Discussion in 'Operating Systems' started by mbk1969, May 7, 2013.
"Audio Bus" is only one. Second one is "Audio Controller".
Turn both to MSI mode.
HD Audio is a controller
Realtek is an endpoint
You set the controller to msi, not the endpoint.
Btw, that means that interrupt mode has nothing to do with your problem - delay. Audio enhancements are the culprit.
You can try to upgrade audio components (Sonic Studio, Nahimic) from here - https://github.com/AlanFinotty/AAFDCHDriverMod
Got something strange, it's not related to the utilities but my one NVMe Controller is using a limit of 65 while it's max limit is 33. Does that seem wrong?
"Limit" value is from registry, and "max limit" is from device properties (got with Win API).
I assume that drivers for your controller are just set "limit" value to some generic/common number.
No harm from that because the actual limit is calculated as minimum value among "limit" and "max limit".
If I disable MSI Mode on an XHCI Controller (native MSI) and disable XHCI Hand Off in BIOS, it probably isn't the same as using an EHCI Controller because "I am emulating", is that correct?
MSI mode has nothing to do with "xHCI vs EHCI". Interrupt mode lies much deeper than any "*HCI" standards.
In BIOS you do not disable xHCI, but only "Hand Off", right?
I know that MSI has nothing to do with xHCI or EHCI. Either you didn't understand my question correctly or it's not easy to answer.
Correct, in BIOS I can only disable "xHCI Hand Off".
In the end I'd like to use "EHCI" with modern mainboards, the only clue I have is that EHCI didn't have MSI Mode on my old system.
XHCI handoff is not required on Windows 10 or 8, its an OS compatibility setting for systems that don't natively have handoff support in the kernel.
Since MSI mode has nothing to do with USB standards it can`t change anything in USB controller (and the way it works with USB devices).
"Hand Off" setting in BIOS is not about disabling xHCI mode, it is about handling the xHCI controller at different stages of the boot - BIOS and then OS. On modern OS (with proper support of USB 3) that setting should be set to "Disabled".
is the onboard sound a Realtek one?
If yes, which motherboard brand is it?
Does someone know, if the Realtek HDA Codecs use MSI-X mode on linux by default?
Yeah, Realtek. Board is an MSI B550-A Pro.
You can see that in the /proc/interrupts file. Devices in MSI mode are marked with "IR-PCI-MSI". It also shows the current interrupt count (how many times there was an interrupt request) per CPU core.
Edit: Oh, it doesn't distinguish between MSI and MSI-X. But does it matter?
On my ASUS board (also using Realtek) MSI Mode is not enabled by default.
Which driver did you use?
I downloaded the audio driver for your motherboard and checked the INFs.
There is no mentioning of MSI Mode support.
I don't have any linux distro installed atm and I'm too lazy to create a bootable live disk/stick.
So I can't check this.
But looking at the source code of the linux HDA drivers, it seems like, some devices have MSI Mode disabled by default.
And there is something probing by the driver to check for MSI Mode support.
I'm asking because I get some very rare stutter on my system, like once a week.
The entire system stutters for ~1 second.
I wonder if this related to MSI Mode. I enabled it on my sound card + NVIDIA GPU + NVDIA HDA.
The stutter reminds me a bit of my old system where the stuttering was immediately noticeable.
The default one that Windows uses. I did not install anything from Realtek's site.
I didn't install anything from MSI's site either. The only thing I installed manually was the chipset and Ryzen CPU power plan driver from AMD's site.
I don't think MSI mode has anything to do with this. I haven't seen it affecting anything, ever. I just force the GPU to MSI mode on Windows now because... it seems natural not to use legacy stuff. But it didn't improve anything. The system behaves 100% exactly like it did before.
Hi, in basically every resource that I've come across using this tool there's a single number under the IRQ, in fact on an older install of windows that was the case for me as well.
However, for some reason I have this long list of numbers and was wondering if it's a potential problem, or something to not be concerned about? I just reinstalled windows today too. Thanks!
List of numbers in "irq" column means that such device uses multiple interrupts (multiple MSIs).
If you do not perceive any issues then do not change anything. Otherwise you can set limit value (in "limit" column) to the "1" or to the number of CPU cores.
Also you can disable unused devices, like USB host controllers, NVM Express controllers. You can check for a devices plugged into those controllers by switching a view to "Devices by connection" in Device Manager. And better disable in BIOS.
Also, are you sure you installed AMD chipset drivers?
@mbk1969 I don't think I have any issues, though my latencymon results aren't ideal - neither are my "mouse tester" polling frequency vs time plots. That's to be expected with a 5950X instead of a 10900K I suppose? I believe the Fresco Logic controller is my Sonnet USB 3.0 card. Both of my M.2 slots are being used, one is an SN850, and the other is an SX8200 - can I still disable these controllers (I'm assuming no)? As a side note I definitely do not use the HD audio controllers, so I guess I can disable that? I have BT and WiFi disabled along with a bunch of other stuff in the BIOS.
I'm confident that I installed the chipset drivers, is there anything that makes you think I may not have?
I'll be setting interrupt affinity today to get drivers off of core 0, along with other small tweaks to improve my DPC latency, but just wanted to make sure that I'm going about this stuff the right way beforehand. Thanks for your time.
No, since both are used.
One of them you have to use, otherwise you would lack audio from your rig. Unless you use external USB audio device.
Seeing Microsoft drivers for AMD USB 3 controllers. I was sure AMD has their own USB drivers. (But maybe only for USB 2?)
Do waste your time on that only if you know how to estimate the effect or know perfectly well OS kernel and device drivers intrinsic "kitchen".
In terms of interrupt affinity, I was going to use latencymon to measure the driver execution impact after switching x or y to a different core, I remember last time I did it there was a clearly measurable difference.
I'll look into the AMD USB 3.1 driver issue and will disable Nvidia USB as well - also yes, I use an external USB DAC.