This post is co-authored by Carlos Morales, Senior Director of Deep Learning Systems, AI Products Group, Intel
The new Dell Ready Solutions for AI – Deep Learning with Intel accelerates AI insights, optimizes TCO and offers a fast on-ramp for deep learning workloads.
In a quest to bring the promise of artificial intelligence to life and capitalize on the massive amounts of data generated on a 24×7 basis, many organizations are rushing to pull together the different technology elements needed to power deep learning workloads. Today, this quest just got a lot easier with the launch of a new Ready Solutions for AI based on Dell and Intel innovation.
The Deep Learning with Intel solution joins the growing portfolio of Dell Ready Solutions for AI and was unveiled today at International Super Computing in Frankfurt. This integrated hardware and software solution is powered by Dell PowerEdge servers, Dell PowerSwitch networking, and scale-out Isilon NAS storage and leverages the newest AI capabilities of Intel’s 2nd Generation Intel® Xeon® Scalable processor microarchitecture, Nauta open source software and includes enterprise support. The solution empowers organizations to deliver on the combined needs of their data science and IT teams and leverages deep learning to fuel their competitiveness.
Dell Technologies Consulting Services help customers implement and operationalize Ready Solution technologies and AI libraries, and scale their data engineering and data science capabilities. Once deployed, ProSupport experts provide comprehensive hardware and collaborative software support to help ensure optimal system performance and minimize downtime. Additionally, Education Services offers courses and certifications on data science, advanced analytics and more.
AI simplified
The new Deep Learning with Intel solution simplifies the path to AI-powered applications with the fully featured container-based Nauta deep learning platform which offers an innovative template pack approach that eliminates the need for data scientists to learn the intricacies of Kubernetes. In addition, Dell’s data scientists have built use case examples for image recognition, natural language processing and recommendation engines to help customers understand the capabilities of the solution’s architecture.
Deep Learning with Intel is also pre-configured with the TensorFlow distributed deep learning framework, Horovod and all the requisite libraries for data modeling. This simplified path to productivity is both easy to setup and easy to use and empowers your data scientists to spend their time building models that generate value instead of wrangling with IT infrastructure.
Faster, deeper AI insights
Once the Deep Learning with Intel solution is up and running, you’re positioned to accelerate model training and testing with the power of 2nd Gen Intel® Xeon® scalable processors. This next-generation processor is at the heart of Dell PowerEdge C6420 servers used in the Deep Learning with Intel solution, and together with the newest software optimizations for TensorFlow and supporting libraries, model training time is greatly reduced. The processor also includes new Vector Neural Network Instructions (VNNI) that radically speeds up deep learning inference workloads with more efficient 8-bit integer data formats and instructions to power through four times as much data as was possible with 32-bit single precision floating point methods.[1]
The solution is integrated with the multi-user open-source Nauta software platform that enables containerized training workloads which ran up to 18% faster than the same workloads on a bare metal system. As your organization needs grow, the Deep Learning with Intel solution enables near-linear scaling, achieving 80% of theoretical maximum performance when the number of compute nodes is scaled from one to 16.[2]
Enhanced TCO
Finally, the new Ready Solutions for AI is great from a total cost of ownership perspective, especially when compared to cloud and hardware-accelerated solutions. For deep learning training workloads as an example, the three-year TCO is 24% less on Deep Learning for Intel relative to a leading public cloud service, while providing double the compute time (24 hours per day versus 12 hours per day), and ten times the storage capacity (100TB versus 10TB).[3]
Public cloud AI service costs can vary widely, and monthly charges can be surprisingly high when inadvertent mistakes lead to runaway processes that consume excessive CPU time or generate massive volumes of data. On the other hand, the Dell Deep Learning with Intel on-premises solution provides managers and financial accountants with known and predictable expenses, while enabling your organization to drive standardization and control over your infrastructure and data.
Key takeaways
The new Dell Ready Solutions for AI – Deep Learning with Intel is an ideal choice for organizations looking to leverage container-based environments to run both single node and distributed deep learning training and inferencing workloads. It simplifies the path to productivity for data science teams and IT and delivers better-than-bare-metal performance. And like all Dell Ready Solutions, this solution is based on a linearly scalable building‑block approach so your deep learning environment can grow to meet your changing needs as time goes on.
Here’s the bottom line: With the included software, servers, networking, storage and services, all optimized for AI workloads, the Deep Learning with Intel solution gives you just about everything you need for an AI-powered organization. Just add data and stir.
To learn more
For a closer look at the capabilities of the Ready Solutions for AI – Deep Learning with Intel, visit Dell.com/ReadyForAI.
[1] Intel news release, “Intel Shows Breadth of Data-Centric Platform with Cascade Lake Advanced Performance and Xeon E-2100,” November 4, 2018.
[2] Enterprise Strategy Group Technical Validation Report, Dell Ready Solutions for AI: Deep Learning with Intel, April 2019
[3] Enterprise Strategy Group Technical Validation Report, Dell Ready Solutions for AI: Deep Learning with Intel, April 2019
