You are doing this in the same computer? The packetloss, I assume, is reported inside valorant?
SQM is not going to prevent drops (drops being the signal AQMs use to tell endpoints to slow down, sqm will tend to drop overall less than a FIFO and a bit earlier resulting in better overall performance).
Hard to tell without data.
So maybe you could run a few experiments so we can try to dive a bit deeper.
Fist let's see how you configured sqm:
Please post the output of the following commands:
ifstatus wan | grep -e device
tc -s qdisc
Then lets see how Origin downloads behave, for this you need to prepare to start a large enough Origin download (no idea whether that can be done on demand or whether you have to wait for new updates to be offered):
- clean the sqm start, by stopping and starting sqm (please post the output of the tc-s qdisc call):
/etc/init.d/sqm stop ; /etc/init.d/sqm start ; echo "" ; tc -s qdisc
- start your Origin download
- wait for say 10 minutes (assuming the download is large enough) and run tc again and post the output:
tc -s qdisc
- also run 'iftop -e $YOURWANINTERFACE` (replace $YOURWANINTERFACE with your true wan interface name) and try to see how much bandwidth the Origin download consumes in Upload and Download direction.
The we can compare the tc output for the two cake instances from before and during the download which should tell us something about the properties of Origin downloads.
Then repeat the same but instead of Origin, play valorant so we get an idea about valorant's network demands. (Please include the iftop step as well, as I have some indication that Valorant might cause significant network traffic by itself and might not fall into the existing ideas how little traffic FPS games produce, probably due to its 128Hz ticks*)
However, I would have thought that nobody would even attempt large scale downloads on the same machine they are playing a real-time critical FPS game like valorant on....
*) If we assume you play with 19 others on a server and that each world tick the server sends one full MTU packet per player (including yourself) at 128 Hz we can expect:
((1544*8) * 128 * 20)/1000^2 = 31.62112 Mbps
That is a lot of traffic. Now Valorant might use smaller packets and/or package more players into individual packets, so this is likely a worst case estimate, but even if Valorant only send 1/3 of this, we are still talking about 10 Mbps.
Now with your 255Mbps link this number should not be a big problem, but if Origin would use say 24 parallel flows each of the 25 flows now would get 255/10 = 25.5 Mbps which for our 31 Mbps estimate would already be to little.
One remedy, if you download and play on different machines, would be to try cake's per-internal-IP-fairness mode, where the Download machine and the Valorant machine each would get 255/2 = 127.5 Mbps guaranteed, which would comfortably fit valorant. And since Valorant only uses 31 of these 127 the rest would still end up being used by the download. However this assumes that game and download use different internal IP address.
The other remedy, if both uses are on the same computer, would be to simply not do this at the same time, play something less time critical than a FPS then, like chess