When powering your latest Machine Learning (ML) or Artificial Intelligence (AI) solution, hardware can make all the difference. But how does one fairly and accurately compare platforms? The MLperf benchmarks provide the level playing field you need, and for this latest v.20 round of comparison, servers from Dell Technologies dominated in several categories.
The MLPerf benchmarks: what they mean
The MLPerf is a suite of performance measurements established by the MLCommons (https://mlcommons.org/en/), an industry organization set up to drive progress in machine learning via benchmarking. Their goal is to democratize the field of machine learning and promote its use industry-wide.
Before diving into the categories in which Dell EMC Technologies' platforms outperformed the competition, it's helpful to review the use case that the MLperf benchmarks support. That way, you can match them to your specific solution's needs.
Image classification- refers to an AI's ability to recognize and interpret the meaning of an image. For example, an application that enforces traffic laws may be coded to identify license plates and send citations to speeders.
Object detection- is used by an AI to find and classify objects within an image. For example, with this ability, an AI might be able to identify specific items at a cash register and total them for shoppers.
Speech-to-text- enables applications to convert a user's voice into text. The dictation app available on most smartphones is a basic example.
Natural language processing- refers to a system's ability to convert spoken words into actions. Therefore, instead of merely recognizing speech, an application using natural language processing can, for example, convert text from English to Spanish.
Recommendation Engines- analyze user data to make product or service recommendations. A classic example would be Amazon's suggestion feature that presents additional products based on user behavior and buying patterns.
Dell EMC PowerEdge servers - MLPerf V2.0 results
The latest MLPerf v2.0 results were recently published, and Dell Technologies' servers finished first in the following categories:
Performance per accelerator with NVIDIA A100 GPUs - A popular choice for Deep Learning Systems, NVIDIA A100 GPUs were used in Dell EMC PowerEdge XE8545 and PowerEdge R750xa servers.
Use cases: image classification, objection detection, speech-to-text, medical imaging, natural language processing, and recommendation engines
System performance among PCIe-based 4-GPU servers - The Dell EMC PowerEdge R750xa was the ultimate winner.
Use cases: image classification, object detection, speech-to-text, natural language processing, recommendation engines
Lowest multi-stream latency with MIG instance in Edge - In the multi-instance GPU (MIG) systems category, the Dell EMC PowerEdge XE8545 bested the competition.
Use Cases: image classification, object detection
T4 inference results - This category looked at systems that used NVIDIA T4 GPUs. The Dell EMC PowerEdge XE2420 finished first.
Use cases: image classification, speech-to-text, recommendation engines
Highest performance per watt with NVIDIA A2 GPU - Among systems paired with an NVIDIA A2 GP, the Dell EMC PowerEdge XR12 won.
Use Cases: image classification, object detection, speech-to-text, natural language processing, recommendation engines
Dell Technologies offerings for ML/AI solutions
As the need for core and Edge-based AI/ML applications has grown, so have the platforms that Dell Technologies builds to support them. Here's a more in-depth look at the systems tested for MLperf mentioned above:
Dell EMC PowerEdge XR12 - The PowerEdge XR12 is a 2U server that is MIL-STD and NEBS Level 3 compliant and designed for use in telecom, military, retail, and back-office locations. They are ruggedized to withstand dust, shock, and extreme temperatures and support up to 36 x86 cores running 3rd Generation Intel Xeon Scalable processors.
Dell EMC PowerEdge R750xa - The PowerEdge R750xa is a dual-socket/2U server that supports eight channels per CPU, PCI Gen 4, and up to 8 NVMe drives. It offers the performance of a 3rd Generation Intel Xeon Scalable processor and is also considered the optimal solution for GPU-based workloads.
Dell EMC PowerEdge XE8545 - This server supports Two 3rd Generation AMD EPYCTM processors with up to 64 cores per processor in a 4U configuration optimized for AI performance. It also offers 6-7x the machine learning performance over current accelerators and supports PCIe gen 4.0.
Dell EMC PowerEdge XE2420 - This platform is another server offering designed to perform at the Edge. It offers temperature resistance in a small form factor, driven by two 2nd Generation Intel Xeon Scalable processors. This is recommended in telco, manufacturing, and retail Edge locations.
Give your AI/ML solution the power it needs
When building and launching your next AI or ML-based solution, don't forget the difference that hardware can make — especially when deploying at the Edge. With the right platforms in place, and your application performing at its best, you're free to innovate and gain an edge over the competition.
UNICOM Engineering is proud to be a Dell Technologies Titanium OEM Partnerand Intel Technology Provider. Our award-winning team is ready to enhance your AI solutions for optimal performance and reliability. Learn how UNICOM Engineering helps our customers bring their applications to market faster with hardware solutions powered by industry-leading components from Dell Technologies and Intel. Schedule a consultation today to learn how UNICOM Engineering can assist your business.