The resulting chips should find homes in high-performance computing, networking, gaming and other applications that require fast access to data stored in memory. And they should appear in physical products by the first half of next year. Eventually, this will have applications in cloud computing and even data analysis. According to the consortium, a single HMC offers a 15x performance increase and uses 70 percent less energy per bit when compared to today’s memory.

When I covered the launch of the effort in 2011, I explained the problem this new chip would solve:

The chip industry is really good at making a CPU that does calculations faster, but it hasn’t been able to make memory chips fast and dense enough to feed the cores enough information to keep up with the CPU’s capabilities. So what chips are left with is a massively large brain that stands idle sometimes while it waits for information to come to it. That idle time burns power and reduces the overall performance of a computer — and it’s becoming a bigger deal as both power and performance are being pushed to the edge.

Since that story was written, the consortium — founded by Samsung and Micron — has grown to more than 100 companies. The members now include big name vendors and users such as ARM, IBM, Microsoft, SK Hynix and HP as well as many smaller and specialty chip firms. The details of the announcement today are about how other chips will connect with the hybrid memory cube.

The structure of the chip is very different from traditional densely-packed DRAM modules. It’s stacked, which is a common way chipmakers have tried to pack in more memory — but instead of stacking the DRAM and connecting it via interconnects on the outside of the chip, the HMC has holes through the module with nanowires connecting the memory modules.

This process, known as Through Silicon Vias or TSV is of growing interest in the industry as it slogs down the path of making 3-D chips. But anytime you change a core silicon element significantly you have to figure out a lot of things –from the hardware layer all the way up to the applications. Today’s spec details how things will shunt bits to and from the hybrid memory cube. From the release:

“The achieved specification provides an advanced, short-reach (SR) and ultra short-reach (USR) interconnection across physical layers (PHYs) for applications requiring tightly coupled or close-proximity memory support for FPGAs, ASICs and ASSPs, such as high-performance networking, and test and measurement. The next goal for the consortium is to further advance standards designed to increase data rate speeds for SR from 10, 12.5 and 15Gb/s up to 28Gb/s. Speeds for USR interconnections will be driven from 10 up to 15Gb/s. The next level of specification is projected to gain consortium agreement by the first quarter of 2014.”

In short, the HMC will support 10 gigabit per second data rates at a minimum for both when they are the same board (short reach) and when they are packed even more tightly (around two to three inches) and get faster over time. Wicked fast.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• The future of mobile: a segment analysis by GigaOM Pro
• Who and what to watch in the new era of the living room
• CES 2012: a recap and analysis

This is how I have broken them down, and where I think things are heading based on my talks with vendors and customers in all of these industries, but I hope to hear from others who may have different opinions. Let’s get to it.

Enterprise: This is the traditional IT system, comprised of what might be a mix of conventional servers that may or may not be virtualized. This is the world where people buy HP(shpq), Dell( s dell) or IBM servers, specialized data warehousing solutions and oodles of Cisco and Juniper networking gear. These companies probably also have a few specialty Sun or Power PC machines supporting older applications that they don’t want to touch.

Their legacy applications depend on this stuff and they have a lot of legacy applications! This is an area where there’s plenty of money to be made, but it’s not where the growth in infrastructure spending will come from. For startups, especially those commercializing open source technology that’s interesting to the enterprise, this is a market that many overlook at first, and only later find themselves getting called into it with requests for proposals on Hadoop distributions or software-defined network help. The companies will buy some SaaS offerings and maybe a few test infrastructure-as-a-service efforts for non-crucial apps, but aren’t leaving their old infrastructure behind. They also don’t want to mess with tinkering, so issues like openness aren’t as essential as ease of deployment and management.

Webscale: Companies such as Google, Facebook, Yahoo, sections of Microsoft and other large online properties fit in this category. For these companies, IT isn’t just a cost of doing business, it’s the enabler for their business, much like a kitchen is for a restaurant. So just how Taco Bell tries to streamline a few ingredients into many cheap menu options in order to keep costs low, these companies streamline their hardware into a few highly optimized pools of computing for what will likely become several services offered on their platform.

These companies buy servers by the rack, and have the engineering resources to write code and implement new technologies that can save them money or speed up their ability to deliver services. Issues like interoperability and openness are important to them because they want their pieces to be as modular and as programmable as possible so they can tweak it to their needs. While their numbers may be few, they are a huge and growing segment of the market.

Cloud: I divide cloud into two categories. The more modern clouds, such as Microsoft Azure or Amazon’s Web Services that look more like the webscale architectures, and telco and service provider clouds that oftentimes have more of the enterprise gear (or specially built gear from enterprise vendors for the cloud.) An example of such gear would be Cisco’s Unified Computing System. In general the clouds following the webscale model may have some equipment closer to the data-specific or HPC (high-performance computing) clouds, but are content to put the engineering effort into making their clouds work optimally themselves. They will also have similar cost imperatives, continually worrying about how to continue scaling out without driving up costs.

As for the telco clouds and hosting companies that are getting into cloud environments, the industry seems to be in a state of flux, as the providers of such clouds realize that their first efforts were often times built around replicating enterprise environments and trying to force them to scale. Some, like Rackspace, have taken a different tack, throwing their hats in with Open Compute and the webscale guys while others have tried to use their existing gear (and the expertise of outside engineers) with cloud optimized software such as Open Stack or the products from Joyent or Eucalyptus.

HPC and Data: This is the area where I’m having the most trouble, in part because data analytics is so new. I lumped them together in the last few years mostly because both styles of computing are able to take advantage of massively parallel compute architectures and require fast interconnects. But there are differences, especially as low-power ARM-based chips tied together with fabrics enter the equation. Those smaller chips can be more energy-efficient for processing data, because many of the problems associated with data can be broken up into really small bits whereas in high performance computing powerful cores are still the norm — they might be massively parallel and distributed, but these are some of the brawniest cores out there. I think as we continue in 2013, we’ll see these categories split.

I’m also curious how many categories we will have in the next three years or so and what pieces each style of computing will borrow from another. As I said before, this is my mental map of computing influenced by the need for performance, cost and fast interconnects combined with the people on the ground at the buyer and their willingness to tinker. I’d love to hear from y’all about your take on these categories and where they might be headed.

Finch image courtesy of Flickr user Keith Ellwood.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Cloud computing infrastructure: 2012 and beyond
• Infrastructure Q1: IaaS Comes Down to Earth; Big Data Takes Flight
• Infrastructure Overview, Q2 2010

While it seems that chip industry has changed a lot in the last five years or so, that’s nothing compared to what will have to happen going forward. In the last five years the advent of iPhone — which brought real computing on a small, mobile form factor — has upended Intel’s dominance in the PC market. Just last week mobile chip maker Qualcomm surpassed Intel in market cap.

And on the server side and in high performance computing, the demand for more performance in a more efficient package has led to the adoption of new accelerator chips. The Top 500 list of the world’s fastest supercomputers came out Monday and showcases this trend with 62 supercomputers on the list using an accelerator. The fastest computer, called Titan, contains a mix of x86 chips and graphics processors to reach speeds of 17.59 petaflops. It was only four years ago that a GPU-powered computer had even made this list.

The Intel Xeon Phi coprocessor

Already the mobile industry uses a mix of cores to divide different processing jobs on a mobile device as a way to send the right job to the right processor, but also to save battery life by turning cores off when not in use and using a lower power core if possible for a job.

But accelerators can only take the chip industry so far, and are aimed at the high performance computing market. A multipetaflop machine still consumes a lot of power. One way to reduce that power consumption and boost performance is to cram more transistors on a chip by shrinking the space between transistors and reducing the overall size of the chip. This is called moving down the process node, but as the chip industry approaches feature widths of 20 and 14 nanometers, manufacturers are resorting to ever more complicated structures and materials to ensure that the chips perform well.

We’ve covered Intel’s 3-D transistors, Soitec’s new wafers, new materials and a variety of other breakthroughs all designed to keep this process shrinking. But researchers at McGill University have demonstrated that once you get to the atomic scale at levels of sub-five nanometers, the electric current on the chip gets all wonky — delivering a fourfold decrease in current. The researchers theorize when they shrink the channel that the electrons pass through they introduce “quantum weirdness.”

The “quantum weirdness” changes the way the current flows through a channel and so far those changes behave in unpredictable ways because the particles researchers are working with are at the atomic or sub-atomic scale. Which relates the research the top-level chip guys are doing to quantum theory and Schrodinger’s Cat. While Schrodinger’s famous thought experiement is an effort to articulate that the act of observing a quantum particle may change its state, the thought experiment is based on the fact that at subatomic scale materials behave differently.

From a write-up on the research:

As feature sizes in future chips shrink to the level of atoms, the resistance to current no longer increases at a consistent rate as devices shrink; instead the resistance “jumps around,” displaying the counterintuitive effects of quantum mechanics, says McGill Physics professor Peter Grütter.

This means that as chipmakers develop smaller processors or build any electronics with nano materials, the laws of physics will demand they change their materials and likely the structure of those materials, as well as conduct basic research to understand the new rules of playing at the atomic scale. Fortunately scientists at universities and researchers inside labs at place like IBM and HP are thinking about quantum physics and how it will change the rules of semiconductor manufacturing. And yes, there are some that are skipping the electronics part of the equation and skipping straight ahead to quantum computers, but I think our digital bits are here to stay for a few more decades.

Schrodinger’s cat image courtesy of Wikipedia.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Supercomputers and the Search for the Exascale Grail
• Listening platforms: finding the value in social media data
• The Internet of things: creating tomorrow’s health care

That’s why Morgridge Institute, a non-profit biomedical researcher based in Madison, WI, used Cycle Computing’s software atop Amazon Web Services infrastructure to process and index human stem cells to build an extensive knowledge base. Morgridge won Cycle Computing’s Big Science Challenge earlier this year which gave it access to Cycle’s technology and Amazon’s cloud, and it has now completed its run, which topped out at 8,000 cores and represented about a million hours, or 115 compute years of work, according to Jason Stowe, CEO of Cycle Computing. All for $0.0175 per compute hour.

Building an encyclopedia of cells

Morgridge’s goal was to create a knowledge base, or an index of associations between genes and the types of cells those genes could turn into, Stowe explained in an interview. “They can turn into pluripotent stem cells, which are very similar to embryonic cells in that they can potentially differentiate into any sort of cell.” The problem with using adult stem cells in the past was that a liver cell stayed a liver cell — pluripotent cells allow researchers to design the cells they need.

Morgridge Institute’s Victor Ruotti.

Victor Ruotti, molecular biologist at Morgridge, was thrilled to get the resources. “We wanted to take cells and compare them to other cells. We took as many samples as we had and gathered as much information as we could. We compared every cell type to every other tissue we had and built a database to say which cell types there are.”

That accumulated knowledge can help researchers figure out how to build the types of cell structures they need for experimentation and research.

The sample size was not that large, 124 samples, but each sample had more than 20 million data points to be compared. “When you multiply that out, it’s a very complex and resource intensive problem,” Ruotti said. Netted out, that’s a total of 1,003,303 core-hours against 11,955 pairs of samples processed, Cycle said.

The work ran on a mix of Amazon EC2 spot instances, including some high-memory instances on Centos.

Goal: curing disease in a petri dish

“We’re in an exciting phase in stem cell research [with pluripotent cells]. If a doctor needs cardiovascular cells to work on vessel or artery constructions, he can get them,” Ruotti said. “Medicine will change because doctors will be able to treat diseases in a petri dish. We can take a disease and simulate it in the lab, treat it in the lab, hopefully cure it there, and then implement the same cure in patients.”

Cycle Computing is determined to show that important workloads can run efficiently on low-cost public cloud infrastructure. Last year, it helped  Schrödinger Pharmaceutical spin up a 50,000-node Amazon cluster for its computational drug design work.

“We did a large run for a Big 5 pharma last year and the most common comment we got from the articles was, ‘I wonder if I could play “Call of Duty 4″ on this massive supercomputer.’” Stowe said. “What worried me about all the glitter was people would miss what is truly gold, which is that scientists can access world-class compute infrastructure at very reasonable cost and in a short time frame.”

Feature photo courtesy of  Flickr user CodonAUG

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• What Amazon’s new Kindle line means for Apple, Netflix and online media
• Microsoft Azure: What It Is, What It Costs and Who Should Care
• How tomorrow’s mobile-centric data centers will look

The problem for computational chemists and biologists is there’s a trade-off between accuracy and speed. This  ”Naga” compute environment built by Cycle atop the Amazon cloud eased that tradeoff, said Ramy Farid, president of New York-based Schrödinger, which specializes in computational drug design.

“We’ve got these really accurate methods but they would take months on a normal cluster. The problem is we want to do the best possible science fast,” Farid said in an interview this week.

That’s where Cycle Computing comes in. The company has made its name building high-performance computing atop AWS infrastructure and has previously deployed 10,000 and 30,000 core clusters on the cloud. For Schrödinger, it upped the ante to 50,000 cores. The alternative in this case would be for Schrödinger to build its own 50,000-core cluster or log time on a supercomputer, said Cycle Computing CEO Jason Stowe.

HPC for rent

“Practically speaking, the latter is impractical for a for-profit company and is generally restrictive. If you’re an academic wanting time on the San Diego Super Computer, for example, you’ll have months of wait time to get approved and even then you get a limited-time window — so if something with your software is not working at that time, you’re out of luck.” And, big supercomputers don’t tend to run the kinds of software these companies want to run. “The beauty of the cloud is it runs your flavor of Linux and other software,” Farid said.

On the other hand, building a 50,000 core cluster could easily cost $20 million to $30 million, he said. The Schrödinger project, by contrast, cost about $4,850 per hour to run.

For this trial, the cluster had access to all regions of AWS and used all of them in some capacity. The application used the various EC2 APIs to provision the resources.

“All the compound data for analysis was uploaded into S3 [Amazon's Simple Storage System]. The cluster was provisioned along side it and grabbed data from S3 to run the calculations and then pushed it back into S3,” said Crowe, who will be talking about the implementation Thursday at an Amazon event in New York. The application also took advantage of some Amazon IP address and DNS capabilities.

Finding a needle in a very big haystack

Farid could not talk much about the specific research purpose other than to say the goal was to find compounds that could be developed into drugs that fight a type of cancer.  The use of the huge cluster enabled Schrödinger to use the more accurate version of its Glide software — in the past it would have had to use the less accurate screen and perhaps miss some compounds that could be extremely useful.

“The problem they’re solving is amazing. It’s like the target is a lock and the confirmation is a key — what Ramy’s software does is let you simulate 21 million keys potentially matching that lock. There are typically a number of false negatives and the reason is you can’t sample all the orientations [of key to lock] properly so it will not show a match that could actually be a match. You could miss amazing drugs that could have an impact,” Crowe said.

The use of this huge cluster enabled Schrödinger to run its much more accurate, but much more compute-intensive version of its screening software and find compound candidates the other software may have missed.

The result of the three-hour run? “We identified  a number of compounds that we will purchase and test,” Farid said.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• 2012: The Hadoop infrastructure market booms
• A near-term outlook for big data
• NewNet Q2: Google closes the quarter with a bang

Cray's XK6 supercomputer

It looks like Oracle has some competition when it comes to selling big iron for big data. On Wednesday, Cray, the Seattle-based company best known for building some of the world’s fastest supercomputers, said it’s getting into the big data game. A new division within Cray, called YarcData, will leverage Cray’s experience working within data-intensive environments for customers such as Boeing in order to woo large-enterprises with big data needs.

Cray was short on details in a press release announcing the new division, but new YarcData SVP and GM Arvind Parthasarathi, formerly of Informatica is quoted saying, “YarcData is the nexus of the world’s most advanced technologies from Cray being applied to solve the world’s most challenging Big Data problems.” The natural leap is that Cray will design parallel-processing systems capable of incredible data throughput — something already required in the supercomputing space, where incredible processing capacity would be wasted without a steady data stream — but that will support today’s popular big data tools (e.g., Hadoop, analytic databases and predictive analytics software).

This type of system could be very valuable for organizations such as banks and intelligence agencies that want to run big data workloads as fast as possible — even process streaming data in real time– and the deep pockets to pay for Cray’s presumably pricey systems. Despite the fact that big-data framework Hadoop gained popularity in part because it’s designed to run on commodity hardware, there’s always a place for high-end hardware when milliseconds really do matter, and there’s something to be said for pre-configured systems that take the guesswork out of building a big data environment, as I explained recently in a piece for GigaOM Pro (sub req’d).

Cray isn’t alone in pushing this high-performance, enterprise-focused big data vision, though. Oracle made a splash in October when it announced a Big Data Appliance that marries Hadoop, R, NoSQL and other technologies to the high-end hardware Oracle obtained when it bought Sun Microsystems. IBM also has an extensive big data software portfolio complemented by a systems business that includes supercomputers, as well. And although it doesn’t have an HPC pedigree like the others, Teradata has years of experience building systems optimized for analytics.

Cray won’t likely become a household name in the big data world, and its notoriously secretive customers might never divulge what they’re using its analytics products for, but there certainly is a market — however small — for super-big, super-fast and super-expensive data.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Buying into big data appliances
• The importance of putting the U and I in visualization
• A near-term outlook for big data

But there are a few recent developments that, taken together, are creating an even more powerful efficiency boost: one that puts resources that were once limited to well-funded corporations and research universities within the reach of a new generation of startup founders. Perhaps it’s time entrepreneurs took advantage of this new environment to solve larger problems, instead of building yet another lifestyle app.

The way I see it, the big components at play here are:

New money is investing in big ideas

Even though the larger economy is rocky, there are a lot of people just itching to pour money into the next big technological thing — hence pre-launch photo sharing startups that net $41 million in funding. While many tech investors are focused on funding sure short-term bets (i.e. the tried and true realm of web and mobile apps), there’s a budding sect aggressively looking to invest in larger, long-term innovations.

Peter Thiel’s Breakout Labs is one of the most explicit examples of this. As we reported at the program’s launch last month, Breakout Labs will aim to fund nascent research proposals: opportunities too early stage or radical to attract dollars from VCs or government grants. Basically, Thiel, who recently told the New Yorker he doesn’t consider the iPhone to be a major technological breakthrough, is saying: Enough with the toys and games. It’s time for us to make something big.

Supercomputers are going mainstream

The next “something big” in tech might not require all that much money to make.

If you want to build something really complex — think aeronautics, new pharmaceutical drugs, medical devices, jet engines, and the like — you need high-performance computing (HPC). HPC solves advanced computational and scientific problems by using a massive amount of computing power to solve very complicated problems that involve a lot of moving parts.

HPC facility at Argonne National Labs (attribution below)

It has historically been so prohibitively expensive to do HPC that only entities such as governments, militaries, well-funded universities, or huge corporations have the kind of access to the machines needed for computational fluid dynamics problems and the like.

But last year, Amazon started offering HPC as a service with “Cluster Compute,” making high-performance computing available in the same way that EC2 made regular servers available in the cloud. Earlier this month, Amazon souped up its Cluster Compute offering significantly — now, Amazon’s HPC-as-a-Service offering provides access to one of the world’s top 500 supercomputers for around $1,000 per hour. Meanwhile, tools such as CUDA and OpenCL give programmers the ability to harness massive numbers of compute cores without having to learn a special programming language.

This takes HPC out of the realm of scientists and makes programming for massively multicore HPC systems accessible to software engineers. What Amazon’s EC2 did for democratizing the ability to develop scalable web apps, HPC-as-a-Service can do for democratizing the ability to solve computationally heavy engineering problems or build gigantic predictive models.

3-D printing is becoming a reality

Printed engine prototype by Mcor Technologies

Once challenging technology problems have been mastered with the help of HPC, some of the solutions will need to be prototyped and put into physical production. This is still a very labor- and cost-intensive process, which is a big reason why many startups prefer to stay in the virtual realm. But the emergence of viable 3-D printing technology is on the cusp of changing that, making it cheaper and easier than ever before to make a physical prototype of a new design.

How much of a reality is 3-D printing today? It’s now available at the consumer level with a startup called MakerBot, which makes a 3-D printer called a Thing-O-Matic. The Thing-O-Matic costs $1,200 and makes relatively simple items such as small toys and gadgets like bottle openers on demand. Three-dimensional printers from companies such as Mcor Technologies are aimed at making more complex prototypes for enterprise-level applications.

Of course, startups have the option of skipping the prototype step and selling simply the IP of their HPC-developed designs to a larger company. But if a startup wants to have more control over the production of what it has made, 3-D printing brings that much more within small companies’ reach.

What will be the hot startup of the next era?

If everything works out as it should, the smart, early stage entrepreneurs of the near future won’t be thinking about how to build the perfect restaurant recommendation app. Instead, they’ll devote their energy to designing a more efficient airplane wing to conserve jet fuel, or a tiny device that can perform real-time monitoring of kidney enzyme levels, or an even more awesome landing gear apparatus for the next Mars Rover. Starting the next SpaceX or Virgin Galactic won’t need the kind of funding that only an Elon Musk or Richard Branson can provide.

Today, the lion’s share of companies that emerge from incubators such as Y-Combinator and 500 Startups deal in consumer-focused web apps. Here’s hoping that in the near future, incubators will look for startup founders who are taking real advantage of their new-found access to serious tech tools to build bigger and bolder products. It seems to me that driving toward that kind of world is where the attention of the tech industry — and the media that covers it — should focus.

“What’s Next?” image courtesy of Flickr user Crysti

Image of the HPC facility at the Center for Nanoscale Materials at the Advanced Photon Source courtesy of Flickr user Brian Howard on behalf of the Argonne National Laboratory.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Crowdfunding’s rapid growth and future opportunity
• Facebook’s IPO filing: ideas and implications
• Flash analysis: lessons from Solyndra’s fall

Servers? We don't need no stinkin' servers!

Both mobile and high-performance computing are placing huge power efficiency and performance demands on chips, but the $64,000-question is how long until such extreme computing use cases hit the server mainstream. Asked another way, the question becomes, how long until Amazon Web Services adopts ARM-based servers?

Or perhaps it isn’t ARM-based servers, but a variation on an Intel chip that takes its architecture from some of the more innovative and energy-efficient silicon out today. For example, Adapteva, a startup I profiled back in May, released a 64-core chip on Monday that can deliver 70 gigaflops of performance per watt. If you don’t speak gigaflops, that’s okay: It basically has done what Intel and certain countries have deemed impossible with the current generation of silicon.

The government of the European Union, in its quest for an exascale supercomputer, has targeted a goal of getting 50 gigaflops per watt (Intel also thinks this would work). In conversations with folks that design supercomputers, the thinking is that a conventional x86-based machine would require the equivalent of a power plant or two to run. That includes all the networking and other trimmings, but the bottom line is that Adapteva’s chips deliver more flops per watt, and that’s a good thing.

It’s not just supercomputers though. Adapteva’s CEO Andreas Olofsson told me the company is only targeting computing extremes such as supercomputing and mobile phones because that’s where the power efficiency pain point is today. Because mobile phones run on batteries, and no one wants a smartphone that dies after two hours, vendors using ARM’s power-efficient architecture have dominated the mobile sector. When Microsoft adapted Windows to run on ARM, it spoke volumes about the need for power efficiency. Windows is one of the most x86-oriented pieces of software out there.

These shifts in usage profiles and the high demand for compute are creating opportunities for companies like Adapteva, so it’s not too far-fetched to wonder how long until that pain point hits conventional servers.

I often cover companies that are hoping the combination of monolithic applications and a desire to reduce power consumption means webscale and cloud vendors will embrace a new architecture. Companies such as Tilera, SeaMicro, Adapteva, Calxeda and others are all betting the next gear Facebook or Amazon buys will be their hardware or contain their chips.

However, even in its state-of-the-art data center optimized at the server level to be energy-efficient, Facebook challenged the way servers and data centers are built but didn’t touch the silicon itself. So, clearly, the webscale world isn’t champing at the bit to replace the x86-based servers their applications are running on. SeaMicro even has shown charts demonstrating that the CPU is only a third of the power associated with running a server, which means there’s still plenty of fat to trim. Of course, SeaMicro is building a server that trims that non-CPU fat and runs Intel’s Atom chips.

However, the global demand for energy and the supply we currently have are reaching a point where it’s safe to conclude that power consumption will become a greater cost and constraint associated with operating data centers. And at some point, building in cooler climates, hot and cold aisle containment, and even newly designed servers won’t be enough if the silicon itself is too hot.

So the question isn’t if, but when, server companies abandon the PC-style architecture. Perhaps Intel, AMD or Via will continue to tweak x86 silicon until it can perform more calculations using less power, or perhaps it will be time for Amazon or Microsoft Azure to go with ARM, Tilera — or even Adapteva.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• The fourth quarter of 2012 in cleantech
• How tomorrow’s mobile-centric data centers will look
• Cloud computing’s impact on chip and hardware design

These things are expensive.

The face of high-performance computing is changing. That means new technologies and new names, but also familiar names in new places. Sure, cluster management is still important, but anyone that doesn’t have a cloud computing story to tell, possibly a big data one too, might starting looking really old really quickly.

We’ve been seeing the change happening over the past couple years, as Amazon Web Services and Hadoop, in particular, have changed the nature of HPC by democratizing access to resources and technologies. AWS did it by making lots of cores available on demand, freeing scientists from the need to buy expensive clusters or wait for time on their organization’s system. That story clearly caught on, and even large pharmaceutial companies and space agencies began running certain research tasks on AWS.

Amazon took things a step further by supplementing its virtual machines with physical speed in the form of Cluster Compute Instances. With a 10 GbE backbone, Intel Nehalem processors and the option of Nvidia Tesla GPUs, users can literally have a Top500 supercomputer available on demand for a fraction of the cost of buying one. Cycle Computing, a startup that helps customers configure AWS-based HPC clusters, recently launched a 10,000-core offering that costs only $1,060 per hour.

Hadoop, for its part, made Google- or Yahoo-style parallel data-processing available to anyone with the ambition to learn how to do it — and a few commodity servers. It’s not the be all, end all of the big data movement, but Hadoop’s certainly driving the ship and has opened mainstream businesses to the promise of advanced analytics. Most organizations have lots of data, some of it not suitable for a database or data warehouse, and tools like Hadoop let them get real value from it if they’re willing to put in the effort.

New blood

This change in the way organizations think about obtaining advanced computing capabilities has opened the door for new players that operate at the intersection of HPC, cloud computing and big data.

One relative newcomer to HPC — and someone that should give Appistry and everyone else a run for their money — is Microsoft. It only got into the space in the late ’00s, so it didn’t have much of a legacy business to disrupt when the cloud took over. In a recent interview with HPCwire, Microsoft HPC boss Ryan Waite details, among other things, an increasingly HPC-capable Windows Azure offering and “the emergence of a new HPC workload, the data intensive or ‘big data’ workload.”

Indeed, Microsoft has been busy trying to accommodate big data workloads. It just launched an Azure-based MapReduce service called Project Daytona, and has been developing its on-premise Hadoop alternative called Dryad for quite some time.

The latest company to get into the game is Appistry. As I noted when covering its $12 million funding round yesterday, Appistry actually made a natural shift from positioning itself as a cloud software vendor to positioning itself as an HPC vendor. Sultan Meghji, Appistry’s vice president of analytics applications, explained to me just how far down the HPC path the company already has gone.

Probably the most extreme change is that Appistry is now offering its own cloud service for running HPC computational or analytic workloads. It’s based on a per-pipeline pricing model, and today is targeted at the life sciences community. Meghji said the scope will expand, but the cloud service just “soft launched” in May, and life sciences is a new field of particular interest to Appistry.

The new cloud service is built using Appistry’s existing CloudIQ software suite, which already is tuned for HPC on commodity gear thanks to parallel-processing capabilities, “computational storage” (i.e., co-locating processors and relevant data to speed throughput) and Hadoop compatibility.

Appistry is also tuning its software to work with common HPC and data-processing algorithms, as well as some it’s writing itself, and is bringing in expertise in fields like life sciences to help the company better serve those markets.

“Cloud has become, frankly, meaningless,” Meghji explained. Appistry had a choice between trying to get heard of the noise of countless other private cloud offerings or trying to add distinct value in areas where its software was always best suited. It chose the latter, in part because Appistry’s products are best taken as a whole. If you need just cloud, HPC or analytics, Meghji said, Appistry might not be the right choice.

One would be remiss to ignore AWS as a potential HPC heavyweight, too, although it seems content to simply provide the infrastructure and let specialists handle the management. However, its Cluster Compute Instances and Elastic MapReduce service do open the doors for other companies, such as Cycle Computing, to make their mark on the HPC space by leveraging that readily available computing power.

The old guard gets it

But the emergence of new vendors isn’t to say that mainstay HPC vendors were oblivious to the sea change. Many, including Adaptive Computing and Univa UD, have been particularly willing to embrace the cloud movement.

Platform Computing has really been making a name for itself in this new HPC world. It recently outperformed the competition in Forrester Research’s comparison of private-cloud software offerings, and its ISF software powers SingTel’s nationwide cloud service. Spotting an opportunity to cash in on the hype around Hadoop, Platform also has turned its attention to big data with a management product that’s compatible a number of other data-processing frameworks and storage engines.

Whoever the vendor, though, there’s lots of opportunity. That’s because the new HPC opens the doors to an endless pipeline of new customers and new business ideas that could never justify buying a supercomputer or developing a MapReduce implementation, but that can enter a credit-card number or buy a handful of commodity servers with the best of them.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• 9 Companies that Pushed the Infrastructure Discussion in 2010
• Infrastructure Overview, Q2 2010
• Cloud computing 2013: how to navigate without a map

The concept of an accelerator is a familiar one in super computing, where the addition of a specialized massively multicore chip, such as a graphics processor or custom chip, is becoming more common. But unlike a GPU, the 64-core Adapteva chip only operates at one watt. To understand how powerful that is from an energy efficiency perspective, a four-to-eight-core server chip could operate at anywhere from 60-120 watts. And the challenge of building the next generation of supercomputers is constrained by the power demands such massive supercomputers would require.

But Andreas Olofsson, the founder and CEO of Adapteva, isn’t focusing on the HPC market at first–despite asserting that his design can scale to a 4,096-core design that would run at 64 watts. He said that while there is plenty of talk about low-power computing, “As long as you can plug something into a wall, the need for low power goes down significantly. It’s only a little bit painful.” However, in the mobile world where devices need to run all day, yet avoid bulky batteries, power consumption is at a premium. So that’s where Adapteva will focus for its big push (the company has some military applications as well).

The company began in 2008 and has managed to raise $2 million in funding from angels and boardmaker BittWare, its first customer. Amazingly, with that small amount of funding it has managed to have three versions of its chip built, making the startup incredibly capital efficient. However, the goal isn’t to build chips for the mobile market, but to license the technology, much like ARM, the firm behind the most common architecture in mobile phones, does. BittWare will manufacture the Adaptevea chip design — called the Epiphany– on its boards.

But in a highly competitive market, and especially on smarthphones, where space on the board is at such a premium, will device makers really embrace an unproven and as-yet-unneeded chip? Olofsson has two more difficult tasks to accomplish (since he’s apparently taken care of the hard task of building and designed a 64-core chip that runs at 1 watt for less than $2 million.) He must explain to board makers, chip firms and device makers why gadgets need this rather foreign accelerator chip, and he has to convince them that it makes sense to process data on the phone, rather than ship it over the cellular network.

The first task is made easier by the low-power envelope and by the fact that the full 64-core system on a system is fairly small — about 8mm square Olofsson said. Check out the model of the A5 system on a chip used inside Apple devices provided below to see how much space the Epiphany chip can take up. It would have to replace existing GPUs in this case.

The second task may be made easier by people’s desire to handle tasks such as speech or facial recognition or intense video games on their mobile devices. Olofsson argues that the latency inherent in sending even voice recognition to a server is problematic and that gameplay is impossible. Plus it costs more in terms of data charges and can drain the battery. “If you can keep the radio quiet and use the processing locally the battery life gets better,” he said.

Like many visionaries pushing a new technology he’s not entirely sure how the Epiphany could change mobile computing, but he’s certain that by boosting performance on smartphones to this degree it will. I’m eager to see if mobile device makers agree.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• What the Google-Motorola deal means for Android, Microsoft and the mobile industry
• Bluetooth to Feel Blue as Personal Area Network Battles Loom
• HTML5’s a Game-Changer for Web Apps