The fifth edition of the Open Compute Summit held in San Jose, California, reunited industry leaders and innovators who shared with the audience their breakthroughs from the previous event onwards and their plans for the future.

Andrew Feldman, General Manager and CVP with AMD, took the stage to deliver a short but compelling presentation which summed up AMD’s accomplishments during 2013, as well as dared to make a couple of bold forecasts.

“Five years ago I participated in the first Open Compute Summit. It wasn’t clear at the time that this was going to be a ‘something’,” admitted Feldman. “When you have 3,500 people registered, and the sort of activity you’ve seen here, it’s clear that Frank (Frankovsky) and his team were on to something. It’s a pleasure to participate.”

Kicking off his presentation, Feldman asked the audience to take a look at two pictures taken from the same vantage point but 8 years apart, and notice how the way we interact with the world has changed during that time-frame.

“Not only did it change the way we interact with the world, but it put the data center into our daily experiences,” explained Feldman. “Every one of these videos on the right was stored in the cloud and ended up in a data center. Every one of them was emailed or tweeted about.”

New directions

“For me, the interesting thing in the direction the data center will go is the client side. That’s where the benefit of the data center is delivered. What’s interesting about these devices is they can’t do any interesting work. You have an Angry Birds’ player,” mocked Feldman, “not an interesting device.”

The data center of today changes our lives in so many ways, allowing the same device to display all the world’s information, one’s friends and social map, search, or even directions on how to get to a good taco restaurant.

“That’s what is so fascinating about the dynamic of a market we’re in right now. The handsets, the tablets, they present the answers; the real work is being done in the data centers,” noted Feldman.

He declared he’s not particularly interested in the Internet of Things, and he doesn’t know whether or not we’ll have beagles with radio-transmitting collars sometime in the future, but he’s sure of one thing: “we’ll have more users and more devices.”

“That will add tremendous demand on the data centers, forcing us to rethink the type of infrastructure we put there,” believes Feldman.

The General Manager and CVP with AMD had some simple observations:

• only a third of the world has internet access

“We don’t have to think of anything new, we have 2 thirds of the world wanting to have what we have here today.”

• we have 6.1 billion cell phones subscriptions but only 1 billion smartphones

“We know the internet is going to be delivered to these people by a phone. And this puts tremendous pressure on the data center.”

• 31 percent of US adults own a tablet

“In existing markets our propensity to spend more time online with new devices is progressing with dazzling speed. In three years we went from basically nobody to a third of the population with a tablet. Nowadays I can even watch TV with a tablet. That means more time online and more work on the data centers.”

Contemplating what we ought to be building, what technology we ought to invent and what 2021 would look like, Feldman decided to ask some questions:

.

• Can we envision a world with less compute?
• With less data in the cloud?
• With larger less parallelized workloads?
• With graphic being less important?
• With power and space less important?

.

These are the important questions that need to be answered as we think about what servers and which processes to build.

Looking forward back

Andrew Feldman is certain that in order to make better decisions about the future, one must always look carefully at the past.

“One of the things we’ve learned in the history of compute is that smaller, lower cost, higher volume CPUs have always won. There is no exception to this.”

Feldman backed his affirmation with some supporting numbers:

Last year we had more than 8 billion ARM CPUs shipped (compared to about 13 million x86 server CPUs shipped). Also interesting and relevant factors: OEM’s and ODM’s want ARM to win, as well as the largest customers. ARM server CPUs will hit the market in 2014, and this “leads me to believe that ARM will play a monstrous role, as well as ARM CPU, in the building of our future data centers.”

New announcements

“We are announcing today the first 28 nanometer server CPU from an established server vendor, sporting an A-core SOC, 64GIG of DRAM, based on ARM A-57 core,” specified Feldman. “It will be sampling in a few weeks.”

“Also in our booth you can see a development kit: it’s a motherboard, with processor down, assorted I/O, allowing software, development, as well as hardware testing.”

“We’re also announcing today the ARM-based open data center. This is compliant with the Group Hug standard and it’s available for our partner integrators to begin building with immediately.”

Lessons learned and forecasts

“One of the things I’ve learned from the networking industry – where collaboration and community is a bad joke – is that we are racing forward (both in the ARM community and the open compute) with a combination of standards, collaboration and community that is accelerating the changes that are happening – particularly in the ARM ecosystem,” stated Feldman.

“While many complain that this ecosystem is not fully mature for software, I would say that we reached adolescence.”

Andrew Feldman wrapped up his presentation with some bold prognostications:

• By 2019 ARM commands 25 percent of the sever market
• Custom ARM CPUs are the norm for megadata centers
• Smaller more efficient X86 CPUs will be dominant in the x86 segment
• Every major server player will have a fabric based offering
• AMD will be the leader in ARM CPUs, and substantial share of the x86 CPU market
• AMD will be the leader in ARM servers