SG Broadband Community

i see, so its still better to have a dedicated NIC. i also have a older machine that has a Intel NIC that i used to have a dedicted card for. i thought it was the same thing, but its good to know that onboard NIC is worse than dedicated even if its the same chipset.

I leave it at 4 queues as opposed to 8 queues. I have a Marvell AQtion NIC. I forgot the model number, but its a recent on-board NIC from an Aorus Master Z790 with the latest drivers. Have you ever personally found 8 queues to work better with a certain brand? Would 8 queues be more for a corporate ...

Has anyone else noticed that using "8 Queues" for the "Maximum number of RSS Queues" under NIC properties slows down your internet speeds? I have a 13900kf configured with E-Cores and HT disabled, so I have 8 active cores running, which correlates with setting this value to "8 Queues".

Could ...

After doing some googling, I decided that with modern connections nowadays, 1000 should be enough for a tineout. Considering even W11 feels this way, I would make it default for TCPOptimizer as well.

It looks like for Windows 11 Canary 27723, the InitialRto setting defaults to 1000 as opposed to it being 3000 previously. Would you say this is a more logical and modern setting nowadays? Or should we leave it at 2000/3000 still for gaming?

Doing some googling and reading articles, I came to the conclusion that TCP Pacing suits more of a server environment, as it prevents "bursts", and delays traffic. Common sense tells me this will increase latency, so Im going to say keep this setting off for gaming.

https://ieeexplore.ieee.org ...

My apologize, I should of mentioned I did indeed restart and did this test 3 times each with it set to 0 and 1, making a total of several computer reboots.

Contrary to the documentation, it appears this setting has no effect on speeds and toggling TCP1323 Window Auto-Tuning. I get an average of 925mbps whether its on or off. This is the results I got before ever messing with the setting. To be on the safe side, and to follow the documentation, the ...

Is the TCP autotuning diagnostic tool required when using TCP Window Auto-Tuning? I definitely have a philosophy of disabling bloat for lower latency. Does it need to be enabled for TCP Window Auto-Tuning to function? Does this setting improve how said option works? Is it better to leave enabled for ...

I did some benchmarks between disabled, normal, and experimental. Heres the results.

disabled:
https://ibb.co/VS2BFtj

normal:
https://ibb.co/J7n0V6x

experimental:
https://ibb.co/0QHkFKG

Using normal setting defintely has the max throughput like you said, while disabled nerfs my throughput ...

Is that the scenerio the test creates?

What would be the best setting for the TCP Pacing Profile option? I currently have it set to "slowstart" according to this tweak guide.

https://sites.google.com/site/tweakradje/windows/windows-tweaking

It does mention its for Hyper-V 2019 servers, so im very skeptical and feel it should be ...

So running "Get-NetAdapterLso -Name *" does confirm its disabled, but TCPOptimizer still reads enabled. Do you think its a bug with TCPOptimizer reading the setting? Heres a pic to show you what I mean. Im on the wired Ethernet profile.

https://ibb.co/djy68mG

I was switching between CTCP and CUBIC for testing, but wanted an easier way to test BBR2 than using netsh.

The biggest change I did to increase my score was to actually completely disable TCP Auto-Tuning and all Offloading options except Checksum offloading, via TCPOptimizer and NIC properties.

When disabling Large Send Offload (LSO) via TCPOptimizer and doing a reboot, the setting claims to still be enabled when opening TCPOptimizer after reboot. When going to NIC properties, LSO says its disabled, for both V1 and V2, along with IPv4 and IPv6. I have a Marvell AQtion 10GBASE-T NIC with ...

Please add BBR2 option to Congestion Control Provider setting for TCPOptimizer.