JEDEC, the consortium of manufacturers of both volatile and non-volatile memory, i.e. RAM and Flash, has just presented a new type of memory module which they have dubbed MRDIMM and that it brings with it a feature that could change the configuration of future PC processors and motherboards. What special features does it have and how will it influence future Intel and AMD CPUs?
PC RAM is generally more about latency than bandwidth. This is due to the common nature of central processing units, or CPUs, which makes their performance vulnerable to latency, as they have no way to mask it. Instead, a GPU or graphics chip does it. The problem arises when we are talking about bandwidth-dependent applications that are limited in this regard. We’re talking about integrated graphics, RAM currently being a bottleneck preventing further scaling, and deep learning and machine learning applications that are becoming increasingly popular.
What is special about MRDIMMs?
It’s very simple, the fact that it allows you to combine the bandwidth of two memory modules. Currently, we have two types of RAM memory modules for our PC: DIMM or SO-DIMM for our PCs. However, MRDIMM would not be a new standard, but rather would seek to solve the bandwidth problem.
How would that work? Suppose we have two DDR5-4800 modules, which means 4.8 Gbps bandwidth on a 128-bit bus with the CPU. The problem is that each module contains a 64-bit data bus, but if we have four slots occupied, only 50% of all possible bandwidth is used due to not enough pins for CPU data.
Well, at this year’s Memcom, JEDEC presented their proposal for a new standard module which they named MRDIMM. The concept is simple: place something as basic as a multiplexer between two DDR memory modules so that they combine their bandwidths and become a QDR Module. In other words, a 128-bit bus at 4.8 Gbit/s would become a 64-bit bus at 9.6 Gbit/s. The bandwidth stays the same, but we’re fetching an extra 64 data pins and that allows us to double the bandwidth.
In the example, they show 8.8 Gbps per second memory achieved by combining the bandwidths of two 4.4 Gbps DDR5s. His projects? To be able to reach 17.6 Gbps in a few years. That is to say double the current memories and all without having to launch a new generation of memory in the form of DDR6 in the middle.
This will mean changes inside the processors
In any case, this also means changes to the processor’s integrated memory controller, whether Intel or AMD. In any case, we are more than sure that these changes are known to both manufacturers, who already have their future chips and motherboards prepared for these new memory modules. The harm of this solution? A reduction in the number of available channels and therefore higher contention, as well as higher memory access latency.
In any case, it will be a great help in mitigating the current problem of integrated GPUs. Who seems to have remained in no man’s land. The reason is simple: lack of RAM bandwidth means they can’t scale in performance. Those who assume that adding more cores to them or increasing their clock speed even when paired with DDR5 memory is futile, as these improvements are held back by the lack of transfer speeds.
Although MRDIMM memory does not solve the contention problem, it does provide enough bandwidth for future integrated GPU designs to scale beyond current capabilities. Who knows whether or not this will help APUs gobble up the lower range of graphics cards. Which already has signs of abandonment by NVIDIA and AMD due to rising chip manufacturing costs and the low margins they provide.
