Targeting Greater Solution Energy Efficiency Through Memory Optimization

Aug 04, 2022

Solution Energy Efficiency

With rising energy costs, customers are looking for solutions that not only perform and add value but can do so at a reasonable price of ownership. As a result, power consumption and solution energy efficiency are increasingly considered.

The Energy Consumption Challenge

As we push our hardware to accomplish more tasks with more data, we must regrettably admit that there's a consequence: increased power consumption. And it's not hard to believe when you consider that data itself is created, moved, and stored in the form of electricity.

In the past, the relationship of power consumption to the size of transistors, known as Kennard's Law, allowed for the continued increase of transistors on a chip without increased power consumption. And at the same time, chip manufacturers were able to double the number of transistors on a chip every two years, a phenomenon known as Moore's Law. Therefore, we enjoyed increased processor and memory performance without increasing power usage for many years.

Unfortunately, due to various factors, both 'laws' have slowed in the past decade. And in response, manufacturers increased CPU performance by stacking chips and improved memory performance through sophisticated caching. Still, with global energy costs on the rise, the threat of increased cost of ownership looms.

NMC - A Long-Term Power Solution

Near Memory Compute (NMC) is a series of initiatives created to allow for processing data in locations (within a server) that are physically closer to where it resides. It's a concept that has been around since the 1960s but until now has been primarily theoretical.

Given today's power consumption challenges, however, it may take steps closer to reality. Intel, in their AI discussion community, presents three key reasons why NMC may become a reality:

  1. Kennard scaling slowdown- With chip manufacturers seeing less and less power consumption economies from more powerful chipsets, the power consumption issue mentioned above is becoming more evident.
  2. Heterogeneous computing - Whether it's CPUs combined with GPUs or accelerators, processors now have the power to solve complex problems more efficiently.
  3. Intel-pioneered memory interconnects (CXL) - Now run by an industry consortium, CXL is a technology that provides improved connectivity between memory layers and error correction.

Current Memory Products that Improve Power Consumption

According to Intel, while the full benefits of NMC have yet to be realized, today's memory and processing innovations will serve as its foundation and are waiting to be leveraged by solution providers. When considering these technologies, it's essential to understand that improvements in the processing and storage of data go hand in hand with the efficient use of power.

For example, with its 3rd Generation Intel Xeon Scalable Processors and family of related products, Intel is offering technologies like:

Intel Optane Memory - Older forms of memory need more power because they provide the same access to all of your data all of the time. As a result, performance was slower, and systems wasted energy on ongoing background processes. Intel's Optane memory has the intelligence to recognize and remember what data you access most frequently and store them in higher-level cache memory. In addition, Optane memory systems channel power and data to the applications used the most, so energy isn't wasted.

Intel Optane SSDs - Unlike older forms of data storage, like HDDs, SSDs provide significant energy savings. That's because they take up less space and require less power and cooling. To meet the needs of core and Edge installations, Intel offers its Optane SSD, which they claim is the fastest data center SSDs available.

Intel Optane Persistent Memory (Pmem) - Intel's Optane Persistent Memory (Pmem) provides the same benefits as other Optane products with added non-volatile failover. So data remains in memory even when a server is powered-down. As a result, it can offer protection against data loss in a power outage and the ability to power down specific systems when not in use.

High Bandwidth Memory (HBM)- High bandwidth memory enables memory to deliver more data to the CPU to enhance performance. As a result, less RAM space is needed as a holding area for data waiting to be processed.

Not only does Intel's Optane family allow for this kind of connectivity, the 3rd Generation Intel Xeon scalable processors can operate in a special HBM mode to allow for more channels of data entering the processor than ever.

To Improve Your Solution's Energy Efficiency, Partner with UNICOM Engineering

As an Intel Technology Provider and Dell Technologies Titanium OEM partner, UNICOM Engineering is ready to maximize solution energy efficiency and performance. Our deep technical expertise can drive your transitions to next-gen platforms and provide the flexibility and agility required to bring your solutions to market. And our global footprint allows your solutions to be built and supported worldwide by a single company. You can understand why leading technology providers trust UNICOM Engineering as their application deployment partner. Schedule a consultation today to learn how UNICOM Engineering can assist your business.

Category: Solutions