Discussion in 'Operating Systems' started by mbk1969, May 7, 2013.
available on 1803 too:
Not "targeted IRQ", but "targeted DPC":
By default, the kernel places DPC objects at the end of the DPC queue of the processor on which the DPC was requested (typically the processor on which the ISR executed).
A device driver can override this behavior, however, ... by targeting the DPC at a particular processor. A DPC aimed at a specific CPU is known as a targeted DPC.
So device driver can override both - the priority of DPC, and the target CPU of DPC. Independently.
Than I just have not noticed it...
Yeah, my bad I wrote the wrong thing.
I do not. And people who tried different settings have not noticed any changes. Which of course does not mean that the setting was ignored. It can mean that we do not know how to estimate the change.
I will repeat: Device driver can override both - the priority of DPC, and the target CPU of DPC. Independently.
I did a little test, made two HP plans one with stock default in interrupt steering mode and one with any unparked processes.
I ran LatencyMon and looking at CPU tab some interesting differences, the any unparked process one populates much more all "ISR count" (default has all on core 0) and DPC count. I will upload pic's of the two with nothing happening on the desktop.
The default is on the right and any unparked process is on the lleft.
I guess people who tried it have not set that exact value.
Btw, do you think this is good - to spray DPCs over all cores?
I really don't know, in the stats tab some of the values are better with the any unparked process, but on the main page, both ways are good results.
I was just trying different things to help with my mouse issue in BFx games when CPU% goes high, I seem to lose some control over mouse movement. Capping FPS does help a lot and hardly happens then except in extreme heavy action periods.
EDit: probably best fix is just 6+ core cpu with HT.
unless the cores are constantly unparked, its a bad idea.
there are no parked cores ever, its a i5 3570k (4c/4th) in any of power plans.
And I always disable core parking.
Ok, I tried all the options and they seem to all work, does what it says in the description, which is best, I have no idea as my values are all good and each test run has slight variances to it.
One thing I noticed in almost all tests were USB2 ( USBPORT.SYS - USB 1.1 & 2.0 Port Driver, Microsoft Corporation) came up most times for Highest ISR routine execution time (µs) and Highest reported total ISR routine time (%).
I don't even have anything connected to USB2 ports.
Microsoft USB drivers are probably bad.
I am lucky someone (Fernando) hacked usb 3.0 extensible host controller drivers. https://www.win-raid.com/t834f25-USB-Drivers-original-and-modded.html
Otherwise i wouldnt be able to use them at all. signal cuts out often, so copying over usb , even using dualshock 4 was not possible. it is now.
Maybe its the same for you, its worth a shot to find proper ones.
It can be because USB2 controllers do not work in MSI mode - hence the lengthened ISR times.
Unfortunately there are no USB2 controllers drivers from Intel.
probably right that is the only non MSI mode device now, other than SMBus which I don't even think is used.
So overall it seems to me use processor 0 setting is very similar to default and since I don't see any kind of improvement I put it back to default. I mainly just wanted to see if it works and if it would spread across all cores.
"Maybe" if I had a latency issue I might see a difference but think driver version would be better path to fix.
I am a complete noob when it comes to this, but for me it makes sense setting only the GPU to "High" and leaving the rest on "undefined". The GPU is what takes the lead when you play any game, switching other devices doesn't make sense unless you feel you have a reason too. For example, if I worked with Audio often and that was my job, I'd set the Audio drivers to High because I would need the most of them.
EDIT: Turns out touching these settings and changing some of them to "high", like the GPU, made games worse. DmC5 started to give me random stutter (and believe me, this game is super prone to stutter, strange stuff with the RE Engine, just Google it), with the MSI modes set as Windows has them, the game is butter smooth, zero hitches, only setting them to negative number in the tool seems the way to go.
The only devices on "high" on my system are network drivers, no idea why, but that's how that driver configures itself so I leave them alone. The rest are on "undefinded", trying to change them did not give me any positive results. So don't set them to "high" believing its like a game setting or that high is faster/better/more responsive, do it just for testing and see what results you can get. I still believe is better to leave these priorities to Windows, because as I said, it didn't gave me any "boosts".
Hello i MSI moded my graphics card but want know how to make it back to default ? nvm found it out had to just re tick it btw does msi mode boost performance in games because i dont see any difference thx for the help
MSI mode does not boost performance. It can reduce the CPU load a bit, and it can help when things go wrong with some PCI devices.
So if your game experience is good you can ignore this tweak.
It can in the very narrowest of cases where a cpu is far slower than the gpu and all you're doing is crunching numbers.