DRAM the essential choice
16Gbit server, desktop and notebook memory modules over an illustration of traces and chips of a circuit board
There are a lot of similarities between the developments of memory and processors in the industry. The wafers producing the chips are getting smaller and smaller, allowing more chips on every wafer. More chips on the wafer allow for more capacity, which will be used either for more storage, or faster throughput. Or both.
The new standard will be DDR5. The capacity will be double that of DDR4. We are moving from 8 to 16Gbit chip density. In a laptop with two slots with Unbuffered SODIMM you were limited to 32GB, now you are going to get 64GB.
This is a natural evolution that happens every time we experience a technology leap forward in the memory industry. It will be good news. Making servers faster; by allowing them to have a higher performance. It will also allow for better performance laptops for use in hybrid working environments post COVID-19. Plus, it will help support the collaborative tools like Zoom and Teams that monopolise resources in your system. Having more memory is essential.
16Gbit transition is a good time to take stock
This transition from 8 to 16Gbit is a good time to question whether you should be migrating or updating the hardware.
The main challenge, according to Adrien Viaud, Senior Technology Manager at Kingston Technology, is compatibility. “Your capacity to upgrade in the future is compromised if you are using Intel Skylake and older systems since they are not compatible with the new generation 16Gbit DRAM components.” he explains. “16Gbit DRAM allows you to double the capacity per memory module, however, a system compatibility check is required. Kingston’s team can help research whether your system is compatible”.
As Ferdi van der Zwaag, Team Leader Benelux and Business Development Manager EMEA puts it “Our task is to guide the customer through the process to buy the right kind of memory. It is more complex than just installing the latest memory. For every customer we need to determine the right combination of the appropriate capacity combined with their requirements for speed”.
Key parameters to consider
In terms of usage, it all depends on the applications that will run on the server. Essentially based on three important parameters: The performance that the application requires, the memory throughput capacity and the power usage.
Viaud uses this analogy. “A highway has different lanes, so first you look at the architecture. For instance, Intel Cascade Lake based platforms feature a 6 memory channel architecture (6 lanes) and AMD Rome offers 8 (8 lanes). Think of your data as cars. To get the best throughput you want one car per lane. If you put all your cars in the same lane, you will create a traffic jam which will be detrimental to the memory performance of the system.” That is the trade-off that will have to be balanced optimally: throughput or capacity. The type of processor used on the server also dictates the memory speed. The highest end processor will give you the best performance, if you choose a less powerful processor the memory will clock down because the processor cannot handle the speed of the memory.
Older generations may not be compatible. When a transition like this is happening, you may have systems that will lose support. You might need to stock or find parts that you could use for a longer period. It is important to think of the future and of scalability: what if you want to add memory in a year? If you have filled your server with low-capacity modules you will be stuck and you will have to remove the preinstalled modules and replace them with new modules.
As we move down in terms of lithography, we start seeing benefits in terms of power consumption. Because the capacity of each module is much larger, the energy efficiency increases accordingly. Filling half the slots of the server with higher capacity modules will give a better throughput and will be more power efficient as filling all the slots would. Another advantage, of course, is it will still be possible to add more capacity later, as there will still be available slots. You have options in the bios, power settings, performance etc. to configure it all.
There are different memory technologies available, UDIMM, RDIMM and LRDIMM however they cannot be mixed. The system will fail to boot. Therefore, it is important to know what memory technology is installed prior to upgrading.
The best memory partner for your solution
“There is no right or wrong, we need to help customers make the right choices for their system. I have seen mistakes made” adds Ferdi van der Zwaag “Kingston is like the satnav that will help you navigate through this landscape”.
Kingston’s advice is to start with the solutions pages and then use the “Ask an Expert” service to tell the team what you want to do. You will get advice on what the best technology for your application is and what configuration to select, which capacity, what throughput, what speed and what power consumption you can expect from the setup.
We invite you to listen to the podcast with Simon Besteman, Adrien Viaud, and Ferdi van der Zwaag.
Listen to podcast
#KingstonCognate introduces Simon Besteman
Simon Besteman is a French and Dutch national and a veteran of the ICT and data center industry. He has over 20 years of experience working at ISPs, suppliers, and data centers and as a management consultant for a wide range of organisations. Simon has held various senior management positions in multinational companies in the areas of service, sales and marketing, operations management and strategic development.
Currently, Simon is the CEO of ISPConnect, the Dutch coalition of hosting providers. As a leading representative of the industry he is a frequent blogger on industry and policy matters, a keynote speaker at congresses and conferences, and a participant at Dutch government round tables on telecommunication, data center and internet regulatory matters. He sits on the boards of various industry groups, with a focus on education, employment and governance.
