Out of the many questions I got on the comments section of my "GMMK Pro: The ultimate guide" video, since I posted it in June 15th 2021, one in particular caught my attention:
And after further investigating this topic, I think I found out the culprit for this question: the written review of the GMMK Pro by RTINGS.com
While I don’t review keyboards in my channel from a gamer perspective necessarily (I am a PC gamer, but my keyboard videos are usually more focused on typing experience than gaming features) I decided to take a deeper look into this input latency matter, since I suspect that a lot of my viewers probably use their keyboards for work and school, but also with a bit of gaming on the side.
And just as a side note regarding my own input latency perception when using the GMMK Pro for my day-to-day activities:
Well, I could not perceive any difference whatsoever when comparing it to any of my many keyboards in that regard (I did update the GMMK Pro’s firmware before I started using it, so that could have fixed whatever issues its controller had out of the box, as we’ll discuss later in this article). But it is important to disclaim that I only use my keyboards for typing and when I do game, I exclusively use a Razer Tartarus Pro Gaming Keypad with optical switches that have analog capabilities (which, among other things, means I can manually set how short/fast I want the actuation distance to be).
Now, moving on the issue at hand: keyboard input latency (or lag)...
As you can imagine, a keyboard with high input latency or lag is less than ideal for gamers, since the action would take longer to happen on the screen after a key press when compared to a lower latency keyboard. And for that reason, gaming peripheral companies such as Corsair, Logitech, Razer, Steelseries and others, have been spending a lot of marketing dollars advertising their low latency gaming boards. And, maybe as a direct result of this, people are starting to pay attention to that matter and trying to come up with ingenious ways to measure keyboards and mice input lag.
And here lies our first problem: as it turns out, it is very difficult to measure such a thing in a reliable manner! When we’re talking about input latency for these types of peripherals on a USB connection, we are usually referring to milliseconds that go between:
pressing a key;
closing a circuit on a switch;
having that signal processed by the local controller on the keyboard (or mouse);
having that signal traveling through the cable to the USB port of the computer;
having that signal traveling to the USB controller on the motherboard (which can be on a “north-bridge” chip or inside the CPU itself depending on the PC configuration);
having that signal processed by the CPU;
then having that signal sent by the CPU to the OS residing in RAM (likely Windows, Linux or MacOS in this case);
having the OS handing that signal to the game software also running in RAM;
which will process that signal (through interactions with the OS and the CPU) and decide how to show that “action”;
then having the game software handing that processed signal (or “action”) back to the OS;
which will then send it through the PCI-E bus to the GPU for rendering;
having that rendered image signal sent from the GPU to your monitor through a HDMI or Display Port cable/connection;
and finally having the monitor’s own controller converting the signal to an image frame to display the “action” on its panel.
Easy to see how many variables one would have to account for when trying to test the overall input latency on a gaming system, right?
So, even after measuring the total latency time between key-press/click and its resulting action on the screen, one would then have to consider all these variables before he/she could even start to determine what are the main latency bottlenecks in the system, such as that of a keyboard for example.
While this endeavor is technically possible, it is not a trivial one. And one reason why very few people have invested any time and money to come up with a scientific way to measure this thing is that, for 99% of people, a difference between 10 or 20 milliseconds is not even relevant to their game play (and certainly completely irrelevant for typing), since most of us don’t have fast enough reflexes to take advantage of these tiny millisecond gains. The other 1%, as it turns out, are the people more likely to be successful professional e-sports players…
Even before I became aware of RTINGS.com and their keyboard input latency measurements, I had been following the work of a YouTube channel called Battle(non)sense who has been, so far, the most scientific approach to testing mice and keyboards’ input lag I have seen so far. And boy does he have an elaborate system to measure latency! I strongly advise you to check this video where he shows and explains his system and methodology, which is mostly centered on a particular combination of hardware and software, provided by Nvidia, called LDAT, or “Latency and Display Analysis Tool“.
The hardware part of this combination is of particular importance here, since it provides a more accurate way to measure the action on the screen than using high speed cameras with slow motion capture capability, since the later requires a human to count frames and interpret events such as keypresses and their resulting screen actions. As per Nvidia’s own description of this hardware component:
“LDAT is a discrete hardware analyzer that uses a luminance sensor to quickly and accurately measure the motion-to-photon (click-to-muzzle flash) latency in a game or application”
One particular issue that Battle(non)sense raises in his video is that there are two ways to approach keyboard input latency measurement: straight out the switch pins on the PCB or through a solenoid pressing the key (like your finger would). The first approach eliminates the variability in a given switch travel distance until actuation, which usually varies between 1 and 2.5 millimeters on most mechanical and optical style switches, which would be ideal to measure how fast the keyboard controller is in processing the signal (since you are eliminating the switch variable in this case). The second would give a more realistic (and holistic) view of the overall input latency of a given keyboard, since you’re including all the variables.
And, as you can imagine, this is a VERY IMPORTANT aspect of such measurement, since 1 extra millimeter of travel distance can add at least 1.8 to 2 milliseconds of latency (for an above average player)! So, in his two videos covering this topic (the other is this one where he focusses on Corsair's new keyboards with their 8,000 Hz polling rate) where he measures the latency of different keyboards from EVGA, Corsair, Asus, Cherry and Wooting, he shows these two different measurement techniques and how your choice of switches will have a rather important impact on your final input latency results. He goes on, of course, to cover a ton of other super important (and interesting) points on this subject that you owe it to yourself to watch if you care about this subject.
Then, going back to the RTINGS.com review, I compared their methodology to the one from Battle(non)sense and found out that they don’t use Nvidia’s LDAT (Nvidia did not provide the LDAT kit to that many media outlets and they do not sell it directly on their website). As a result, RTINGS resorted to using the high-speed camera method mentioned above in conjunction with NVIDIA’s Reflex Latency Analyzer tool that is included in its GeForce Experience software. You can check RTINGS.com explanation of their keyboard input latency method here.
While this method has validity, and while RTINGS seems to do the best they can to account for the switches’ travel distances to actuation, I can’t help but being a bit more sceptic of their results here… When checking all the different grades they gave to a ton of keyboards on the specific issue of input latency, you can clearly see that the majority of high grades were, unsurprisingly, given to keyboards that have shorter actuation distances (all targeted at gamers, naturally) with a clear advantage on those with optical switches (also unsurprisingly, since optical circuits are inherently faster than mechanical ones).
There are some glaring exceptions of course, such as the stelar result they had with the Anne Pro 2 for example,
which uses fairly mundane Gateron Brown switches with its rather large 2.5 mm actuation distance. And then there are some real head-scratchers such as their results with the Razer Cynosa V2, a rubber-dome/membrane keyboard, with a much better input latency result than other proper mechanical keyboards…
Which is something a bit difficult to grasp, if you ask me, since membrane keyboards will only actuate when they bottom out, while mechanical keyboards’ switches usually actuate somewhere in between the top and the bottom out positions (which one would think should give the “mechanicals” an edge over rubber-dome/membrane boards).
These results raise a bunch of questions about how well RTINGS is being able to account for switches travel distances in their scores, without even going into other aspects of their measuring system that are not clear in their description, such as if these tests are all made using the same computer for example. It would be reasonable to assume that they probably do. But, if they don’t, then, all of a sudden, we have a bunch of other variables that could significantly affect these results (different cables, CPUs, GPUs, OSs with different patches and applications installed, etc.).
When comparing Battle(non)sense results with those from RTINGS.com for two of the same keyboards they both tested, this is what we have:
Note: numbers are the average results after multiple tests by both outlets
The first thing we see there is a pretty significant difference between these results. Unfortunately, until I find a third media outlet that is testing these same keyboards for input latency, it becomes impossible to “triangulate” these results to have a better idea of testing methodologies and its differences, errors and inconsistencies, as well as to tell whose numbers are closer to real world results.
While I do trust Battle(non)sense measuring system a bit better, my point here is not to throw shade on RTINGS.com testing methodology. Rather, my goal with this article is to show that you should take these very new (and still difficult to audit) input latency measurements with a few pounds of salt! And it is my humble opinion that no one should base a keyboard purchasing decision solely on results like these. One should consider much more important things, such as comfort, noise and typing experience for example, which will make a much bigger difference in your day-to-day use of a keyboard (even for gaming), than these microscopic differences in input latency measurements.
As it relates to the GMMK Pro, we unfortunately have only RTINGS.com numbers to go by and no other results to compare them to, since Battle(non)sense has not tested the GMMK Pro yet, and thus we don’t have a second opinion on this particular keyboard’s input latency situation. So, until more information comes out from more media outlets, with equally good resources at their disposal, as well as similar minimally reliable testing methodologies, I’m not going to point my finger to the GMMK Pro (or any other keyboard for that matter) on the input latency issue.
Having said all that, it is important to point out that the first batches of the GMMK Pro had firmware issues right out of the gate (key chatter and misfires as reported by Brandon Taylor on his review of the keyboard) that were supposed to be fixed by the firmware update you were forced to do when installing the Glorious Core application on Windows. At least in my experience, these issues were completely addressed and fixed by that first firmware update, since I never experienced any of those problems.
I’ve also seen other anecdotal reports online where people claim that the Glorious Core app doesn’t really change the pooling rate of the keyboard when you choose 1,000 Hz instead of 125 Hz or 500 Hz (I’m not sure how they were even able to tell the difference though) and that the only way to ensure that you are indeed getting 1,000 Hz is by setting it through QMK and re-flashing the keyboard completely. While I do not recommend taking this route, it is at least an option for anyone brave enough to dive into the “black magic” world of QMK.
And to close this article, I’d go out on limb here and say that if absolute input latency performance is of any concern to you, then you should probably stick to keyboards brands that actually focus (and invest heavily) on that aspect, such as the products from Corsair, Razer and Steelseries, instead of going with custom or semi-custom keyboards that are more focused on typing experience. These gaming focused companies offer switches with short actuation distance and with optical and “hall effect” circuits that, while not the easiest ones to type on (you’ll make more typing mistakes on switches with super short actuation), are guaranteed to give you lower input latency numbers for gaming, even if the difference happens to be so small to be almost imperceptible to most non-professional gamers out there.
I do believe professional e-sports players use these keyboards because of their gaming performance and not solely because of sponsorship deals. They probably wouldn’t accept to be forced to use a slow keyboard that hurt their gaming performance just for sponsorship money if that meant stop winning championships.
While the mechanical keyboard community loves to scoff at pre-built keyboards from these brands because of their “gamer looks” and rattly stabilizers, I don’t think most professional gamers care about those "typists concerns" when they’re fragging and head-shooting their adversaries for a living… and so shouldn’t you, if absolute gaming performance is your #1 priority.
If for nothing else, this would be another good reason to mod these pre-built boards for better sound or to have more than one keyboard in your collection, right? 😉
Disclaimer: any products linked in this article are affiliated links that will result in a small commission (paid by the retailer in question with no additional cost to you) to this website's owner.
Comments