[Flint 2 / OWRT24] A+ Bufferbloat (+0ms) but BAD Hitreg in CoD - Is Single Core SQM the bottleneck?

Routers
1

Hi everyone,

​I bought the Flint 2 (GL-MT6000) specifically to minimize latency and jitter for competitive gaming on PS5 (Call of Duty & EA FC).

​My Setup:

  • ​ISP: 1 Gbps Symmetric Fiber (Spain / Digi).

  • ​Protocol: PPPoE over VLAN.

  • ​Firmware: Latest op24 (OpenWrt 24 / Kernel 6.6).

  • ​Connection: Wired (Cat 6a S/FTP).

​Current "Hardcore" Approach:

To eliminate any potential source of jitter, I am using a script that disables almost every offloading/multicore feature to force a single, consistent queue path through the CPU:

  1. ​Network Acceleration OFF: Both Hardware and Software flow offloading disabled.

  2. ​Irqbalance OFF: Service stopped.

  3. ​Packet Steering OFF: Disabled in Global Network Options.

  4. ​SQM Enabled: CAKE with piece_of_cake.qos + squash_dscp=0.

​The Paradox (The Problem):

With this configuration, I achieve a perfect A+ (+0ms) on Waveform, BUT I have to cap my speeds down to ~300 Mbps. If I go higher, the single core saturates.

​However, the in-game experience is BAD:

Even with the A+ result, Call of Duty feels unresponsive. I am experiencing significant "shoot first, die first" moments and inconsistent hit registration. It feels worse than standard ISP routing sometimes.

​My Current SQM Config:

sqm.eth1=queue
sqm.eth1.enabled='1'
sqm.eth1.interface='pppoe-wan'
sqm.eth1.download='300000'
sqm.eth1.upload='300000'
sqm.eth1.qdisc='cake'
sqm.eth1.script='piece_of_cake.qos'
sqm.eth1.linklayer='ethernet'
sqm.eth1.overhead='34'
sqm.eth1.squash_dscp='0'
sqm.eth1.squash_ingress='1'
sqm.eth1.qdisc_advanced='1'
sqm.eth1.iqdisc_opts='nat'
sqm.eth1.eqdisc_opts='nat'

Questions for the experts:

  1. ​Is "Single Core" hurting UDP processing? Could disabling Packet Steering/Irqbalance be causing internal processing latency for game packets, even if the buffer is empty? Is it better to enable Software Offloading + Packet Steering to reduce CPU load, even if it adds theoretical jitter?

  2. ​QoSMate (eBPF) vs SQM (iptables): Would switching to QoSMate (QoSify) provide better performance/hitreg on this hardware? Does eBPF handle high-frequency UDP packets better than tc-cake?

  3. ​Hardware Upgrade: Is this 300Mbps/Single-Core limit simply the ceiling of the Filogic 830 CPU handling PPPoE? Would upgrading to the Flint 3 (GL-BE9300) solve this bottleneck and allow full Gigabit SQM with proper hitreg?

​I am confused because the tests say "Perfect", but the game feels "Bad" and with low speed. Any advice is appreciated.

​Thanks!

3

Hi,

This seems a bit odd, since Waveform has shown perfect results.

Could you please confirm the following:

  1. Do you observe similar behavior in EA FC?
  2. When playing Call of Duty, does its built-in latency indicator show anything out of the ordinary?
  3. During gameplay, are there any other devices on your network that might be consuming significant bandwidth?
1 Like
4

What does libreqos say?

If it also shows good here, maybe the issue may not related to internet.

I used to play competitive games, but more in gta online.

One trick most players don't know is optimizing your screen settings, especially this:

It could be that you play against a player with a much higher refresh rate, this means if you use a monitor with 60hz, vs someone with 240hz he is able to respond faster than you and if he is also pro smart and strafes behind edges of walls, you lack the amount of fps and he appears to you behind walls and such, even though that is not true your monitor lacks the necessary frames.

hz is the refresh rate of the monitor, but you also need to be capable to reach that amount of fps, some propertairy functions may cause extra delays such as framegen with in combination with nvidia reflex, often both are required for framegen to function (it still delays, so I advise against it) only a small hand of monitors have it the nvidia reflex but it isn't a requirement, you also can play without framegen and reflex that is maybe better, framegen however is not a solution it's the whole fake frames it adds that they even need reflex (a pretty old tech) to fix some of the response times lol 25ms is still really bad.

But because hz need to be higher doesn't mean you play better, you have to adjust and things become much more intense, imo max 240hz is what the human eye can intercept, higher is bs most 'pro' gamers ive seen also don't go higher than 240hz :smiley:

With vsync you can lock the fps to your monitor hz, so it will not go higher if that happens you will have a stuttering issue.

Refresh goes hand in hand with input lag it is important this works both correctly.

Also in most online games, often you have to shoot to the direction you expect the player to move, the closer they are the more accurate it hits, but the further away the more away you need to shoot for the bullet to land / travel... I have tried headshotting airjets but I had to move like 300 gta meters away for it to work, you have to measure and try then, each game has this different.

1 Like
5

Hi, thanks for the reply! Here are the details:

  1. EA FC Experience:
    Yes, the behavior is noticeable in EA FC as well. It feels "heavy" or unresponsive, suggesting some internal latency processing issues even if the bufferbloat test looks good on paper.

  2. The Main Issue (The Speed Cap):
    The most concerning part is that to achieve that A+ rating, I have to cap my 1Gbps connection down to 300 Mbps (300000) on both Download and Upload.
    Capping a Gigabit line to 30% capacity seems excessive for this router. If I raise the limits above 300 Mbps, the single-core CPU saturates (since I disabled offloading/steering to avoid jitter), but the gaming experience seems the same.

  3. Call of Duty In-Game Meter:
    The in-game latency meter is stable (flat 15-20ms), and I don't see packet loss icons. However, the gameplay feels desynchronized ("instant deaths", shoot first/die first), regardless of the flat ping line.

I suspect the issue might be related to how the Flint 2 handles PPPoE over VLAN combined with SQM. It feels like the CPU struggles significantly with the encapsulation overhead when SQM/CAKE is active, forcing me to lower the speeds drastically to maintain packet pacing stability.

6

Hi, thanks for the suggestion!!

  1. LibreQoS Test Results:
    I just ran the test on bufferbloat.libreqos.com and, just like Waveform, I get an A+ grade.
    So, the SQM logic is definitely working correctly, keeping the latency flat... BUT only because I have capped the speed at 300 Mbps.

    image
    image1659×1842 267 KB

  2. Monitor & Input Lag:
    I appreciate the advice, but I have already ruled out the display as the issue. I am playing on a high-end setup optimized specifically for low latency:

  • Monitor: LG Ultragear 27GP95RP-B (Nano IPS).
  • Settings: Running at 120Hz on PS5 via HDMI 2.1, Response Time set to "Fast", and unnecessary post-processing (e.g. Black Stabilizer high values) disabled to minimize input lag.

So, it's not a 60Hz vs 120Hz issue. The hardware side is optimized.

7

Based on the current information, the issue does not appear to be network-related.
Both Waveform and LibreQoS tests show perfect results, and the in-game latency counter also indicates normal operation.

Because FPS gameplay can be influenced by many non-network factors (frame rendering time, monitor response, motion blur, and more), we recommend consulting gaming-focused communities for deeper analysis.

Regarding Flint 2 reaching only about 300 Mbps when limited to a single core, this behavior is expected. In your setup, one core is responsible for forwarding, PPPoE, VLAN processing, SQM, and related tasks.

If you require higher throughput, enabling Packet Steering can help distribute the load.
This typically does not introduce noticeable latency variation.