FixNetworkBufferbloat script contains harmful and obsolete tweaks = FUD

[ArtanisInc]

The script applies numerous tweaks that are now considered obsolete, counter-productive, or based on networking myths. For most users on modern systems (Windows 10/11), these changes are more likely to degrade network performance and stability than to improve them...

Here is a breakdown of the most problematic modifications (not exhaustive):

1. TCP Auto-Tuning is Disabled
   The Problem: TCP Auto-Tuning is a critical Windows feature that dynamically adjusts the TCP receive window size. Disabling it cripples throughput, severely limiting maximum download speeds on any modern high-speed connection (Fiber, 5G, etc.). It's a practice that has been advised against for over a decade.
   Code Reference: netsh int tcp set global autotuninglevel=disabled

2. Downgrading the TCP Congestion Algorithm
   The Problem: The script replaces the modern, highly efficient CUBIC algorithm with the older CTCP. CUBIC is the default standard on nearly all modern operating systems for a reason—it performs better across a wider variety of network conditions. This is a clear downgrade.
   Code Reference: netsh int tcp set supplemental internet congestionprovider=ctcp

3. Aggressive & Risky ACK Behavior
   The Problem: The combination of TcpNoDelay=1, TcpAckFrequency=2, and TcpDelAckTicks=0 creates a risky "send immediately, acknowledge patiently" behavior. Setting TcpDelAckTicks to 0 disables the delayed ACK timeout, forcing the system to wait indefinitely for a second packet before sending an ACK. On networks with irregular packet flow, this can introduce latency instead of reducing it.
   Code Reference: sp ... TcpNoDelay 1, sp ... TcpAckFrequency 2, rp ... TcpDelAckTicks 0

4. Useless and Myth-Based TTL Modification
   The Problem: The script changes the default Windows DefaultTTL value from 128 to 64. This modification has zero impact on ping, latency, or speed. It's a classic "cargo cult" tweak that provides no benefits and, in very rare cases, could theoretically cause packets to be dropped on extremely long routes.
   Code Reference: sp ... DefaultTTL 64

5. Disabling Modern TCP Recovery Features
   The Problem: The script disables rack and taillossprobe. These are beneficial, modern TCP features designed to improve performance and recovery over lossy networks (like Wi-Fi or 4G). Disabling them is a step backward in network resilience.
   Code Reference: netsh int tcp set supplemental internet rack=disabled, taillossprobe=disabled

As the script's own comments wisely state: "You should upgrade to a router with fast cpu and ram having Smart Queue Management"...

[AveYo]

FixNetworkBufferbloat - AveYo, 2025.09.15
test on waveform.com/tools/bufferbloat , speed.cloudflare.com , speedtest.net
You should upgrade to a router with fast cpu and ram having Smart Queue Management
This script is no SQM, but just a short term network limits configuration!
to skip the dialog use cmd parameters or rename script ex: fix_network_bufferbloat upload.bat
Upload fix limits up speeds, Download fix limits single-part dl, but games benefit!
Phone tethering or poor wireless signal? Should select Both

Are you serious?!
Script is god damn bleeding edge for anyone stuck with 4G tether or xxx/50 DL/UL underprovisioned scam isp
No chance in hell you can achieve such latency reductions with just a set-it-and-forget-it script
That's what bufferbloat is about - latency, not throughput!
And I was the first to figure out how to manipulate windows QoS without extra 1000ms latency..

- TCP Auto-Tuning Disabled = first step at reducing latency on unstable connection (and a GUI choice)
- Downgrading the TCP Congestion Algorithm = CUBIC is quite aggressive and I want TCP to take a backseat over realtime UDP
- Aggressive & Risky ACK Behavior = FUD - educate yourself, TcpNoDelay 1 disables Nagle's algorithm which has not been suitable since the '80s, TcpAckFrequency 2 is default (send an ACK every other packet) and TcpDelAckTicks was literally deleted
- Useless and Myth-Based TTL Modification - just think for a second; anything above 64 hops is such a bad routing and added latency that it's practically unusable, specially for the mentioned 4G tether or xxx/50 DL/UL underprovisioned scam
- Disabling Modern TCP Recovery Features - I'd wish.. but FYI, those are readonly and default enabled

Go to line 38, set $PAUSE_FOR_SUMMARY = 1 and run the freaking script in a vm or whatever it's more verbose
Run SG TCP Optimizer after and it will show a very sane preset

[ArtanisInc]

Fair enough... I get that your script targets edge cases like 4G tethering and weak uplinks. But let’s be clear:

On modern broadband (fiber/5G), disabling Auto-Tuning and downgrading CUBIC is a regression, not an improvement. If latency is truly the goal, the modern solution is BBR2 but some apps like steam won't work as intended...

Forcing TTL=64 does nothing for latency: TTL only decrements per hop, and no real route ever comes close to 128 hops. Lowering it doesn’t speed anything up... It just risks breaking edge cases.

Your ACK tweaks may reduce delay in some flows, but globally forcing them can actually increase packet overhead and lead to congestion instead of improving responsiveness.

The “disabled” TCP recovery flags are cosmetic at best, misleading at worst.

I’m here for the technical discussion, not the drama. If you have any documentation or detailed analysis on your tuning choices, I’d be happy to review it.