Science Fiction Coming to Life? New 4-Socket Servers drive AI and ML closer to reality!

Why 112 cores in a single server? Because, as HAL 9000 would say: “This mission is too important for me to allow you to jeopardize it!” If you thought you would never need such calculation power, now is the time to change your mind and tame the data avalanche!

It’s what CIOs prefer! Dual socket servers are the world’s leading sellers at the moment, and demand keeps rising, according to data published by IDC. “Dual socket infrastructures are useful for datacentre deployments, both in terms of power consumption and cost per server”, explains Lloyd Cohen, the analyst in charge of the server market at IDC.

If most servers have two processors, it’s because they suffice to effectively run most applications, but with new workloads come new infrastructures. With ever greater data volumes to be managed at enterprises, their operations are ever more value related and their processing more and more demanding on resources.

“A large part of what’s driving the digital transformation revolution is machine learning and data analytics”, says Adam Shepherd (ITPro). “There is not one single industry around the world that is not interested in artificial intelligence and machine learning”, confirms Ravi Pendekanti, Senior Vice President, Server Solutions Product Management & Marketing, Dell EMC.

Databases and/or Data Lakes are therefore at the heart of this revolution. So, to speed up the access to the data it contains, your server infrastructure needs to cross a threshold. This is the gist of the latest four-socket server announcements made recently at Dell Technologies World.

More data for smarter working! – HAL 9000: “I know I've made some very poor decisions recently, but I can give you my complete assurance that my work will be back to normal.”

The new PowerEdge R840 and R940xa systems are specifically designed for these new data-intensive uses. They stand out thanks to their four processor sockets. Naturally, this is not the first time that Dell EMC proposes such infrastructure systems. However, thanks to the latest generation Intel Xeon Scalable chips, these four slot systems will now support up to 112 cores. To assess the performance gains, Principled Technologies compared the number of Monte Carlo Simulations run per minute by the new PowerEdge R840 and by its 12G generation predecessor, the R820. Simulations serve to predict the future value of an investment portfolio. Based on the results of the analysis, not only was the R840 3.5 times faster, but it also reduced statistical errors by 54%.

There is a simple reason for this. The more simulations the algorithm makes, the more accurate the results. This enhanced reliability will find its full value in a machine learning or artificial intelligence scenario. The more and faster your infrastructure can absorb data, the smarter the software platform will be to extract relevant data for your business. The processor itself is not the only component playing a part in this task.

More than a processor, a precious accelerator! – HAL 9000: “I am putting myself to the fullest possible use, which is all I think that any conscious entity can ever hope to do.”

To accelerate processing, IT managers also draw on two major advantages of state of the art servers. Firstly, the GPU (Graphics Processing Unit). Thanks to its parallel processing capacity, the GPU is especially suited to AI. With four sockets, a 1:1 ratio is achieved: four CPUs for four GPUs, thereby considerably speeding up exchanges between components. This is what the R940xa – that stands for for Xtrem Acceleration - proposes.

The second hardware aspect you can count on to speed up the process is FPGAs (Field Programmable Gate Arrays). Here too, programmable circuits work in tandem with the CPU to off-load processing that can be performed in parallel and speed request execution - check out more details in the article by Robert Hormuth.

In conclusion, according to a study by Forrester Consulting, 61% of IT decision makers feel that the lack of dedicated processors, like GPUs and FPGAs, are one of the major hardware challenges facing artificial intelligence. This is a challenge that is now being met … HAL 9000 would be proud!

Let's keep in touch ... feel free to comment my posts on LinkedIn and follow me on Twitter!