Many experts in our embedded industry will try to convince you that a “by-8” (x8) DRAM module is superior to a “by-16” (x16) module. In some respects, those experts would be right. But it’s not always the case, as we will show you here. In most configurations, the x8 module will yield higher bandwidths compared to the x16 configuration.
However, there are many other characteristics that must be considered in your memory selection, particularly depending on your intended application and its specific needs. In fact, we are going to show you that the advantages of moving to x16 may outweigh that performance loss. Most notably is a cost reduction of about 30%. If you’re designing for such applications as kiosks, automated teller machines (ATMs), or arcade gaming platforms, the performance received from the x16 DRAMs should suffice.
First, being completely transparent, let’s explain why there’s a difference in performance. Very simply, the number of “bank groups” within the x16 chip is half of what’s found inside the x8 IC. The purpose of that bank group is to increase I/O bandwidth without increasing the DRAM’s internal bus-width. Hence, fewer bank groups, fewer paths to increase performance.
While I can’t cite any “official” comparison documentation, we have been able to procure some testing data. In three separate tests, the performance of the x16 equaled about 90% of the x8 modules. These tests were performed with readily-available off-the-shelf memory.
Now let’s looks at the cost. To be clear, not all x16 memory ICs are cheaper than x8 devices, but there are many examples where that is the case with a cost reduction of about 30%.
One of the bigger advantages from the designer’s perspective is that no changes are required on the board to migrate from x8 to x16 memory. The only consideration that must be made is to perform sample testing, which ensures that the DRAM module is completely compatible with the CPU and motherboard. Seasoned engineers should be familiar with the sample testing process, which is a common practice, and is typically performed routinely, regardless of the memory choices that are made.
The majority of embedded applications will not be impacted by the 10% performance degradation. Places where you may want to consider the x8 models include those which require the ultimate in performance, like 5G applications, deep learning, artificial intelligence (AI) and machine learning, Edge computing, smart medical applications, and higher-performance computing (HPC).
One real-world example where x16 memory is more than adequate is in an arcade gaming platform that was deployed in Japan. Here, cost was high on the priority list, as was the promise of stable performance over a long time period. Innodisk provided its DDR4 2400 4-Gbyte SODIMMs, with the result being 100% tested and optimized performance. That particular module is certified to CE, FCC, and RoHS standards.
Now that we’ve convinced you there are instances where x16 memory is a better option, we want to introduce you to Innodisk, a provider of industrial embedded flash and DRAM products and technologies. The company prides itself of being a service-driven provider, supplying customers with exactly the storage products that fit their needs, in terms of cost and performance.
For further technical information or inquiries, please contact us!