The platform: Windows 10, 64GB (fastest latency) RAM, GPU with 11GB video RAM, 8-core (16 threads) Intel CPU running at default (no overclocks anywhere) 5.1 GHz. Page file on separate PCIe channel (not South Bridge), other cache/temp file/folder classes on separate PCIe lanes too.
With latest firmware, drivers, versions, updates, etc. Processing is kept "local" in cases where I have a choice.
Mozilla: Firefox 135.0 (64 bit)
The behaviour: regular Firefox in a few windows with a few tabs each & private Firefox in the same way.
The setup: disabled Firefox's GPU co-processing as I need the RAM for Adobe's Camera Raw, Lightroom Classic, Photoshop. Because Firefox uses too much in there. Workarounds to prevent aMeSs-Windows from hogging the GPU (e.g. set all folders to text directories).
The understanding: Research B/M degree in software engineering with AI. Yes, a browser is a multiprocess thing.
The issues: (1) Firefox still manages/managed incidentally to hog more than 1GB of GPU-RAM, while not formally allowed. (2) My impression is that after closing ALL private windows/tabs, Firefox does not terminate the processes and does not release the memory to the pool. (3) With only one window/tab in the foreground and only that one in the visible GUI layer, all open tabs (included from closed tabs?) keep actively updating active content at the same rate and priority, and this causes CPU level context switches that have a similar effect as infinite "paging" in anachronistic times. The reality is that in browsing, we open a bunch of pages - in absence of some mental support by the software - as reminder that we need to verify something, still (as "to do").
Is Mozilla convinced of their memory management? Can Mozilla run non-foreground stuff at much lower priority? Sorry to day, as all-time only FF user and its ancestry, but you are competing with MS and Adobe now for hogging the system. Yes, there is a price for optimisation, and we could not afford the hardware in the 80s, but not everybody keeps up with the gamer-Jones's. My workstations and their systems configuration gave me 98th percentile in SPECmark. I would not expect >any< problems at that.
