Today, Intel made two fairly large announcements simultaneously with the announcement of their Omni Scale Fabric and the integration of it into their next generation of Xeon Phi chips. Additionally, Intel has worked with Micron to enhance Knight’s Landing with high-performance on-package memory. This on-package memory is also known as Hybrid Memory Cube or HMC.

The Knight’s Landing next generation Xeon Phi product announced today will use Intel’s Silvermont CPU architecture which is modified (or as Intel says, enhanced) for HPC. The expectation that Intel is setting is that these cores will deliver three times the single threaded performance of the previous generation and still be Intel Xeon Binary compatible.

Knight’s Landing – MCDRAM

As for the on-package memory itself, the HMC, it will support up to 16GB at launch while only taking up 1/3 the space (compared to Knight’s Landing vs Knight’s Corner) compared to GDDR5. They are also claiming five times the bandwidth compared to DDR4 using the same amount of memory. However, to be fair, this 5x bandwidth comparison versus DDR4 isn’t necessarily a fair one since DDR4 is still in its infancy and a more appropriate comparison would be DDR5 since DDR3 in servers is fairly slow. Intel and Micron are also claiming 5x the power efficiency when compared to GDDR5, based upon comparisons between Knight’s Landing and Knight’s Corner (the previous generation). But do keep in mind that Micron says they will only be shipping 2GB and 4GB parts this year, making the 16GB parts for Knight’s Landing 2015 parts.

Knight’s Landing – MCDRAM

The Omni Scale Fabric from Intel is designed to deliver maximum bandwidth and scalability between Intel’s future Xeon and Xeon Phi products. With interoperability between Knight’s Landing and Xeon processors coming with the 14nm generation of Xeon server processors. This Omni Scale Fabric will have PCIe adapaters, edge switches, director systems, Intel’s own silicon photonics and open software tools. All of these are designed to make the upgrade to Intel’s Omni Scale Fabric less painful.

Omni Scale Fabric

In terms of performance, Knight’s Landing will deliver “3 TFLOPS of peak theoretical double-precision performance” which is based on preliminary and “expectaionts of cores, clock frequency and floating point operations per cycle.” Which really means that this hasn’t really been benchmarked exactly quite yet and we will have to see what the end product delivers when Intel actually starts shipping commercial systems in the second half of 2015.

There is also already an operational supercomputer using Knight’s Landing in their system and that’s NERSC’s Cori supercomputer which already employs Knight’s Landing and has already benchmarked the performance at over 3 TFLOPs per Knight’s Landing node.

NERSC

The post Intel's New Knight's Landing Xeon Phi Combines Omni Scale Fabric with HMC appeared first on VR World.

Last week ARM invited a group of journalists and analysts to Austin Texas to hear about their server, mobile, and wearable developments. ARM and their partners presented in-depth explanations of their version of the ARM architecture.

On the first day of the conference, HP’s Dwight Barron gave an overview of their Moonshot system.  They have been refining the specifications since its late 2009 inauguration.

Moonshot’s design differs from the traditional servers which have been the general-purpose workhorses of the data center. These boxes have proved to be jacks-of-all-trades, able to run operations for organizations of every shape and size. They started with proprietary operating systems and a warehouse sized room with less computing power than today’s smartphones. In the past, one could choose from several operating systems and server architectures. Today, operating system options are limited and most run on the Intel x86 architecture.

Barron explained that the cloud and mobile applications have changed the assumptions of traditional IT departments. Now IT has to balance rack space density, power consumption, thermal efficiency, and costs. Power is one of the major controlling factors in the data center.

The cloud uses more power than Japan.

The microserver SoC (System on a Chip) typically has a TDP (Thermal Design Power) of 15 to 20 Watts or below, compared to 90 plus Watts for a high-end server. The microserver chassis has the circuitry related to networking, storage, and cluster communications along with integrated cooling and power supply. Thus, the shared resources reduce the complexity of the overall design.

Barron showed a typical dual processor, HPC server motherboard versus the SoC based Server motherboard. The dual processor general purpose motherboard requires system RAM for the processors, dual GPU’s with their RAM as well as the overhead of a PCIe switch, storage controller, and NICs. Each of those separate chipsets consumes power and creates heat. The SoC server motherboard starts with an integrated die having all those features in silicon, which greatly reduces required power and significantly decreases heat. The SoC cartridge and the Moonshot integration will reduce the latency of the wires between the traditional HPC server’s sub-systems.

New era app focused silicon

To prove that Moonshot is a viable approach, HP put in a production configuration. At the HP.COM website there are approximately 100 applications for browsing and downloads. HP.COM’s website gets approximately 300 million hits per day. It was running on 46 HP legacy servers consuming just over 115,000 Watts per day. Replacing that legacy configuration with six Moonshot systems lowered the power consumption to 6,000 Watts and reduced the rack space usage by 89 percent.

HP 300m hits 94% less power

Barron gave a lengthy explanation of why mobile users and cloud services require application focused silicon for the servers. HP will offer server cartridges from multiple vendors – AMD, Applied Micro, Intel, and Texas Instruments. Applied Micro and Texas Instruments are based on the ARM architecture. This provides customers with the option of picking a server cartridge tailored for specific services, like High Performance Computing (HPC), gaming, telecommunications, finance, seismic imaging, Big Data analysis, web serving, and video analysis, to name a few.

TI  has developed hybrid ARM processors that mix anywhere from one to four Cortex-A15 (32-bit) cores with one to eight TMS320C66x digital signal processors into a single SoC. These combinations in the Moonshot can be used in application specific work such as pure cloud infrastructure workloads – servers, switches, routers, network control planes, and telecommunication switches with applications such as VoIP and LTE.

Applied Micro’s X-Gene 1 eight core SoC with ECC (error-correcting code) memory. The platform is capable of running a full LAMP (Linux, Apache, MySQL and PHP) software stack. The X-Gene implements the ARMv8 ISA which is a full 64-bit architecture that is backwards compatible with 32-bit ARMv7. The CPU features hardware virtualization acceleration, MMU virtualization, and advanced SIMD instructions.

Barron said that ARM architecture and HP’s Moonshot will bring new performance levels and reduced power consumption to the data centers.

There was an interesting announcement in last Friday’s Digitimes. “Foxconn Electronics (Hon Hai Precision Industry) and Hewlett-Packard (HP) will establish a joint venture specifically for producing servers for cloud computing and offering related supporting services, according to Foxconn.”
“The partnership reflects innovation in HP’s server business model through combining Foxconn’s R&D capability and manufacturing expertise, with HP’s market leadership in cloud computing products and related services to enable both companies to offer cloud computing solutions which will change existing market game-playing rules, HP CEO Meg Whitman said.”

Paul Teich, CTO and Senior Analyst, Moor Insights & Strategy, said, “HP and Foxconn’s partnership should help both of them address substantial challenges in continuing cloud computing R&D investment in spite of purchasing pressures that might lead to a ‘race to the bottom’ for prices and margins.”

BSN is planning a hands-on evaluation of the performance of an HP Moonshot with the Applied Micro X-gene cartridge. We will let you know the results as we did with our May 2011 comparison of ARM to x86.

The post HP Moonshot Using ARM 64-bit SoC appeared first on VR World.