sMartins wrote:I typed again the :die command, maybe you can check if there's some difference to understand which effects or such trigger those low perfomances.
No relevant difference that I can find, at least.
sMartins wrote:I resetted all to default settings a couple of times ….. and now looks back up to 60 fps, tbh it's very weird cause I didn't manage yet to discover which settings trigger so slow fps, mostly due to the fact that I had less stuff enabled previously, such as anisotropic filter turned off, max perfomances power supply and stuff like that, for more perfomances, not lower.
Hmm, what you're seeing
could possibly be the same thing as an effect I've been seeing, namely that what happens when you're very low on video memory is that the rendering performance drops very significantly, probably due to OpenGL swapping textures between video and system memory. That being said, the only reason I'm getting so low on video memory from time to time is that my video card only has 1 GB of video memory and Chrome uses like half of that (:P), so it can happen sometimes and sometimes not (depending on how much VRAM Chrome is using at the moment) when I'm running two clients at the same time. Using antialiasing increases the client's VRAM usage by quite a lot, so that tends to increase the likelyhood of it happening.
What is worth mentioning about this phenomenon is that, once a client has entered the state where it's partially pushed out of VRAM, it seems that it can't get everything back into VRAM even if I close the other client, so once that happens, that client needs to be restarted, so it could possible account for what you're seeing. What is strange about it is that if it happens on your GF970, however, you'd have to have some other VRAM hog in the background that is much, much worse than my Chromium. But you could verify whether this is the cause by checking your total VRAM usage when it happens. If it's close to full, it's likely the culprit. That being said, I don't know how to check VRAM usage on Windows.