HPC

IBM's BlueGene/Q super chip grows 18th core

It's nice to have a spare


Hot Chips The mystery surrounding the number of cores in the 64-bit Power processor that will be at the heart of the 20 petaflops "Sequoia" BlueGene/Q supercomputer has been finally cleared up.

Back at the SC10 supercomputing conference in November 2010, a software engineer working on the BlueGene/Q system told El Reg that the processor module at the heart of the system would have 17 cores: one to run the Linux kernel and the 16 others to perform mathematical calculations. IBM also said at the time that this chip would be a variant of the PowerA2 "wirespeed" processor, but geared down to 1.6GHz from its 2.3GHz design speed.

In February 2011, when Argonne National Laboratory said that it was going to take a 10 petaflop super based on the BlueGene/Q design (basically half of the Sequoia machine that is going into Lawrence Livermore National Laboratory), IBM told El Reg that it was just a 16-core chip, nothing funky.

For whatever reason, neither turns out to be true. The BlueGene/Q processor, the company revealed at the Hot Chips conference at Stanford University late last week, actually has 18 cores: 16 cores for doing work, one core for running Linux services, and a spare that is intended to merely increase the yield that IBM Microelectronics can get out of its chip fabs but which can, according to George Chiu, senior manager of advanced high performance systems at IBM, be activated and used in the system, in theory.

Chiu was very clear, however, that he was not making any promises that this 18th core would be used as a hot spare in any BlueGene/Q supers, but merely that the capability is there.

Big Blue detail

IBM gave out a lot more detail on the BlueGene/Q processor at Hot Chips, and Chiu walked El Reg through the details. The chip looks like this:

IBM BlueGene/Q chip

The BlueGene/Q custom Power processor

Like other processor designs these days, the BlueGene/Q processor is an example of a system-on-a-chip design, which tries to cram as many components of the system board onto the chip. The BlueGeneQ processor is based on the Power A2 core that IBM created for networking devices and experimentation, and this is the block diagram of the core:

IBM BlueGene/Q A2 core

The BlueGene/Q processor's Power A2 core block diagram

This Power A2 core has a 64-bit instruction set, like other commercial Power-based processors sold by IBM since 1995 but unlike the prior 32-bit PowerPC chips used in prior BlueGene/L and BlueGene/P supercomputers. The A2 core have four threads and has in-order dispatch, execution, and completion instead of out-of-order execution common in many RISC processor designs. The A2 core has 16KB of L1 data cache and another 16KB of L1 instruction cache. Each core also includes a quad-pumped double-precision floating point unit, which is blocked out thus:

IBM BlueGene/Q quad FPU

The quad-pumped FPU in each BlueGene/Q core

Each FPU on each core has four pipelines, which can be used to execute scalar floating point instructions, four-wide SIMD instructions, or two-wide complex arithmetic SIMD instructions. These instructions are extensions of the of the Power instruction set. The FPU has a six-stage pipeline and has permute instructions to reorganize vector data on the fly; it can do a maximum of eight concurrent floating point operations per clock plus a load and a store.

Crossbar connection

At the heart of the BlueGene/Q chip is a crossbar switch, which links the cores and L2 cache memory together. This crossbar runs at half the clock frequency, at 800MHz, and it has a peak bisection bandwidth of 563GB/sec. It connects the processors, the L2 cache segments, the networking logic, and other parts of the chip together.

Like IBM's commercial Power7 chip, the BlueGene/Q processor has 32MB of embedded DRAM as cache memory, but this is implemented as an L3 cache on the Power7 and it's an L2 cache on the BlueGene/Q processors. Each BlueGene/Q chip has two DDR3 memory controllers, which have ECC scrubbing and which support 16GB of total memory running at 1.33GHz.

These memory controllers interface with eight slices of L2 cache, handling their cache misses (one controllers for each half of the 16 cores on the chip). The DDR3 memory and the BlueGene/Q chips are soldered onto the same processor card, and the two memory controllers provide 42.7GB/sec of bandwidth from the processor out to main memory on the card.

The logic for IBM's 5D torus interconnect is also embedded on the chips, with 11 links running at 2GB/sec. Two of these can be used for PCI-Express 2.0 x8 peripheral slots. The 14-port crossbar switch/router at the center of the chip supports point-to-point, collective, and barrier messages and also implements direct memory access between nodes.

The spare

The 17th core runs Red Hat Enterprise Linux, and it's designed to offload operating system services from the other 16 working cores on the BlueGene/Q processor to eliminate distracting OS noise and jitter. This 17th core will take care of interrupt handling, asynchronous I/O, MPI pacing, and RAS event handling, which you need in a system with 1.57 million working cores. In the event of a core failure, the system can remap the cores, bringing the 18th core online if it hasn't already been blocked off because it has a booger on it.

But the 18th core was really added to increase chip yields.

The BlueGene/Q processor is 359.5 square millimeters in area (18.96 millimeters per side in an actual square), and it has around 1.47 billion transistors. It is implemented in the 45 nanometer copper/SOI process that IBM used to make the Power7 chips. The cores used in the BlueGene/Q chip will all spin at 1.6GHz, with IBM varying the voltage as necessary around its 0.8 volt target to keep that clock speed rock solid and therefore be able to synchronize events across those 1.57 million cores.

By lowering both the clock speed and voltage from the Power A2 processor, IBM is giving up some performance per chip, but it is able to dramatically lower the active and leakage power of the processor. The 16-cores are designed to deliver 204.8 gigaflops at a power draw of 55 watts.

Here's how the full BlueGene/Q system gets integrated:

IBM BlueGene/Q hierarchy

The BlueGene/Q system (click to enlarge)

The compute cards and the optical interconnect modules have water blocks on them to keep them cool. The system is designed to use water at between 60 and 65 degrees Fahrenheit. The Sequoia system that Lawrence Livermore is getting will have 96 racks of BlueGene/Q nodes to reach its 20 pteaflops performance level, and it is expected to consume around 6.6 megawatts.

BlueGene/Q is expected to start shipping to Lawrence Livermore and Argonne national labs, which are financed by the US Department of Energy, next year. ®

Narrower topics


Other stories you might like

  • Lonestar plans to put datacenters in the Moon's lava tubes
    How? Founder tells The Register 'Robots… lots of robots'

    Imagine a future where racks of computer servers hum quietly in darkness below the surface of the Moon.

    Here is where some of the most important data is stored, to be left untouched for as long as can be. The idea sounds like something from science-fiction, but one startup that recently emerged from stealth is trying to turn it into a reality. Lonestar Data Holdings has a unique mission unlike any other cloud provider: to build datacenters on the Moon backing up the world's data.

    "It's inconceivable to me that we are keeping our most precious assets, our knowledge and our data, on Earth, where we're setting off bombs and burning things," Christopher Stott, founder and CEO of Lonestar, told The Register. "We need to put our assets in place off our planet, where we can keep it safe."

    Continue reading
  • Conti: Russian-backed rulers of Costa Rican hacktocracy?
    Also, Chinese IT admin jailed for deleting database, and the NSA promises no more backdoors

    In brief The notorious Russian-aligned Conti ransomware gang has upped the ante in its attack against Costa Rica, threatening to overthrow the government if it doesn't pay a $20 million ransom. 

    Costa Rican president Rodrigo Chaves said that the country is effectively at war with the gang, who in April infiltrated the government's computer systems, gaining a foothold in 27 agencies at various government levels. The US State Department has offered a $15 million reward leading to the capture of Conti's leaders, who it said have made more than $150 million from 1,000+ victims.

    Conti claimed this week that it has insiders in the Costa Rican government, the AP reported, warning that "We are determined to overthrow the government by means of a cyber attack, we have already shown you all the strength and power, you have introduced an emergency." 

    Continue reading
  • China-linked Twisted Panda caught spying on Russian defense R&D
    Because Beijing isn't above covert ops to accomplish its five-year goals

    Chinese cyberspies targeted two Russian defense institutes and possibly another research facility in Belarus, according to Check Point Research.

    The new campaign, dubbed Twisted Panda, is part of a larger, state-sponsored espionage operation that has been ongoing for several months, if not nearly a year, according to the security shop.

    In a technical analysis, the researchers detail the various malicious stages and payloads of the campaign that used sanctions-related phishing emails to attack Russian entities, which are part of the state-owned defense conglomerate Rostec Corporation.

    Continue reading
  • FTC signals crackdown on ed-tech harvesting kid's data
    Trade watchdog, and President, reminds that COPPA can ban ya

    The US Federal Trade Commission on Thursday said it intends to take action against educational technology companies that unlawfully collect data from children using online educational services.

    In a policy statement, the agency said, "Children should not have to needlessly hand over their data and forfeit their privacy in order to do their schoolwork or participate in remote learning, especially given the wide and increasing adoption of ed tech tools."

    The agency says it will scrutinize educational service providers to ensure that they are meeting their legal obligations under COPPA, the Children's Online Privacy Protection Act.

    Continue reading
  • Mysterious firm seeks to buy majority stake in Arm China
    Chinese joint venture's ousted CEO tries to hang on - who will get control?

    The saga surrounding Arm's joint venture in China just took another intriguing turn: a mysterious firm named Lotcap Group claims it has signed a letter of intent to buy a 51 percent stake in Arm China from existing investors in the country.

    In a Chinese-language press release posted Wednesday, Lotcap said it has formed a subsidiary, Lotcap Fund, to buy a majority stake in the joint venture. However, reporting by one newspaper suggested that the investment firm still needs the approval of one significant investor to gain 51 percent control of Arm China.

    The development comes a couple of weeks after Arm China said that its former CEO, Allen Wu, was refusing once again to step down from his position, despite the company's board voting in late April to replace Wu with two co-chief executives. SoftBank Group, which owns 49 percent of the Chinese venture, has been trying to unentangle Arm China from Wu as the Japanese tech investment giant plans for an initial public offering of the British parent company.

    Continue reading
  • SmartNICs power the cloud, are enterprise datacenters next?
    High pricing, lack of software make smartNICs a tough sell, despite offload potential

    SmartNICs have the potential to accelerate enterprise workloads, but don't expect to see them bring hyperscale-class efficiency to most datacenters anytime soon, ZK Research's Zeus Kerravala told The Register.

    SmartNICs are widely deployed in cloud and hyperscale datacenters as a means to offload input/output (I/O) intensive network, security, and storage operations from the CPU, freeing it up to run revenue generating tenant workloads. Some more advanced chips even offload the hypervisor to further separate the infrastructure management layer from the rest of the server.

    Despite relative success in the cloud and a flurry of innovation from the still-limited vendor SmartNIC ecosystem, including Mellanox (Nvidia), Intel, Marvell, and Xilinx (AMD), Kerravala argues that the use cases for enterprise datacenters are unlikely to resemble those of the major hyperscalers, at least in the near term.

    Continue reading

Biting the hand that feeds IT © 1998–2022