Solved Chromium slow

i believe that could be the case for me too, which i'd guess it actually comes from the journaling updating the journal every minute or so, and so there'd be times where the information was so much to write that the system'd freeze for that tiny amount of time

maybe i should try booting with journaling disabled to confirm, shouldn't hurt to try right
 
UPDATE: i currently have journaling disabled right now and it seems to have gotten just a little better, what even is the rationale behind having soft-updates and journaling enabled at the same time anyways

however i just noticed this little process running in the background, syncer, which apparently is responsible for "flushing volatile cached file system data to disk" every moment
i guess this one is the bigger culprit here then? not really sure how to check processes by drive I/O so i'm guessing stuff again
 
I'm no expert, but iostat -w 1 will at least print out io information every second, so you could have numbers. Perhaps you can relate that to running processes such as syncer.
 
well it doesn't say which process is causing such I/O spikes, but i guess it should at least be good enough to confirm that the lagging might be because of that
 
UPDATE: i was right:

1761517194565.png


523.90 MB/s··· this is literally ~95% of my SSD's speed··· what the hell could possibly be writing so much data to it?
i wonder if there's a way to list processes by how much I/O they're using
 
Some browsers do different stuff with different Xorg DDX (Firefox software-renders if using Intel DDX but hardware-accels with modesetting; there could be other things with amdgpu vs modesetting and NVIDIA)
 
i see, well i was able to reduce the issue even further by changing the following sysctl parameters:
Code: 
# try increasing the file limits
# so that i (might) get a smoother
# experience on chromium
kern.maxvnodes=1048576 [original was 520k-ish]
kern.maxfiles=524288 [ditto]

# try increasing my SSD's
# performance too
vfs.read_max=128 [original was 64]
not sure what exactly about these made it work better but hey they did
i might try to keep messing with those numbers until i completely eliminate the issue, for now i seem to be on the right track

also,
Firefox software-renders if using Intel DDX but hardware-accels with modesetting
Click to expand...
what exactly does that mean? is it why i don't get hardware acceleration at all in firefox, even tho' i even have FFmpeg compiled with VA-API/Vulkan support enabled?

EDIT: just found out, i checked sysctl vfs.numvnodes and that shit was at nearly half the original value of vfs.maxvnodes, so i guess that tunable was already enough to help
 
ruby R53 said:
what exactly does that mean? is it why i don't get hardware acceleration at all in firefox, even tho' i even have FFmpeg compiled with VA-API/Vulkan support enabled?
Click to expand...

In about:support a little down there's Graphics -> Compositing; it'll have just WebRender if it's hardware-accelerated (no "Software" mention)

Firefox's graphics blocklist disables hardware-acceleration if Intel DDX is used on Xorg (likely bugs). Setting about:config -> gfx.webrender.all true force-enables hardware-acceleration. The default DDX is modesetting and shouldn't be blocked from hw-accel.
 
You must log in or register to reply here.
Back
Top