IBM opens up Power chips, ARM-style, to take on Chipzilla

Google joins consortium along with Nvidia, Mellanox, Tyan


With its embedded Power chip business under assault from makers of ARM and x86 processors – and to a lesser extent MIPS chips – and having lost the game console business to AMD, IBM had to do something dramatic to expand the addressable market for its Power processors. And that something, which Big Blue has just rolled out, is called the OpenPower Consortium. which takes a few pages from the ARM Holdings playbook to breathe some new life into the Power architecture.

Search and infrastructure giant Google has joined the OpenPower consortium, as has GPU chip maker Nvidia, networking chip maker Mellanox Technologies, and motherboard maker Tyan.

The OpenPower consortium has been in the works for a while, Brad McCredie, chief technology officer for IBM's Systems and Technology Group, tells El Reg, and is part of its efforts to bolster the Power platform.

Precisely how long, McCredie was not at liberty to say, but it stands to reason that when Tom Rosamilia, currently general manager of Systems and Technology Group, was given the job of chief strategist by then-new CEO Ginni Rometty in early 2012, doing something to shake up the Power chip business was high on the to-do list.

The plan, according to McCredie, is to open up the intellectual property for the Power architecture and to allow customizations by licensees, just like ARM Holdings has done brilliantly with its ARM processors, and – as Intel has been hinting – the world's biggest chip maker is thinking about doing as well.

"The way the industry is innovating is shifting," says McCredie. "Clearly open and multi-party collaboration is becoming much, much more the way the industry is innovating with technology. And so we are taking our Power IP, opening it up as well as decomposing it – the processor, the firmware, all of the key pieces – to enable people to innovate around Power platforms. Web 2.0 companies definitely consume hardware this way."

There is a confluence of different factors that have compelled IBM to open up the Power architecture. McCredie said that in the past, system makers of all kinds – be they PCs or servers – were content to get the performance or total cost of ownership goals they have for machines by innovating at the motherboard or system level. But now, this is shifting and companies want to innovate at the chip level, particularly as processors, networking, and storage functions are being brought down onto system-on-chip, or SoC, packages.

There is also plain old physics playing a role, with chip etching technologies now sufficiently small that these different components can be put into a single die or within a single package.

The third factor is that the IT industry wants choice – meaning something other than an x86 option. "I am hearing very strongly that we need to have choice out there from our partners and others in the industry," says McCredie.

The governance model for the OpenPower Consortium is just being established, but the basic idea is that IBM will control the Power instruction set, much as ARM Holdings does with the ARM instruction set. In days gone by, people might have wanted to innovate at the instruction set level, but McCredie says this is less important today, and being able to add functions to a Power core (or a collection of them with cache memories and such) is what people want to be able to do.

The other important thing is to have open source firmware, which IBM will contribute, to go along with an open source operating system, in this case Linux, and an open source hypervisor, with KVM being the choice, that will appeal to hyperscale data center operators. IBM is not opening up its AIX variant of Unix or its PowerVM hypervisor, so don't get too excited.

Finally, while IBM would love for potential future Power chip designers to use its wafer baker in East Fishkill, New York, that will not be a requirement. Without naming names, McCredie says that IBM will work with consortium members to find alternate fabs for Power chips.

"You can't call it OpenPower if there is only one place to manufacture the chips," he quipped, another veiled dig at Intel and its aspirations to allow customizations of its x86 processors so long as it does the fabbing.

GlobalFoundries and TSMC are the two obvious non-IBM foundry choices.

Google, as usual, is not saying much about its intentions regarding the OpenPower Consortium and its possible use of Power-based systems in is monstrous infrastructure.

"We believe in openness and we are looking forward to the innovation that the OpenPower Consortium will bring to the datacenter hardware and software industry," a Google spokesperson tells El Reg. "The consortium has the potential to establish Power architecture as a viable option for applications running within Google's datacenters."

Google cautions that it is "early days" and that it is "looking forward to seeing what can come out of this consortium."

Nvidia, which is an ARM licensee, is not about to become a Power licensee, says Sumit Gupta, general manager of the Tesla Accelerated Computing business unit. But Nvidia is very excited about the prospects of marrying Power processors and Nvidia GPUs for both HPC and general purpose systems.

"It provides an alternative to x86 in both the HPC and in the enterprise markets," says Gupta. "And I personally see this about IBM backing CUDA, just like IBM backed Java in the enterprise nearly twenty years ago. They bring a high performance CPU to the party, and we have a high-performance GPU. And because we are cooperating, we can build a much better solution than other situations that we are in. Having choice in the ecosystem is good for the entire ecosystem."

Mellanox is going to be working on integrating its server adapters tightly with Power chips, and Tyan will presumably be working on alternative motherboards to the ones that IBM has manufactured for its own use. (Tyan executives were not available for comment at press time.)

The OpenPower Consortium is not restricting the licensing of any particular Power chip technology, but McCredie says that it is really focused on next year's Power8 chip and onwards. The Power8 chip, he explained, will have on-chip PCI-Express 3.0 controllers – something IBM has been sorely lacking this year with the Power7+ chips – which will put it on par with Intel's Xeon chips.

Interestingly, the Power8 chip will get rid of the GX++ InfiniBand links used to plug in external peripheral drawers for Power Systems machines and replace it them something called the Coherently Allocated Processor Interface, or CAPI for short. This is an overlay that will ride atop the PCI-Express 3.0 mechanicals to provide coherent memory addressing for CPUs and external coprocessors like Nvidia GPUs. ®


Other stories you might like

  • Running Windows 10? Microsoft is preparing to fire up the update engines

    Winter Windows Is Coming

    It's coming. Microsoft is preparing to start shoveling the latest version of Windows 10 down the throats of refuseniks still clinging to older incarnations.

    The Windows Update team gave the heads-up through its Twitter orifice last week. Windows 10 2004 was already on its last gasp, have had support terminated in December. 20H2, on the other hand, should be good to go until May this year.

    Continue reading
  • Throw away your Ethernet cables* because MediaTek says Wi-Fi 7 will replace them

    *Don't do this

    MediaTek claims to have given the world's first live demo of Wi-Fi 7, and said that the upcoming wireless technology will be able to challenge wired Ethernet for high-bandwidth applications, once available.

    The fabless Taiwanese chip firm said it is currently showcasing two Wi-Fi 7 demos to key customers and industry collaborators, in order to demonstrate the technology's super-fast speeds and low latency transmission.

    Based on the IEEE 802.11be standard, the draft version of which was published last year, Wi-Fi 7 is expected to provide speeds several times faster than Wi-Fi 6 kit, offering connections of at least 30Gbps and possibly up to 40Gbps.

    Continue reading
  • Windows box won't boot? SystemRescue 9 may help

    An ISO image you can burn or drop onto a USB key

    The latest version of an old friend of the jobbing support bod has delivered a new kernel to help with fixing Microsoft's finest.

    It used to be called the System Rescue CD, but who uses CDs any more? Enter SystemRescue, an ISO image that you can burn, or just drop onto your Ventoy USB key, and which may help you to fix a borked Windows box. Or a borked Linux box, come to that.

    SystemRescue 9 includes Linux kernel 5.15 and a minimal Xfce 4.16 desktop (which isn't loaded by default). There is a modest selection of GUI tools: Firefox, VNC and RDP clients and servers, and various connectivity tools – SSH, FTP, IRC. There's also some security-related stuff such as Yubikey setup, KeePass, token management, and so on. The main course is a bunch of the usual Linux tools for partitioning, formatting, copying, and imaging disks. You can check SMART status, mount LVM volumes, rsync files, and other handy stuff.

    Continue reading

Biting the hand that feeds IT © 1998–2022