My roommate started a 4K Netflix binge one night while I was mid-ranked-match in CS2, and I went from a clean 28 ping to full-on rubber-banding in about ten seconds. I blamed my ISP, then my router, then genuinely considered moving my PC to a different room with its own line. None of that would have fixed it, because the problem wasn’t bandwidth at all — it was something called bufferbloat, and once I understood what it actually was, the fix took about fifteen minutes and didn’t cost me anything.
The Part Nobody’s Speed Test Shows You
Here’s the confusing part: my download speed was fine the entire time. A regular speed test would have told me nothing was wrong. Bufferbloat only shows up under load — your ping is fine sitting idle, but the moment your connection gets busy, it can jump from 20ms to 500ms even though you’re nowhere near your max bandwidth.
The actual cause is buffers. Routers and modems hold extra data in a queue so packets don’t get dropped when the line is busy, but oversized buffers mean packets sit there for hundreds of milliseconds before they’re sent out. Your game sends small, constant, time-sensitive packets, dozens of times per second, telling the server exactly where you are and what you just did. Those packets get stuck in line behind someone else’s video stream or a cloud backup, which sends much bigger chunks of data less urgently. The queue doesn’t know your packet matters more — it just processes things in order.
The upload side is usually the real culprit here, and this is the specific mechanic most explanations skip entirely: a typical home connection might only have 40-80Mbps of upload, and something as ordinary as a video call, a Steam cloud save syncing in the background, or an iCloud photo backup can fill roughly 200ms worth of that upload buffer on its own. Your download speed never moves. Your game still spikes.
This is also why cable and DSL connections get hit harder than fiber. Fiber has enough dedicated upload capacity that the same buffer rarely fills; cable’s shared, lower-upload architecture chokes much faster under the exact same load, because there’s simply less headroom before the queue backs up.
How to Actually See It Happening (Not Just Take My Word For It)
On PC (CS2 and Similar Games)
If you play CS2, open the built-in Telemetry overlay, which replaced the old net_graph command. Watch your ping and jitter live during a match. You’ll see the exact moment someone else’s traffic hits the network, because your ping trace goes from a flat line to a visible sawtooth pattern. That visual is the whole diagnosis: flat means fine, jagged means something on your network just started competing for the queue.
On Xbox
Go to Settings, then Network, then Test network connection and statistics. This built-in tool shows your ping, download and upload speed, and your packet loss percentage directly — no third-party app needed. Run it once with the house quiet, then run it again while someone streams or downloads something, and compare the packet loss numbers.
On PS5
PS5’s built-in connection test doesn’t expose packet loss as cleanly as Xbox’s does, so the practical test here is elimination rather than a direct readout. Pause downloads, cloud uploads, game updates, and anyone else’s video calls or streams, and see if the lag clears immediately. If it does, you’ve confirmed it’s a queueing problem on your own network, not your ISP or the game server.
Any Platform: The Waveform Test
You can also run the Waveform bufferbloat test from any browser on any device on your network. It loads your connection with real traffic in both directions and grades your latency under load from A to F, separately for upload and download. If you score a C or worse, especially on the upload side, this is very likely your problem.
The Actual Fix
CAKE vs. fq_codel: What You’re Actually Turning On
The setting you want is called SQM, Smart Queue Management. It works by intentionally managing your router’s queue instead of letting your ISP’s oversized buffer do it, so small, time-sensitive packets like game traffic and voice calls get pushed ahead of big bulk transfers instead of waiting in line behind them.
Two algorithms show up under this label. CAKE is the more effective and more actively developed option, but it needs more CPU power from your router to run well. fq_codel is the lighter, older fallback, better suited to older or lower-end router hardware that would choke trying to run CAKE at high speeds. If your router supports both, CAKE is the better choice; if it’s an older or budget device, fq_codel is the safer one.
If You’re on an eero Mesh System
This is hiding in plain sight, worth pairing with our breakdown of whether gaming routers are actually worth it. Open the eero app, and look for a setting called Optimize for Conferencing and Gaming — this is eero’s plain-English rebrand of SQM. It’s a single toggle, no separate firmware or app needed. One caveat worth knowing: it isn’t available if your eero is running in bridge mode, and turning it on may slightly cap the bandwidth ceiling for any one device in exchange for more consistent latency across all of them, which is the entire point.
If You’re on ASUS Hardware With Merlin Firmware
You’ll need Merlin firmware version 386.2 or newer. From there, install the CakeQOS-Merlin script, then enable Cake directly in the router’s web interface under Adaptive QoS. Cake and standard Hardware Acceleration can’t run at the same time — the script disables Hardware Acceleration automatically when you turn Cake on, and the router needs a reboot to apply it. If you already have an older CakeQOS-Merlin script installed from before version 2.x, uninstall it first to avoid conflicts with the newer version.
If You’re on OpenWrt
Install the luci-app-sqm package (or just sqm-scripts if you don’t need the graphical interface). In the Luci web interface, go to the SQM-QoS page, select your WAN interface, and check the Enable box. Set your Download and Upload speed fields, then go to the Queue Discipline tab to choose Cake specifically rather than leaving it on a generic default.
The Number That Actually Matters More Than the Algorithm
Whichever option you’re setting up, set your upload and download limits to about 85 to 90 percent of your real tested speed, not your plan’s advertised speed. If your actual measured download is 100Mbps, set the SQM download field to roughly 85,000kbps, not 100,000. Setting it too close to 100 percent defeats the entire purpose, because the router needs headroom to actually manage the queue itself instead of just recreating the same full-buffer problem your ISP’s equipment already has.
If Your ISP Has Rolled Out Low Latency DOCSIS
Some newer cable modems are starting to fix this at the hardware level with something called Low Latency DOCSIS, or LLD, which brings modem-side bufferbloat down without you touching a single router setting. It’s worth checking whether your ISP has deployed it on your specific modem model, since if they have, you may not need to configure SQM manually at all.
If You Set All This Up and It’s Still Not Fixed
There are two common reasons SQM doesn’t fully solve it. First, if there’s another router upstream of yours, like ISP-provided equipment in modem-router combo mode, that device’s buffering becomes the real bottleneck, and your SQM settings downstream of it won’t help much. In that case, put your ISP’s equipment into pure bridge mode if it supports one, so your own router is the only thing managing the queue. Second, Wi-Fi itself adds its own buffering and retransmission delay on top of anything happening at the router level, so if you’re still seeing spikes after enabling SQM, connect the device that’s lagging directly via Ethernet before assuming the fix failed. On a high-bandwidth fiber connection with plenty of spare upload capacity, the practical benefit of SQM can also be small enough that you won’t notice much difference, which is a sign your connection was never the bottleneck in the first place.
If you’re weighing this decision, it’s also worth reading Cloud Gaming Isn’t Broken by Slow Internet โ It’s Broken by This Number Speed Tests Don’t Show.
FAQ
Why does my internet speed test say everything’s fine but I still lag when someone else is on the network?
Speed tests measure your connection at idle, not under load. Bufferbloat specifically only appears when the connection is busy, which is exactly the condition a normal speed test doesn’t create or catch.
Does this affect console gaming too, or just PC?
Yes, on any platform — it’s a network queueing problem, not something specific to PC. Xbox has a built-in test that shows packet loss directly; PS5 users should use the elimination method of pausing other network activity and checking if the lag clears.
Do I need to buy a new router to fix this?
Usually not. Most modern routers, including eero mesh systems and ASUS routers on Merlin firmware, already support SQM — it’s just switched off by default and often labeled in a way that doesn’t obviously connect to “fixing lag.”
Will SQM slow down my other downloads while I’m gaming?
Slightly, and that’s intentional. SQM prioritizes small, time-sensitive gaming and voice packets ahead of large bulk transfers, so a big download might take marginally longer to finish, in exchange for your game not spiking every time it’s running.
Related Guides
- Are Gaming Routers Actually Worth It
- Best Gaming VPN to Reduce Ping in 2026
- Can a VPN Improve Gaming Lag? The Honest Answer
- How to Reduce Gaming Network Latency
1 thought on “Why Your Game Lags Every Time Someone Else Streams Netflix (And How to Actually Fix It)”