On Tue, 30 Aug 2016, Greg wrote:
I suspect the previous USB chipset was causing me more problems than typical,
as I am now very much enjoying the OOTB experience. I did see updates to the
driver as well (workarounds/adding delays) in the time that had elapsed.
I am told that all motherboard chipsets are less than they could be. Intel
or VIA usb chipsets generate more interrupts than NEC (or maybe SIS). The
defining factor it seems is the USB 1.1 part of things. There are two of
them intel/via use the UHCI driver which expects the MB CPU to do lots of
the work. Other chips use the OHCI protocol which does as much work as
possible in the chipset. This one uses a NEC chipset but is USB3 so I
don't know what the control protion is:
http://www.newegg.ca/Product/Product.aspx?Item=9SIAA0D4C34273&cm_re=PCIe_USB_card-_-9SIAA0D4C34273-_-Product
For something more expensive that has OHCI in the spec:
http://www.newegg.ca/Product/Product.aspx?Item=N82E16815114048&cm_re=PCIe_USB2_card-_-15-114-048-_-Product
I do not know where things fall in the OHCI/UHCI with USB3 chipsets... but
I think we can be pretty sure Intel/VIA go the cheap way. Thing is, even
if USB3 doesn't use the O/UHCI part of things, almost all multichannel USB
audio interfaces do. It might be worth while having a list of known good
chipsets/PCIe USB cards.
BTW, even with uhci drivers, I found it made a big difference to:
a) make sure the irq that goes with that driver/usb port is not shared.
b) rtirq lists that usb port separately from the rest (IE. usb3 usb, not
just usb)
c) nothing else is plugged into that port via a bridge/hub whatever.
This meant for me (on my netbook) only using the USB port on the right
side, not using the second USB port on the right side... adding a hub to
the left side USB port for everything else (I was running from a USB hard
drive at the time). I was able to run the (USB 1.1) audio device with jack
at 64/2 with no xruns with this set up. (Atom single core at 1.6Ghz) Test
duration being overnight so 6 to 8 hours. (cron turned off, ht off(well
told Linux only one core), CPU gov performance (also tried userspace at
800MHz with success), setting rtirq RTIRQ_NAME_LIST="usb3 snd usb".
For someone who has done some real world testing:
http://crimeandtheforcesofevil.com/blog/2016/07/25/so-hey-usb-chipsets-totally-matter/
Point seems to be, that those going portable who rely on onboard USB...
you may want to load linux onto a Mac which I am told will have good
chipset, or expect to have higher latency. This is ok for recording, not
for softsynth or effects. Otherwise add a new USB card.
I do not know if the UHCI and OHCI drivers can run side by side or if the
internal USB would have to be disabled.
--
Len Ovens
www.ovenwerks.net
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev@lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-dev
