17/9/2024 Marvel vs Capcom 2 Input Lag Results
Let's try something different. I have traditionally posted all my results on Twitter, however have been trying to update my website, and thought I might take the opportunity to expand my reporting a bit. This may or may not persist.
Marvel vs Capcom 2 has been long awaited, and is definitely one of the fan favourite games. I took the opportunity to look at input lag across a host of different platforms and settings.
Firstly using video interrupt method, PS4onPS5 the results are 57ms, so about 3.5F. This was tested with normal and turbo settings, so presumably the turbo setting increases the game speed via frameskip. This also looked at the command history and the animation (Ryu LP) which were the same. To note, when I looked at the Dreamcast version a long time ago, it came out at 3.5F, so this one seems to be equivalent.
Next we had a look at the Switch version via video interrupt. The result was 68.75ms, so a touch over 4F. This is pretty good. I didn't get the chance to test off a real screen (photo transistor method) as the switch didn't seem to like my HDMI cable, but I presume it will be something similar to this.
Last we checked out the Steam Deck version. Through video interrupt we got a result of 43.89ms, so about 2.5F. This is quite fast, and apparently faster than the Dreamcast version, which I had tested before.
As always, PC doesn't play nicely with video interrupt, so I moved on to the photo transistor method on the Inzone M3.
Firstly, I retested the PS4onPS5 version. The result was ~66ms across all different video output settings (VRR off, auto, auto apply, 120hz off, automatic). This is not unexpected, as the Inzone M3 does tend to draw the screen ~half a frame (8.33ms) after the video interrupt result.
Next I had a look at the Steamdeck. I was able to disable frame limit and set the output to 240hz. I wasn't able to enable VRR, but could allow tearing (although it never seemed to happen). In any event, the result was 20ms, so just over a frame of input lag!. Enabling vsync in game seemed to lead to about 3ms of additional input lag (and I didn't notice any tearing occurring when VSync was disabled.)
Finally moving on to the PC, I went through a large volume of tests. Stating with the monitor at 240hz, GSync on, input lag reduciton at ultra and favouring performance. For the first test I went with the default in game settings of borderless window with VSync off. This gave results of ~19ms! Repeating the test gave results of 15ms! There does seem to be some fluctuation in the results, between about 15-20ms, which means that precise comparisons between modes can be somewhat difficult. In any event, turning on vsync in game didn't seem to add any additional input lag. Switching to full screen with VSync off led to higher results of 28ms, but with screen tearing occurring (so that the bottom of the screen was calculated to be 19ms). Turning on VSync in game had a large effect, removing tearing but increasing input lag to 51ms. Windowed mode returned the results to a more successful 16ms with in game VSync on or off.
Responding to some comments from eventhubs, I went back and retested with GSync off, fixed refresh engaged. All other settings unless I mention them were the same. For borderless I got results of ~16ms for borderless window with VSync off in game. For fixed refresh and fullscreen with vsync off I got 27ms, with tearing, and with vsync on I got 46ms, so again a large jump.
Finally I tested the effect of dropping the monitor refresh rate. I left other settings as gsync on, in game vysnc off, borderless window as that had provided the previously best results. Running a quick test at 240hz with a result of 16ms, I then dropped the speed to 120hz and got a result of 18ms, so not a large effect. Dropping the screen refresh rate to 60hz brought the input lag all the way down to 44ms. So it would seem that so long as your monitor can achieve at least 120hz, you should see a large input lag reduction playing on a reasonably powerful PC.
I also took the chance to have a look at the effect of changing filters, however having compared no filter, filter a, filter b and filter h and found no difference, didn't go through the rest of them.
Given the chance I would like to have a look at at least one of the other games in the collection, probably Marvel vs Capcom. Given the systems are different (Naomi vs CPS2) I would be curious if the differences between formats and input lag is the same.