Are you testing this out on the first level of the game by any chance? There are known roping issues (When ropes swing and move) that cause major lag on any system. If this is the case, grit your teeth and get throught he first level, and wait until you are in the jungle-city. This will be a more accurate depiction of your performance.
Hmm. Very strange. Have you tried disabling Vsync at all? Do you have any form of AA set for Crysis 3? if you do dont use SMAA or anything like that, try disabling it completly to begin with to see what it's like. AA is a real frame rate drainer.
Right. Ok, we may have found your problem. You are saying, if I read the above statement right, that when palying Crysis 3 you are using around 3.5-4gb of your system RAM, and between 800mb-1024mb graphics RAM. This would meant that you are actually maxing out both your system and graphics memory. The fluctuations in FPS could be caused by one program requireing less resources, or you panning the camera on the game to an area with less activity.
Unfortunately, if this is the case you will need to upgrade your system RAM to around 8gb. OR you have to lower in game settings. You are simply using too much RAM and VRAM for the GPU to render any more frames. I believe this is your problem.
So when you run Razer Game Booster, (Which I assume is some sort of utility to boost gaming performance)(I shall refer to this as RGB from now on) you get a higher FPS than with the others, and you use less System RAM. So when you use the RGB, I would assume it kills all unesasery tasks and programs, and thus frees up some system RAM.
FRAPS gives more accurate readings than the tools bundled with motherboards and video cards.
- So much so, that FRAPS itself may impact the performance once you get over 64fps.
- It's not a noticeable impact until you're getting over 128fps though, so with your GPU I wouldn't worry about it.
You can always set FRAPS to Realtime and CPU0 affinity, then disable the number and just have it dump benchmark data to a log file to review later. (Just rendering the frame rate impacts performance).
If GameBooster makes such a marked difference then I wouldn't be surprised if a slightly less trimmed application (MSI FPS) is just enough to trigger a massive slow down on your system.
This is just one of the reasons I recommend IObit GameBooster so often.
You'll never get perfect scaling in SMP. One of the cores will be nearly maxed out (or shared with a HT 'core' slightly over 100% for the pair) and the rest will just do what they can, spinning their wheels.
Not all the workloads will require a full core, and they can't process what hasn't happened yet in the simulated worlds.