Posted in

I stopped guessing at RAM capacity and started measuring it, and my next build just…

The topic I stopped guessing at RAM capacity and started measuring it, and my next build just… is currently the subject of lively discussion — readers and analysts are keeping a close eye on developments.

This is taking place in a dynamic environment: companies’ decisions and competitors’ reactions can quickly change the picture.

Assembling the perfect PC is often an exercise in technical know-how, meticulous financial planning, and a fair bit of foresight. Computing has always been a volatile landscape, and the current hardware prices only mean that every component decision carries more weight than it used to. In that environment, the fear of under-speccing becomes twice as acute, and RAM has always been where that anxiety tends to land first.

Both the decisions to overbuy and the fear of under-provisioning are understandable. After all, operating systems haven’t gotten any lighter, PC ports of games are less optimized, and the conventional wisdom of “more is always better” has gone unchallenged long enough that we’re made to think that anything less than 32GB of system memory means that your workflow will crawl instead of walking. The problem is, this instinct has never been more expensive to indulge. Luckily, you don’t have to rely on your best guess while shopping for your next build, because all the data you need to make a decision is already with you.

If you’ve ever opened the Windows Task Manager to diagnose the cause of a stutter while you’re running a few tabs on Chrome and seen “11 GB in use” on a 16GB system, the natural conclusion you’d come to is that an 8GB machine would’ve been a total disaster. That isn’t necessarily the right conclusion, and understanding why can save you a significant amount of money on your next build.

Windows is deeply opportunistic when it comes to system memory. When it sees available RAM, it fills it almost to the brim with cached assets, pre-fetched data, and standby content it thinks you’ll need next. This is usually perceived unfavorably, but it’s reasonable behavior from an OS perspective. The problem with treating this data as reflective of your memory requirement is the fact that it exaggerates your “in use” figure well beyond what your workload demands on a typical workday. A 32GB system showing 24GB occupied and a 16GB system running the same workload at 8GB are often doing the exact same amount of work. The larger headroom just gives Windows the license to cache more speculatively.

The figure you actually want for this diagnosis is the Active Working Set, which is the memory your system cannot function without before it starts paging to your SSD’s NAND. You can’t access this information via the Task Manager at all, as for this, you need your Resource Monitor.

If you type “resmon” into your Start Menu and pop open your Resource Monitor, the “Memory” tab will comprehensively explain your memory needs, provided you know which data set to look for. A horizontal bar at the bottom breaks your physical memory into four color-coded categories, which represent the memory “In Use”, “Hardware Reserved”, “Modified”, “Standby”, and “Free”.

Here, “In Use” represents your Active Working Set, which refers to memory that running processes are actively consuming. This is precisely the figure that reflects your demand. “Modified”, on the other hand, is a small, transient queue of pages changed in memory but not written back to disk. The category that misleads most is “Standby”, which are pages that processes used recently but are no longer actively needed. Windows tends to keep them cached on the off chance they are requested again, but discards them instantly when something else requires the space. “Free” is completely unoccupied, and a healthy system will often show very little of it. The two categories you need for a valid estimation, therefore, are “In Use”, especially when your system is running the applications you intend to run on your new build, and of course, the “Modified” memory. This is a rather straightforward litmus test I like to run before evaluating memory configuration for laptops I am looking to buy.

In between an average work session running all the apps and services that I normally use, the Resource Monitor shows about 14–15 GB of memory “In Use” alongside 15–16 GB sitting in “Standby.” Simply put, on a usual day, nearly half the installed memory is doing nothing essential. The “In Use” figure usually includes a dozen non-essential Chrome tabs, Discord running in the background (that I most likely forgot to close) and other communication apps, and even with all of that running simultaneously, the system’s demand rarely grazes past the 17GB mark.

I first stumbled upon the Task Manager misattribution by accident, when one of the DDR4 sticks on my home workstation failed, dropping me from 16GB to 8GB. I expected the system to struggle and crawl, but instead, all my applications (and even games) ran almost identically to how they did at full capacity. The reason for this is the fact that my actual working set had never been close to 16GB to begin with, and Windows had simply been filling the available headroom with processes it deemed necessary. Needless to say, I didn’t feel the need to replace the faulty DDR4 DIMM till date, and my workflow hasn’t exactly suffered because of it.

This almost goes without saying, but your workflow will always dictate your memory requirements. A video editor working with 4K feeds and timelines and a gamer running Twitch alongside a AAA title will always land on different numbers. The variance, however, is narrower than what conventional wisdom suggests. The difference between what your system and your workflow demand versus whatever Windows Task Manager suggests is always a touch wider than you’d expect. As such, it’s always a good idea to measure first and buy second, especially given the current RAM economy.