Discussion in 'Operating Systems' started by mbk1969, May 7, 2013.
OK, you are the first one. -)
I have a ZxR too and it works fine as well. And my processor is an i7-6700K since you said something about it.
Intel i5-4670K is 4 core cpu with no ht
So it has 8 logical cores. In any case if network card works you should not freak out.
i bet this limit only applies when not using hardware HPET interrupts, which uses the cpu instead
TSC+LAPICs Low performance (slow timers + syncing) = 2.76MHz
LAPICs low performance (slow timer - no syncing) = 3.5Mhz
TSC+HPET medium performance (slow and fast timer + syncing) = 3.8Mhz
HPET high performance (fast timer - no syncing) = 14.3MHz
when measuring DPC latencies in win10 and 8 make sure you disable dynamic ticks or else u wont get the correct results
posting this when am using MSI mode. impressive right?
Impressive. Would be more impressive if you`d posted results in legacy mode too (with much more big latencies). :nerd:
you mean the setting in the BIOS? I have it disabled. I bet it will be with much more latencies
No. You wrote
So for the comparison sake it will be interesting to see screenshot for "not using MSI mode".
Just tried msi mode on Sound Blaster Zx. While it can get into msi mode it doesn't function properly. If you increase the volume is gets distorted and messes up.
Fresh windows 10 install and im sorta panicking here. Not sure what I can do to fix this big mess, so any tips would be greatly aprpeciated!
Try ticking both hdaudbus boxes in MSI_util.
Also, look in Device Manager under IRQ whether devices have been correctly switched to MSI mode.
Well the hdaudbus switched with no issues.
This is how everything looks now.
I am having issues updating the ACPI driver ( The one provided by Lenovo doesn't even start up the installation process )
And a quick snap of my Dev Mgr
Seems everything is working as intended in there
At my Win10 rig driver "acpi.sys" (device "Microsoft ACPI-Compliant System") has version "10.0.14393.0". Most probably it belongs to Win10 and there is no need in custom driver.
I take it you have a notebook. Have you checked for a latest BIOS? I don`t like notebooks due to a customization OEMs do to them and hence the need for a customized drivers.
Do you use Win10 installed by OEM or you have installed it yourself?
Thanks for the clarification on the driver.
Yes it is a notebook Lenovo Y50-70
Windows installed by myself, not OEM.
It's also running a modified bios because the original one had everything locked. So this has access to power states and such ( if the need ever arises to fiddle with anything that advanced )
Also for all it's worth
i am not sure if post #1 is still to be considered as best practice since W8.1/W10...
it shouldn't be needed to enable MSI-mode manually or to play around with it after kernel 6.2 ...
it should be fine by default. but that's just an estimate... can you elaborate on that?
i am just saying because i can imagine many people coming in this thread, having some medium-high DPC latency and just blindly following the steps in post #1...
maybe this is to be updated with info or some is obsolete?
i guess you might know!
I see that you have forced Windows to use HPET. I don`t like forcing the use of platform clock. Who knows, may be this is the culprit at your rig. After all HPET is related to ACPI.
As for P-states and C-states Windows has control on them through (hidden) power plan settings. You can switch on/off both of them (separately). I mean I would not modify BIOS for that due to possible problems.
I don`t think this whole IRQ thing is different in Windows 8/10. The kernel of Win8/10 is not that different from Win7. The mechanics if ISR-DPC are still there. Hence ISR part can be optimized by switching to MSI-mode.
If you have links on MSDN articles or tech blogs which say that interrupts are processed differently in Win8/10 then share.
Edit: I did not label post # 1 as "best practice". This whole thing is optional. It can give boost in performance and troubleshoot issues of shared IRQ.
Edit: Btw, seeing that MS own AHCI driver started to work in MSI-mode in Win10 makes me think that MSI-mode is right direction.
Disabled it, I figured I'd try both extremes.
Surprising now ( i evne had it turned off before ) things have stabilized...
yes i didn't question the use and correctness of interrupt signaling via this method.
just wanted to remind that the REG_DWORD is mostly correctly initialized for modern devices already.
(just so nobody who has DPC latency problems and comes here starts playing around with it and maybe disables MSISupported for devices, because he finds the value to be set to "1" and just tries everything out...)
people should be aware that it is very improbable that a non-touched reg->HLM\SYS\CCS\Enum.. is the source of their DPC latency "problems"