yo, hello I've been playing around with interrupt affinity policies, looking for what's the ideal configuration for my setup. (I'm into competetive fps so I want low latency) I noticed some strage things going on and thought I share my observations, maybe someone smart can help me understand... Ok, so I tried different interrupt policies for my GPU(1080ti, will post all my specs at the end) and then checked with latencymon to confirm DCPs are generated on the cores specified by the policy. Results: 1) If I chose "MachineDefault", "AllCloseProcessors", "AllProcessorsInMachine", "SpeadMessagesAcrossAllProcessors", or no policy at all(default), all DCPs are on core0. Even on full load there is no spreading of DCPs to multiple cores. 2) If I choose "SpecificProcessors" policy and select a single core, that core will handle all the DCPs -as it should. If I choose multiple cores, all the DPCs will still go to a single core. I have been told that, even on default/no policy, the DPCs should spread at least to 2 cores. What could be the problem here? The next thing I noticed is even weirder. Default policy I get spikes in latencymon when moving the mouse. USBPORT.SYS spikes to above 30us in ISR routine execution time. If I switch my GPU policy to specified processor and choose one or more specific cores(but not all cores) these spikes disappear and I have rock solid ISR routine execution time of below 5us no mater how much I move the mouse. If I choose all cores or any policy other then "specific" I get the spikes back. Tested and retested this a lot of times. So to sum up, my interrupt affinity policies don't change where the interrupts are made for the most part but at the same time very much affecting the latency spikes of my USB. No idea how to troubleshoot, any suggestions are very welcome! Here's the rest of my specs: gtx 1080ti Xeon E3-1230 v3 (4 cores) gigabyte H87-HD3 16gb ddr3 USB 2.0/ehci windows 10, 1909 Note: Both my gpu and usb controller would share IRQ port 16, but I'm using MSI mode for my graphics card, maybe that's still important detail, idk. I tested originally with 441.66 driver and now with latest, same result. Have a good day!