EMC VNX storage
arrays.
Taken altogether, customer demand for UCS is changing the shape of the industry: Cisco is ranked #2 world-wide in x86 blade server revenue market share, with 33.9% share in the US, and a top 5 ranking among server vendors overall.
And as we’ve seen in other news today, UCS technology is very much on fast forward, along with Intel’s, so don’t let the tick tock from traditional server OEM’s lull you to sleep.
This is great!!! B200, C240 and C220 M3 going to pack a great consolidation ratio. Great news for enterprise looking for consolidating CPU intensive workload on the UCS.
Isn’t Ivy bridge the last generation? Haswell has been out for a while now.
Good question. Haswell was announced by Intel back in June. Like Ivy Bridge, it will come to desktops/notebooks before it arrives in multi-socket systems like servers and workstations.
Most of the PC vendors have already released Haswell variations of their lineups for sometime now. Not sure about the Server side. Sounds odd when a UCS refresh with a older gen chip is being released in lieu of a newer gen version being out already.
Take a look at the high-end E5 family (Q3), I’m guessing the v2 are Haswell based.
http://ark.intel.com/products/family/59138
The mid-powered E3 family Haswell versions have been available from Q2.
Correct, that’s the normal course of things for all the server vendors (not just UCS.) Intel will qualify a new architecture for notebooks/desktops several quarters earlier than for servers. the glue for SMP designs takes longer to develop; some server vendors will build ~64 proc systems around it.
Check out the server reference in this article on Haswell:
http://en.wikipedia.org/wiki/Haswell_(microarchitecture)
The release of a more powerful chip is always a good thing. With the growing popularity of virtualization, how will this upgrade in appliance compete against an ESX server with the latest and greatest built from the ground up?
Xeon E5v2 (ivy bridge) is the latest and greatest for servers. If I understand your question right: building up a system on latest gen desktop reference architecture to run ESX and how that might compare to a sever platform on previous generation processor… not certain. I’d have to dig to see how the memory controllers are configured on each because that’s the major lever for virtualization. The server platform is likely going to outperform even between generations because more memory controllers are typically available vs. the desktop/notebook implementations.
Great article! Being more powerful CPU means you can do same work with less energy spent, heat dissipation, which might justify bulk upgrades for a DC. Same time, means your business can grow with more resources standing by quickly.
Of course, simple assumption here, each set up is different but that’s the basis for the upgrade.
Not only new process for CPU is cool for processing by itself, but means memory chips will follow similar improvements so you can add up cache, RAM and storage flash, which helps whole setup when refreshing capacity.
Haswell for servers is not urgently needed, considering better power saving features which help on laptop lines refreshed by mid-2013, but once it’s ready for servers, they can benefit on a little more power efficiency which is always welcomed!