You may have noticed that some computer specifications now list RAM speeds in terms of MT/s or “Megatransfers per second” instead of the traditional Megahertz. Why the change? Glad you asked!

How RAM Specs Worked Before

RAM has traditionally been rated in Megahertz, with one MHz representing a million clock cycles. Computer hardware works in a clock cycle represented by a square wave, with a peak and a trough on the wave representing a single cycle. With each cycle, the RAM can complete a memory operation, such as reading or writing. In other words, you can perform a million transfers every second with RAM running at one MHz.

This made it fairly easy to work out how fast RAM was. If you had 133MHz RAM, it was half as fast as 266MHz RAM. However, with the invention of DDR (Double Data Rate) RAM, it became possible to complete a data transfer at both the peak and trough of the wave. Effectively doubling the data rate, hence the name.

So, 133MHz DDR would effectively be just as fast as 266MHz traditional RAM. To make things simple for customers, most RAM manufacturers choose to market their DDR RAM as “effectively” the MHz rating of regular RAM. So you’d see something like DDR4 3200MHz RAM, but the actual clock speed of the RAM is only 1800MHz.

The Problem With MHz as a Performance Measure

So, what’s the big deal? Well, for one thing, it means that the marketed specifications for RAM are technically inaccurate. However, more importantly, Megatransfers are a measure of actual data that can be moved, which means it’s not a one-to-one analogue for Megahertz. The MT/s specification is the theoretical maximum amount of data that can be transferred in a second. While the RAM will always run at its rated frequency, it will not always run at its maximum MT/s rate, because of the variable workloads involved in computing.

There’s also the future of RAM to consider, since QDR (Quad Data Rate) memory can do four operations per clock cycle, two for reading and two for writing. This is even harder to express in MHz, so something that addresses actual data volumes seems to make more sense.

Megatransfers Make More Sense, but It’s Not the Whole Picture

While I think MT/s is a better measure of RAM performance than MHz, it’s not perfect or even the whole picture. For one thing, RAM latency matters too, but if we stick to data transfer performance, there’s more to it as well.

You see, the actual amount of data moved in a megatransfer depends on the width of the memory bus. That is, the number of bits that can be moved per transfer. The main system RAM on modern computers has been 64-bit for years now, so that means you have an apples-to-apples comparison between two RAM modules as long as they have the same bus width.

However, MT/s wouldn’t be so straightforward for GPU VRAM, since GDDR (graphics DDR) uses a much wider bus. Low-end GPUs have 128-bit buses, while 256-, 384-, and 512-bit buses are more typical.

So, it might make even more sense to simply rate RAM in terms of Megabits per second (Mbps) which tells you exactly how much actual data the RAM can move theoretically. Plenty of memory makers already do. If you look at the specification sheet for a Samsung memory module, you’ll notice that it’s rated in Mbps, not MT/s. This does away with both MHz and MT/s, and tells you exactly what the total bandwidth of the memory is.

It remains to be seen whether MT/s or Mbps wins out at the end of the day, but I think it’s clear that either way, MHz ratings for RAM are no longer the best way to express their performance. For a while we’ll be seeing all three, however, so it’s best to know what the difference is!