TLDR:
The issue is that if I set my fps_max before loading the map, the demo automatically uncaps it once, and then seemingly again after the demo starts. This can be remedied by simply beginning my CapFrame recording later at a less precise moment. I don't know what solution to propose, but maybe something like just a simple start screen with automatic behavior in the benchmark to set it to fps_max 0, and don't override it again in the middle of the demo or halfway through it etc.

DR:
I was finding myself surprised recently at noticing a remarkable difference in the behavior of the game, and stutters when switching between fps_max 0, and fps_max 400. Also I am considering how this might change or be affected by past or future updates. I was running some tests locally in a warmup map, and wanted to use CapFrameX to study the different behavior of values for fps_max.

To my memory, I recall that fps_max used to behave differently, it simply capped frames if the card would be able to push past, and then drop below in "stressful" areas. Please feel free to correct me if I'm wrong here, or just perhaps remembering how CS:GO worked? My card is a 4090 and when set fps_max 0, it gets average frames of 500 on Ancient, swinging as crazy high as 650, and low as 450. When using fps_max 400 or 300, however, it seems to always undulate below the cap, sitting at much lower averages of 200-300 and never gets even close to the cap. Not until I go lower towards my actual refresh rate - do I start to see it behave in a "capped" fashion where it holds at 144 with almost no variance. But ironically - the game is the absolute WORST to play with this, due to no VSYNC in my monitor, and horrible tearing or just a gross feel to it. This persists even at fps_max 200.

I started by trying to test different resolutions and fps_max caps by simply walking through Grail or Ancient, but was finding that the graphs and results I was getting was probably too varied based on my inability to mechanically move. Also scripting something for movement through the map seems to be outside of my knowledge and ability. Which is what led me to discover this nice workshop map, which is amazing, but just seems to not fit me purpose.

I realize that the community probably will overwhelmingly ask "why the f would you cap your FPS" since this has always been an almost unanimous opinion and pro opinion, but I wanted to investigate independently and at least confirm ( and also it might just be the fact the different systems and graphics cards vary on this radically as well ). Even if it's fruitless entirely, I'm still having fun playing around and testing it, and I might also use a similar method to test different BIOS, CPU, and GPU settings tweaks in the future to see if I am being bottlenecked or have some odd things going on with my rig.

Also - I did try using a replay demo, but these apparently are not the same as running natively in client or whatever. The FPS is just radically different and not representative of gameplay. I realize this demo isn't either, lacking the FPS view etc and all the networking ♥♥♥♥, but it would be much more consistent than anything I can put together. I'll look around for more demo maps, but there are only a few from what it appears.

Yes and I'm not just here to brag about the 4090 but it's a great card in my opinion except for the ♥♥♥♥♥♥ power connector, a perfect upgrade from my previous 1080TI which was the effing best card and still in use in my sloppy rig I let other people use. It was my last big birthday purchase to myself with my new rig before I unexpectedly hit some much harder times, so be gentle with me.