Enterprise, Hardware

Intel Announces E5-2600 CPU, New King of Big Iron


On Tuesday, Intel invited a select group of journalists to the announcement of their new Xeon processor E5-2600 product family. They promised significantly higher performance claiming best data center performance per watt, I/O innovation, and trusted hardware security features. They say this will give corporate IT departments? greater ability to scale and deliver data to greater numbers of cloud users.

Introducing the Xeon E5 series
Diane Bryant, Vice President at Intel Corporation lead the whole launch 

The Xeon E5-2600 is one of a family of server chips based on the Sandy Bridge architecture that Intel shipped in late 2011 and says will ship in volume for 2012. Very few Sandy Bridge desktop CPUs are actually in the hands of OEM/ODMs. All Sandy Bridge processors are manufactured with Intel’s 32-nanometer (nm) node process technology. This is not their announced 22nm "Tri-Gate" process that Intel says will ship later this year for the "Ivy Bridge" family of PC processors.

Intel Sandy Bridge-EP core...
The Xeon E5 core

Each core on the Xeon E5-2600 chip has a completely revamped branch predictor for its 32KB instruction cache. The new "front end" of the processor ? the L1 instruction cache, pre-decoder unit, instruction queue, decoder unit, and out-of-order execution unit ? has been redesigned. Thus it can sustain a much higher level of micro-ops bandwidth and use less power by turning off elements of the front end when it can use micro-ops caches added to the chip.

The die size of Intel Sandy Bridge-EP is amazing - bear in mind this is an LGA2011 sized packageWe remember when Intel announced their partnership with VMware in Fall 2007 at their Folsom, CA facility. We were told then it would revolutionize the IT department?s control and work load on fewer servers. Tuesday?s E5-2600 is claiming to deliver up to 80 percent improved performance compared to the prior generation of Intel server CPUs.

Also on Tuesday, IBM announced their new server solutions based on the E5- family. They are designed to expand cloud and analytics capabilities across its entire portfolio, helping make Smarter Computing a reality for x86 clients. IBM claims their new two-socket BladeCenter HS23 server  delivers 62 percent more computing power, and runs 20 percent more virtual machines than its predecessor BladeCenter HS22. The blade server will support up to four times more memory than their previous generation HS22. The HS23 blade is designed for virtualization and cloud deployments, or even database and ERP (enterprise resource planning) installations.

Adalio Sanchez, general manager, IBM System x business said: "IBM is delivering easy-to-deploy cloud and analytics products to help clients align their businesses to manage unprecedented amounts of data, and become much more efficient at turning that information into timely business insights."

Xeon E5 Series servers offer unprecedented amount of computing performance in a standard sized package

Supermicro was at Intel?s San Francisco announcement and will display their E5-2600 platform family at CeBIT 2012 in Hanover, Germany this week. Supermicro is showing Intel?s just announced Xeon E5-2600 platforms with 8 cores per CPU, 16 threads per processor. Each CPU has a free-air cooling system designed to withstand 47 degrees Celsius operating temperature environments. Supermicro has created a new platform architecture called "Fat Twin" which comes with enhanced cooling, power savings, and lower TCO (total cost of operations), along with their own high-efficiency power supplies with integrated UPS (uninterrupted power supply) battery backup.

Industry analyst, Nathan Brookwood of Insight 64, said the adoption of the new PCI-Express 3.0 bus and 10-gigabit Ethernet at the motherboard level will improve throughput for servers. Intel claims the integrated I/O with PCI Express 3.0 can have up to triple the movement of data into and out of the processor. They claim information can be made available faster than ever to support data-hungry applications. This controller will be offered as both LAN on Motherboard (LOM) as well as converged network adapter (CNA) card.

Diane Bryant, Intel vice president and general manager of the Datacenter and Connected Systems Group, was one of the presenters on Tuesday. She gave an excellent overview of the advantages of the E5-2600 family, pointing out that corporate IT server purchases have a three to five year upgrade window. Many of Intel?s corporate customers are still using Intel?s older systems. With the significant performance improvements of the E5-2600, many of them should be taking advantage of the improved features.

Bryant listed some of the key requirements to enable IT to scale, such as higher performance, energy efficiency, I/O bandwidth and security. Intel claims they have the best combination of performance, built-in capabilities and cost-effectiveness. Intel?s new Xeon processor E5-2600 product families are designed to address these requirements, and take over the workload for the next-generation data center powering servers, storage and communication systems.

Fujitsu showed off their 2U E5-2600 based server

On Tuesday, Fujitsu showed their Xeon E5-2600 dual socket 2U rack version. It is expandable to 16 hard drives, up to 7 PCIe Gen 3 cards, and can have as much as 768GB of RAM – all in a 2U rack housing.

Intel 10GbE CardIntel?s new integrated X540 Ethernet Controller chipset and LAN card can be a LAN on Motherboard (LOM) as well as converged network adapter (CNA) card. Insight 64’s Nathan Brookwood said bringing 10GbE onto motherboard
s represents a big step in the right direction for increasing performance levels of lower-cost servers. "It’s been talked about for years, but it’ll be finally coming to fruition," Brookwood said.

Bryant summarized the presentation acknowledging the manufacturers announcing E5-2600 products this week, including: Acer, Appro, Asus, Bull, Cisco, Dell, Fujitsu, HP, Hitachi, Huawei, IBM, Inspur, Lenovo, NEC, Oracle, Quanta, SGI, Sugon, Supermicro and Unisys.

The floor was then opened for questions. Right off the bat, the first question was about AMD?s purchase of SeaMicro. Industry analysts saw the move as a blow to Intel, which had appeared to have formed a tight relationship with SeaMicro as the CPU of choice for their high-density microservers. Bryant said Intel "did look at SeaMicro’s fabric technology. There are probably very few people they didn’t come to and shop their solution to. We were not impressed. We declined and very soon after our competitor acquired them."

The presenter’s dismissive words seemed to say one thing, yet body language seemed to tell another story. Her body language seemed to indicate that some folks at Intel were NOT very happy about AMD jumping out and purchasing SeaMicro and making them their own captive "partner"/division. The proof of the pudding will be whether AMD captures any significant design wins and makes sales strides in the corporate microserver marketplace.

With Intel?s year in, year out proven track record and market share in the x86 server arena, plus the fact that Intel?s last major server CPU design change was over three years ago, they should sustain their dominate position of owning nearly 80 percent of the corporate IT x86 server marketplace. Independent benchmarking has yet to verify all of Intel?s above performance claims.  Today?s x86 ?Big Iron? server market is shifting as microservers come on-line, transaction per Watt of consumed power becomes more important with less focus on Big Iron?s brute force performance solutions, along with the ARM 64-bit IP and MIPS 64-bit IP. These facts could all erode Intel?s secure position in the server data center world.