Page 3 of 3
Re: evaluating xhci usb controller performance
Posted: 04 Sep 2025, 08:53
by Hyote
petergase wrote: ↑01 Sep 2025, 12:59
z690 tomahawk 12700k
Renesas uPD720202 is way better than intels xhci, using microsoft default driver
less usb coescalling? feels like rweverything imod
StarTech.com 4 Port USB 3.0 PCIe UASP - SATA / LP4 Power (PEXUSB3S44V)
I lost the screenshots but a guy on Discord tested these and concluded the default Intel controller to be superior to other USB controllers with ASMedia winning in a few tests that are not relevant in realtime use.
Re: evaluating xhci usb controller performance
Posted: 04 Sep 2025, 08:58
by kyube
Hyote wrote: ↑04 Sep 2025, 08:53
I lost the screenshots but a guy on Discord tested these and concluded the default Intel controller to be superior to other USB controllers with ASMedia winning in a few tests that are not relevant
in realtime use.
I assume you're referring to
https://twitter.com/BEYONDPERF_LLG/stat ... 04789956#m
The sample size (n=1) is too small.
The tester didn't provide repeatable steps to replicate his testing environment and thus allow others to contribute with their own hardware & OS to make up a larger sample size and give concrete conclusions.
His HW choices for the MXA tests are also not mentioned.
I do value his data, however. It's definitely something interesting to explore, as DPC/ISR driver performance is one of the most important parts in the journey of minimising CPU overhead from drivers (decreasing jitter).
I'm not sure what you refer to as “in realtime use”, as everyone seems to have their own definition of it
Re: evaluating xhci usb controller performance
Posted: 04 Sep 2025, 11:14
by Hyote
kyube wrote: ↑04 Sep 2025, 08:58
Hyote wrote: ↑04 Sep 2025, 08:53
I lost the screenshots but a guy on Discord tested these and concluded the default Intel controller to be superior to other USB controllers with ASMedia winning in a few tests that are not relevant
in realtime use.
I assume you're referring to
https://twitter.com/BEYONDPERF_LLG/stat ... 04789956#m
The sample size (n=1) is too small.
The tester didn't provide repeatable steps to replicate his testing environment and thus allow others to contribute with their own hardware & OS to make up a larger sample size and give concrete conclusions.
His HW choices for the MXA tests are also not mentioned.
I do value his data, however. It's definitely something interesting to explore, as DPC/ISR driver performance is one of the most parts in the journey of minimising CPU overhead from drivers (decreasing jitter).
I'm not sure what you refer to as “in realtime use”, as everyone seems to have their own definition of it
Nothing technical, just general computer use and gaming.