Input lag is very important in competitive gaming. Every millisecond counts when you are competing at the highest level. A few milliseconds can be the difference between killing or being killed in any first person shooter and this is especially true Counter-Strike: Global Offensive. I performed a variety of input lag tests for CS:GO to find out which settings affect input lag and what to use for the lowest input lag.
I used a Nikon 1 J1 high speed camera recording at 1200 frames per second to find the time between the button press and the monitor displayed the gun firing in CS:GO. I used a method similar to the one Mark Rejhon came up with for the Blur Busters G-SYNC input lag test. I used an Asus VG248QE monitor with the GSYNC upgrade kit for all of my tests.
The Corsair K95 keyboard has a mode that turns a key's LED on the instant the key is activated. I positioned my keyboard next to the monitor and recorded them with the high speed camera. I used the ffdshow OSD with Media Player Classic to display the frame numbers and recorded them in a spreadsheet to do my calculations. To get the input lag time I counted the frames from when the key was pressed until the monitor displayed the gun firing. Since the camera records at 1200 fps each frame has a timespan of .833 milliseconds.
I ran all the tests in an empty bot match on Aztec as a terrorist. I went to the same corner and recorded as I fired the entire clip of a glock. I recorded each shot as a separate video to make things easier.
I ran two tests comparing multicore on and off. The first test I did was with max_fps set to 1000. With multicore rendering on I hovered around 800 fps and with multicore off I averaged around 400. With multicore enabled the input lag ranged from 5 ms to 21.7 ms with an average of 12.3 ms. Multicore disabled ranged from 1.7 ms to 21.7 ms with an average of 12.1 ms. The difference at such high frame rates is fairly negligible, but when frame rates are higher you would excepct input lag to decrease, which did not happen.
I lowered max_fps from 1000 to 144 for the second test. When the frame rates are equal it becomes very apparent that multicore rendering adds input lag. With multicore enabled the input lag ranged from 5 ms to 26.7 ms with an average of 16.4 ms. Multicore disabled ranged from 3.3 ms to 23.3 ms with an average of 12.8 ms. It is still less than a 4 ms of additional input lag, but it is a significant percentage and will be higher at lower frame rates.
If you are limiting the frame rate and have enough power to reach that limit only using one core you should definitely disable multicore rendering. If your CPU is bottlenecking your frame rate you may want to enable multicore rendering to get a higher frame rate, but if you are GPU bottlenecked you should leave it off.
Instead of every core being combined to render a single frame faster each core produces its own frame. Since the processing all happens at once it can't grab input that hasn't happened yet. The extra frames rendered from the extra cores are based on interpolation from previous data instead of new data.
Spreadsheet Videos