Intel is launching the latest version of its Xeon E7 high-end server chips at a time when data analytics and other compute-intensive workloads are putting attention back onto powerful scale-up systems that run on four or more processors.
Intel officials introduced the Xeon E7 v3 portfolio during an event in San Francisco May 5, saying new features like more cores and technologies that can speed up data processing will enable the new chips to offer up to six times the performance of their predecessors and up to 12TB of memory, supporting both DDR3 and DDR4 memory. The chips can scale from four to 18 cores, and improve security and reliability through enhanced capabilities in the company’s AES-NI, AVX2 and Run Sure Technology.
Executives with the giant chip maker also said they expect the 22-nanometer processors—which sport 5.7 billion transistors—will give organizations an even greater incentive to migrate their high-end workloads from Unix systems to more open and less costly x86 servers, and will accelerate the declining market share of competitors like IBM’s Power8 and Oracle’s SPARC platforms.
Speaking during the event, Diane Bryant, senior vice president and general manager of Intel’s Data Center Group, noted that last year, 2 percent of the servers shipped were RISC-based, and that with the capabilities in the new Xeon E7 v3 processors, that number could shrink. Intel chips own about 94 percent of the market for systems with four or more sockets, while RISC and mainframe systems have about 4.6 percent, Ed Goldman, CTO of the Enterprise IT unit in the Data Center Group, said in March during a workshop in Oregon about the new processors.
The bulk of servers sold are in the two-socket segment, but Goldman said he expects that “technologies like in-memory will get [this part of the market] to grow.” Bryant said the performance and economics of the x86 processors will convince businesses to opt for the Xeon E7 v3 chips. She noted that, compared with the Power8 chips, the new Xeons offer 10 times better performance-per-dollar and 85 percent better total cost of ownership. Intel also claims a 60 percent better performance-per-watt.
What the new processors are addressing is the push for greater data analytics capabilities and real-time processing, Bryant said. Massive amounts of data are being created and organizations want to analyze that data in real time to make better business decisions.
“Data is the new currency,” she said. “It’s the currency of the digital world. … There isn’t an institution or business that can’t benefit from big data and analytics.”
Intel engineers kept that as the backdrop as they worked on the new Xeon E7-8800/4800 v3 chips, which are based on the “Haswell” architecture. They offer 20 percent more cores and threads than the previous version and up to 20 percent more last level cache (which creates up to 39 percent better OLTP database performance), according to officials.
A key new feature is Intel’s Transactional Synchronization Extentions (TSX), a technology that streamlines access to shared memory for multi-threaded applications. Without the technology, many threads are left idling while waiting for a thread that has been locked to access the memory. TSX enables multiple threads to access memory at the same time, which improves the performance of multi-threaded workloads running on the chip, Bryant said. SAP has optimized its latest HANA software to take advantage of the TSX feature.
Intel Aims Latest Xeon E7 Chips at Analytics, Real-Time Processing
The processors are aimed at four- to eight-socket servers, though can support up to 32 sockets, she said.
Intel also has cut in half the number of chips in the family. The Xeon E7 v3 portfolio offers 12 CPUs in four segments: Basic, Standard, Advanced and Segment Optimized for such workloads as databases, low power and high-performance computing (HPC). They vary in the number of cores they have (from four to 18), their speeds (1.9GHz to 3.2GHz) and their power envelopes (115w to 165w), as well as pricing ($1,224 to $7,175). Intel also has simplified the naming structure—those chips with the naming system 88xx are aimed at eight-socket systems, while the 48xx chips target four-socket servers.
Intel’s new chip is a big deal for the company and the industry because of the performance gains, according to Nathan Brookwood, principal analyst with Insight64. Intel was probably startled by what IBM was able to do with Power8 and was able to push what the Xeon E7 v3 could do to respond, Brookwood told eWEEK. It gives Intel an even stronger position in the high-end market that once had been dominated by RISC-based systems, and the TSX technology was able to significantly boost performance.
He noted that SAP optimized its HANA in-memory software for the next Xeon E7 chips, and that Cloudera has done the same with its Hadoop big data solutions. However, Brookwood said he also expects other enterprise software makers—such as Oracle—to follow suit.
“It’s just mind-boggling what Intel has been able to do,” he said. “Everything was improved in a reasonably coherent manner.”
The new chips offer system makers significant advantages, Brookwood said. They have big performance gains, “and because they have the same thermal envelope [as the previous version], their [servers’] performance-per-watt will improve,” he said. “This is as close to a free lunch as you can get from an OEM perspective.”
Those OEMs are responding. Bryant said there are 17 system makers—from Hewlett-Packard and Lenovo to Dell, Oracle, Fujitsu, Cisco System and ZTE—that are developing 45 systems that will be powered by new processors.