Are you trying to test at an increased speed, like +1 or +10? Or do you want to run it at +0 speed and see whether it slows down at all?
You could create a short replay file by starting a custom game with cheats on, spawning in lots of units (e.g. 100 ASF per player) and letting the fight. Obviously a "natural" game wouldn't have enough going on in the first two minutes to make a good benchmark, and you don't want a benchmark that needs 10+ minutes to run.
You could also create a SIM mod to help with running the replay file. Let's say the mod does the following: it records the system time at the start of the match, then ever time the game time reaches a multiple of 60 seconds, it reports the amount of real time since the start of the replay file, in milliseconds, by sending a chat message. For example, if your replay is 2 minutes and 4 seconds long of game time, the mod would report twice, once at 60 seconds and once at 120 seconds of game time.
If the system runs without stalling at all, it should only take 60,000 and 120,000ms to reach those milestones. But if you get a score higher than 120k, that shows how much lag the particular system has.
Or to make things simple, it could report the time elapsed minus the amount of game time. E.g. if it took 125 seconds to reach the 120-second milestone, it would report 5,000 ms of delay. Then you don't even have to do math of substacting out 60k or 120k, you would just get the exact amount of lag.
Such a mod could be made compatible with the Steam version or the FAF version. Probably there would be very little difference between a mod that did that for Steam and a mod that did that for FAF.
So you could run the benchmark only by playing the replay file with that particular mod active. And the end result would be a numerical score showing how much the game lags.
That would also allow you to make different replay files depending on how "challenging" you want the benchmark to be. Of course you could use a time different than 2 minutes but if you want the benchmark to be fast you should focus on making a replay file that is resource-intensive so you only have to run it for about 2 minutes to get a good idea of the score.
Since this benchmark would require the game to actually run, you would also be testing the graphics card. I have no idea whether the built in test benchmark uses the graphics card at all.
This mod would be very easy to make i think. I might even be capable of it.