Furthermore, it is built for AI tasks.

Although IBM's response to the affordable supercomputer has been operational for some time, it has only lately made any concrete disclosures about its so-called Vela project.

Looking to its blog(opens in new tab) for more information, IBM disclosed that the study, which was written by five company employees, addresses issues with earlier supercomputers and their unpreparedness for AI tasks.

The business clarifies the choices it made regarding the use of low-cost but powerful hardware in order to adapt the supercomputer model for this future type of workload.

We need your assistance to create a better website for our users. By responding to our poll and sharing your thoughts and ideas about the tech sector in 2023, you can contribute. Your responses will all be kept private and secret, and it won't take long. Once more, thank you for your assistance in improving TechRadar Pro.

Although it's obvious that these supercomputers can handle demanding AI workloads, including the one built for OpenAI, the company behind the well-known ChatGPT live chat software, a lack of optimization has left traditional supercomputers at risk of lacking valuable power and having an excess in other areas, which results in an unnecessary expenditure.

Although it is well known that bare metal nodes are the best for AI, IBM wished to investigate making these available inside of a virtual machine (VM). Big Blue claims that the outcome is significant efficiency improvements.

We developed a method to expose all of the node's capabilities (GPUs, CPUs, networking, and storage) into the VM so that the virtualization overhead is less than 5%, which is the lowest overhead in the industry that we are aware of, after conducting a substantial amount of research and discovery.

Vela has four 3.2TB NVMe storage drives, 1.5TB of DRAM, and 80GB of GPU memory in terms of node architecture.

According to The Next Platform's estimation, if IBM wished to include its supercomputer in the Top 500 rankings, it would produce approximately 27.9 petaflops of performance, putting it in 15th place as of November 2022.

Although existing supercomputers are capable of handling workloads related to artificial intelligence, the rapid advancements in this field and the pressing need for cost-effectiveness highlight the need for such a machine.

IBM said its latest Q System One has double the Quantum Volume of its predecessor. If Quantum Volume doubles annually for nearly a decade quantum computing will become way more practical.

Foldable phones: Why they won't change the shape of the smartphone industry
IBM is outlining another milestone in quantum computing -- its highest Quantum Volume to date -- and projects that practical uses or so-called Quantum Advantage may be a decade away.

Big Blue, which will outline the scientific milestone at the American Physical Society March Meeting, made a bit of a splash at CES 2019 with a display of its Q System quantum computer and has been steadily showing progress on quantum computing.

In other words, that quantum computing buying guide for technology executives may take a while. Quantum Volume is a performance metric that indicates progress in the pursuit of Quantum Advantage. Quantum Advantage refers to the point where quantum applications deliver significant advantages to classical computers.

Quantum Volume is determined by the number of qubits, connectivity, and coherence time, plus accounting for gate and measurement errors, device cross talk, and circuit software compiler efficiency.

IBM said its Q System One, which has a 20-qubit processor, produced a Quantum Volume of 16, double the current IBM Q, which has a Quantum Volume of 8. IBM also said the Q System One has some of the lowest error rates IBM has measured.

That progress is notable, but practical broad use cases are still years away. IBM said Quantum Volume would need to double every year to reach Quantum Advantage within the next decade. Faster progress on Quantum Advantage would speed up that timeline. IBM has doubled the power of its quantum computers annually since 2017.

Once Quantum Advantage is hit, there would be new applications, more of an ecosystem and real business use cases. Consumption of quantum computing would still likely be delivered via cloud computing since the technology has some unique characteristics that make a traditional data center look easy. IBM made it's quantum computing technology available in 2016 via a cloud service and is working with partners to find business and science use cases. 

Here's how quantum computing and classic computing differs via our recent primer on the subject.

Every classical electronic computer exploits the natural behavior of electrons to produce results in accordance with Boolean logic (for any two specific input states, one certain output state). Here, the basic unit of transaction is the binary digit ("bit"), whose state is either 0 or 1. In a conventional semiconductor, these two states are represented by low and high voltage levels within transistors.

In a quantum computer, the structure is radically different. Its basic unit of registering state is the qubit, which at one level also stores a 0 or 1 state (actually 0 and/or 1). Instead of transistors, a quantum computing obtains its qubits by bombarding atoms with electrical fields at perpendicular angles to one another, the result is to line up the ions but also keep them conveniently and equivalently separated. When these ions are separated by just enough space, their orbiting electrons become the home addresses, if you will, for qubits.




READ MORE:

Watson Anywhere plan looks to free up AI platform across multiple clouds

IBM has announced it is liberating its Watson AI services to work across any cloud platform.

The news was announced at the opening of IBM's Think event in San Francisco and could mean a major expansion to the possibilities of what Watson can do, and where it can be deployed.

Going forward, IBM Watson services will be integrated with IBM Cloud Private for Data, meaning they can be run on any environment, whether that be on-premise or in the field, or across any private, public or hybrid-multi-cloud - meaning companies can now take Watson to their data.

The launch will include services such as the Watson Assistant, which offers the chance to build AI conversational interfaces into apps and devices, and Watson OpenScale, which can deploy large-scale rollouts of AI across multiple devices.

• What is AI? Everything you need to know
• 5 unusual things you can do with IBM's Watson
• How IBM Watson will improve your morning commute

Watson breaks free

IBM says that the move will also allow Watson customers to do more with the platform, as until now, the platform has been limited to IBM's own cloud, meaning many users have had to store and manage data across multiple siloes and with different suppliers.

"Businesses have largely been limited to experimenting with AI in siloes due to the limitations caused by cloud provider lock-in of their data," said Rob Thomas, General Manager, IBM Data and AI.

"With most large organizations storing data across hybrid cloud environments, they need the freedom and choice to apply AI to their data wherever it is stored. By breaking open that siloed infrastructure we can help businesses accelerate their transformation through AI."

The news also comes as IBM looks to finalize the $34bn acquisition of open-source software firm Red Hat, suggesting that Watson could be about to become even more accessible for developers around the world.

The services will be linked to rival, non-IBM clouds using Kubernetes, again potentially opening up more opportunities for open-source developers.



READ MORE:

Does IBM’s announcement of its new quantum computing system mean that quantum computing must be on your technology roadmap?

IBM unveiled a complete "quantum computing system," IBM Q System One, last week. What's more, it chose to do it at CES in Las Vegas. Should you take IBM's claim of a "commercial" quantum computing system to mean that it is time to put quantum on your tech road map? The short answer is "no" -- but, as always, there are exceptions. Since there is a lot of confusion on quantum computers (QC), we think the market needs help to stay grounded in reality.

Here is my take on the announcement: It is a big step forward, but QCs are taking baby steps and the journey is a thousand miles long. To understand this, you need to let go of a bunch of assumptions. After that, you need to appreciate why IBM's announcement is significant without losing sight of the long road ahead of us. Let me break these down for you briefly.

START BY LETTING GO OF YOUR (DIGITAL) ASSUMPTIONS

As I've been diving deeper into quantum computing, I've discovered that I have been making a lot of assumptions because the things that work for classical computing turn out to be different for QCs. Let me help you avoid the same mistakes.

• Universal QCs don't have RAM or CPU clocks. Digital computing power is measured by how many 1s and 0s you can process and how quickly bitsets are processed in and out of CPUs using RAM as a storage buffer. QCs work with states between 1 and 0; however, there is no way to write these rich, complex intermediate quantum states to "quantum RAM." Thus, you can't think of QCs as working the same way that digital computers do because QCs cannot do sequential processing.
• Universal QCs today are batch problem solvers, not general-purpose programmable systems. We naturally assume that a universal QC is programmable and can do many different things using instructions, loops, or object calls. In fact, QCs require a completely different approach to programming logic. Without RAM, QCs are more like batch problem solvers. Digital data goes in, quantum magic happens, and digital data comes out that represents a probable answer to the input problem. This is good for computationally intensive tasks in optimizations, materials, and chemistry, some machine-learning tasks, engineering, etc. It's not good for software as we typically understand it.

WHY Q SYSTEM ONE IS A BIG DEAL (AND WHY IT'S NOT)

With this information as background, you can start to see why IBM's announcement is significant and why it's not the bombshell that some media outlets imply. First, because it's a completely new computing paradigm, it's best to think of our current state as somewhere between Turing's WWII machine and the first mainframe. There is a lot of system and software engineering foundation that needs to be developed and given time to mature.

Here is what I think is significant about IBM's announcement and what is perhaps a bit of hype:

• Significant: It's a system. This is the most important thing about IBM's announcement. Up until now, quantum computing systems have been stitched together the same way we built computers in our garages back in the 1980s. IBM is claiming that it now has a fully integrated computing system. This is very important because instead of optimizing components, IBM can how to start to optimize the system as a whole. Until we can start engineering integrated quantum computing systems, we really can't drive consistent increases in power.
• Significant: It's modular. Not only is it a system, IBM claims that it's modular. That means components can be upgraded separately, which is big because it starts to make the quantum system commercially viable. If you wanted a more powerful version, you don't have to buy a whole new multimillion-dollar box. Given the exponential effects that could dramatically increase quantum computing's power over the next few years, being able to swap out for new components to increase performance is a good thing.
• Hype: It's the "first commercial QC." This is where I urge caution. IBM likes to tout being the first at things. And many times, it is -- but its claims also rev up the market, sometimes too far in advance. I can see the sense in IBM's claim because (A) you can buy it if you have deep pockets and (B) its systematic approach and modular design make it more commercially viable. However, IBM's claims can be interpreted as "ready for primetime"problem-solving. That's far from true, as I cover next.

HOW FAR WE HAVE TO GO

We have a long way to go before quantum computers can solve real problems that businesses have. No QC today (universal or annealing) can do anything better than a digital computer simulating a quantum computer. That is very, very important.

No quantum computer today can do anything better than a digital computer simulating a quantum computer.

What's more

Researchers don't know when QCs will be commercially useful. The technical term is "achieving quantum value." This means that on a cost-for-performance basis, QCs can solve a problem more economically than a classical computer. All the work being done today is searching for quantum algorithms that might someday (hopefully sooner rather than later) be commercially viable. In our report, "A First Look At Quantum Computing," we estimate that it will be three to five years before we see mainstream applications exploiting QCs to solve a few select business problems.

It's likely to take 10 to 20 years before QCs can break PKI encryption. The holy grail of quantum computing is something called error-corrected universal quantum computers. Today, universal QC qubits only stay stable for a short time. Due to the reasons mentioned above, you must solve your problem in that time or go back and start over. If vendors can design sufficient redundancy, they can theoretically create a stable, logical, error-corrected qubit. For reasons, I lay out in my report, "Quantum Is Not An Immediate Security Threat," this is likely to take 10 years. Until then, we won't have QCs that can factorize sufficiently large numbers to threaten PKI encryption.

Quantum machine learning is still largely theory. One of the most exciting use cases for QC is its application to improve machine learning. However, machine learning today involves a lot of optimization and vector math operations that QCs, because of their limitations, don't do very well. Experts are searching for quantum machine-learning algorithms that replace or supplement traditional iterative approaches, but this work is a very early stage. My point is: Researchers don't know exactly how transformational QC will be with AI. On the other hand, Google's quantum moonshot is called the Google Quantum AI Lab, so we know what the tech giant thinks.

What should the smart CIO do? First, read our previous research to get your team smarter, then give us a call to help you identify opportunities to investigate further.



SOURCE:

IBM makes progress on-chip fabrication without using lithography.

IBM has devised materials and processes that could help improve the efficiency of chip production at the 7nm node and beyond.

The company's researchers are working on challenges in the emerging field of 'area-selective deposition', a technology that could help overcome limitations on lithographic techniques to create patterns on silicon in 7nm processes.

Semi Engineering has a neat account of lithographic patterning and why at 7nm there's growing interest in an area-selective deposition.

Techniques such as 'multiple patterning' helped ensure integrated circuits kept scaling, but as chips have shrunk from 28nm to 7nm processes, chipmakers have needed to process more layers with ever-smaller features that need more precise placement on patterns.

Those features need to align between layers. When they don't, it leads to 'edge placement error' (EPE), a challenge that Intel lithography expert Yan Borodovsky believed lithography couldn't solve and which would ultimately impede Moore's Law.

In 2015 he encouraged the industry to investigate area-selective deposition, the area that IBM's researchers are exploring, which may one day be a successor to EUV lithography, the technique Samsung is now preparing for the fab after decades in the lab.

Fabs already use some forms of selective deposition to deposit materials on metal surfaces in devices. Area-selective deposition requires a tool that can deposit different combinations of material sets -- metals on metals and dielectrics on dielectrics -- on a device.

Rudy J Wojtecki, a researcher at IBM's Almaden Research Center, explains IBM's effort to improve the technology:

"With traditional methods of fabrication, this would require coating a substrate with resist, patterning the resist through an exposure step, developing the image, depositing an inorganic film and then stripping the resist to give you a patterned inorganic material.

"We found a way of depositing this inorganic film much more simply, using a self-aligned process, where we immerse a pre-patterned substrate in a solution containing a special material and then add that coated substrate it to a deposition chamber and you're done. We are literally able to grow a component of a device in a controllable manner at the nanoscale."

The group is using one of three main methods for area-selective deposition called 'atomic layer deposition' with a focus on using 'self-assembled monolayers' (SAMs).

It may help pave the way for hardware that better supports AI applications, such as three-dimensional structures.

"Once we develop methods of scaling this process, we can begin to integrate it as we build next-generation hardware, whether it is for new AI hardware or making devices at the 7nm technology node or beyond," Wojtecki said.

IBM isn't alone in developing techniques for area-selective atomic layer deposition, but Wojtecki argues his ability to tailor a chemical structure for demanding applications has enabled the development of "new polymerizations, materials and characterization methods" that may one day become scalable.


IBM says it's found a way of depositing inorganic film much more simply, using a self-aligned process.

IBM AI OpenScale, Multi-cloud Manager and IBM Security Connect equate to a strategic push by Big Blue to be the agnostic integrator and platform provider for key enterprise technologies.

IBM launched IBM AI OpenScale, a platform designed to help enterprises build, run, manage and operate artificial intelligence applications.

The launch of AI OpenScale is part of IBM's ongoing effort to become a management plane for AI and add transparency to so-called black box approaches. If AI sprawl isn't here today it soon will be and enterprises are likely to have a management headache ahead. IBM has been pushing for more AI transparency and tools that allow data scientists, as well as business executives, find flaws in models.

For IBM, AI OpenScale is one front of a multi-pronged strategy to position its wares as being more open and serve as an integrator for data, multiple clouds, and security analysis. IBM's big message: Serving as an agnostic technology provider can better help enterprises.

Beth Smith, general manager of Watson Data and AI at IBM, said AI today is a mesh of tools, models, and frameworks. "People use a variety of tools. Some are rolling your own," said Smith, who added that IBM AI OpenScale is an interoperable system that can support AI implementations.

IBM's AI OpenScale platform, which will be available later this year on IBM Cloud and IBM Cloud Private, will operate AI applications and debug them for things like bias wherever they were built or currently run. The platform will support frameworks such as Watson, Tensorflow, Keras, SparkML, AWS SageMaker, Azure ML and others.

Part of the AI OpenScale launch includes NeuNetS, which is a system that can automate and build AI. Smith noted that NeuNetS can save data scientists time and can narrow an enterprise skills gap.

IBM AI OpenScale is designed to explain how AI applications reach decisions, provide an audit trail and ensure AI models are fair at runtime.

Here are some points and examples of how AI OpenScale would work in practice:

• AI OpenScale manages and optimizes AI applications, but data scientists would build models in the framework of their choice.
• However, IBM AI OpenScale would automate many items in the AI development process. For instance, de-biasing would be automated. "AI OpenScale would bring fairness to an attribute in a model and does it in a way that doesn't alter the base model," said Smith.
• AI OpenScale would leave the original model alone but de-bias it with a new auto-generated model.
• NeuNetS would be used to fine tune AI and models and could speed up development time by months. NeuNetS has been in use at IBM for "several months," said Smith, who said the service would start out as a beta within the platform.

For IBM AI OpenScale to work within an enterprise the client would have to be able to point Big Blue to a direct endpoint where the AI black box resides. IBM wouldn't be able to manage embedded AI in another application. For instance, Salesforce's Einstein couldn't be accessed from AI OpenScale, but the CRM giant could use IBM's platform to manage its models that it embeds into applications.

Separately, IBM launched Multi-cloud Manager, an operations platform based on Kubernetes containers to manage public and hybrid cloud deployments.

The console from IBM is optimized on IBM Cloud but can integrate and manage workloads on clouds from the likes of Amazon Web Services, Red Hat and Microsoft. The Multi-cloud Manager runs on IBM Cloud Private.

According to Big Blue, the differentiator for its cloud management console is that it's based on open standards to manage data and apps across clouds.

Multi-cloud Manager, available this month, includes:

• The ability to interconnect different clouds, unify systems and automate operations.
• A dashboard to manage thousands of Kubernetes applications and data where its located.
• An integrated compliance and rules engine for enterprise policies and security standards.
• Automation to define how, where and when Kubernetes applications are deployed and how they are backed up.
• In addition, IBM outlined IBM Security Connect, an open platform that aims to aggregate and analyze federated security data across multiple systems and environments.
• 
  
• IBM Security Connect is based on machine learning and AI.

According to IBM, Security Connect will open its framework, micro-services, software developer kits and application programming interfaces for integration and development. IBM Security Connect will house the company's current Security App Exchange and all of its security applications.

The company also committed to using existing open security and protocol standards. IBM Security Connect will be available in the first quarter of 2019.


SOURCE:

Managing microservices is a critical issue since enterprises are increasingly built on them.

Big Blue's VP and CTO of cloud told ZDNet that Australia is punching above its weight in technological adoption after the federal government announced handing it AU$1 billion to do just that.

Microsoft's latest financial results show that it is now bigger than IBM has ever been -- but Google got there first.

Microsoft's annual revenues of $110.4bn took it past 'peak IBM', but Google did it even faster. (IBM and Google have different financial years so have yet to announce equivalent results.)

At the time, Microsoft had annual revenues of $7.5 million (not billion) and a staff of 40 that included a new hire from Procter & Gamble by the name of Steve Ballmer. By contrast, IBM had a turnover of $26.2 billion and a staff of 341,279 - it was about the same size as General Electric and Gulf Oil, but made more profits than both of them put together.

Big Blue's as-a-service revenue hit an annual run rate topping $11 billion, but the big growth was put up by the hardware unit.

• Global business services revenue was $4.2 billion, up 2 percent from a year ago.
• 
  
• Technology services and cloud platforms had second-quarter sales of $8.6 billion, up 2 percent from a year ago.
• 
  
• Systems revenue was $2.2 billion, up 25 percent due to growth in IBM Z mainframes. IBM launches 'skinny' Z mainframe designed for 19-inch standard data center rack
• 
  
• For the year, IBM said it expects non-GAAP earnings of at least $13.80 a share with free cash flow of about $12 billion.

Box Skills provide a framework for applying machine learning to the Box cloud content management platform.

IBM has announced new cloud availability zones to boost its capacity in several key regions around the world.

IBM and CA Technologies are combining hardware and software forces to bring more applications and cloud connections to the mainframe.

1. CA Brightside, a new mainframe management suite, will be available on zCloud so developers can create applications for the mainframe using existing tools.
2. CA Service Virtualization to test and modify applications. The tools make it easier for companies to develop and test mainframe applications in the cloud.
3. CA Mainframe Operational Intelligence for performance management and monitoring.
4. CA Data Content Discovery to find and protect data for compliance purposes.
5. CA Mainframe Resource Intelligence, which offers a SaaS-based way to automate management and optimize performance.

Under the partnership, IBM will extend its IBM Cloud Private and IBM Cloud Private for Data and middleware to Red Hat's OpenShift Container Platform and Certified Containers.

• IBM Cloud Private can deploy containers across Red Hat.
• Developers will be able to use IBM's cloud services for AI, the blockchain, and the Internet of things.
• IBM and Red Hat will serve as an integrated hybrid cloud platform.
• IBM also said that IBM PowerAI, the company's deep learning toolkit, will also be available on Red Hat's operating system.
• Red Hat and IBM are two of the larger open source software players, and their partnership dates back to 1999. In 2002, IBM and Red Hat outlined an enterprise Linux partnership, and the two companies have been advancing hybrid cloud development.

IBM got some help by currency fluctuations, but strategic imperatives revenue delivered growth.

1. Digital offerings are driving consulting growth and there were "modest" improvements in revenue in the first quarter relative to the fourth quarter.
   
2. IBM Z sales were driven by the z14 and new workloads. IBM Z sales were up 54 percent from a year ago.
   
3. Storage revenue declined after four straight quarters of growth.
   
4. First quarter revenue in the Americas was flat with EMEA up 1 percent and the Asia Pacific flat. Americas revenue saw growth in Canada and Latin America and a decline in the US.

IBM is hellbent on bringing the power and unique characteristics to a private or public cloud data center near you.

IBM has big ambitions for its barely visible computer, including helping combat fraud with blockchain tech.