I cannot believe this worked. HOURS I SPENT. DAYS I SPENT. WEEKS I SPENT. Trying to figure out why wdf01000 and nvidia drivers were causing latency on Latencymon. All I did was move the Geforce to core 1 like you recommended with the tool and FINALLY it is perfect. I cannot believe this worked. I just can't believe it.
The Affinity tool will only affect devices that have an IRQ, you can check this by sorting device manager to Resources by type. Setting affinity to a mouse will do nothing, you will need to set the affinity on the usb controller the mouse is using. If a device uses MSI-X (Most likely your storage controller) It should be set to Spread messages across processors. The biggest IRQ tweak is setting MSI mode on devices that support it, there are a few tools that will help with that and tell you if it's supported. Finally don't set affinity on a logical core (hyper threaded) it will just cause issues.
how do i know which cores hyper threaded on my pc. i’m getting high latency on latency mon from wdf and some other things and i want to use this method. i’ll on a fresh windows 10 install on new asus rog strix 18 with a 4080. hyper thread is enabled in bios but i have efficient and performance cores or something so i don’t know what i need to do for this method
0,2,4,8 etc are usually real cores. 1,3,5,7 etc are usually hyperthreads. You want to bind the interrupts to the performance cores as they will be used for running games etc and you want the interrupts to be on the same cores as the game thread. @@slowchop
I put my GPU on core 1 instead of core 0. Cyberpunk 1440p High Settings Ultra RT DLSS Quality, minimum frame-rate in the benchmark consistently stayed at 42-45 min FPS. After the change it is now at 62-64 min FPS. This is incredible, thank you! Using a 3090 and 5800x3d.
INCREDIBLY USEFUL/IMPORTANT ARTICLE REGARDING DPC LATENCY AND SAME CORES VS MULTIPLE CORES INTERRUPTS (helpdesk.flexradio What-are-DPCs-and-why-do-they-matter-) In this article he mentions how running interrupts across multiple cores comes with some disadvantages and this can be true in some instances. However, depending on the system and how it handles them can make this statement false. This is due to the fact that 1 core handling 100 different interrupt drivers is simply not feasible. Because windows handles interrupts and processes and background tasks ALL ON CORE 0 or 1 SINGLE CORE. Core to core latency can be an issue as he mentions here, but not as badly as a latency penalty of a buffered core. The main advantage to having core interrupts ALL being serviced on the same core is cache hits and misses. If your cpu notices that all your interrupts from 1 service are handled on 1 core, it will automatically have a higher chance of getting a cache hit which is significantly faster than having to fetch the info from ram. This is the reason why there is a give and take tradeoff when it comes to doing stuff like this and that's why I strongly encourage experimentation and testing for yourself. Architectures sometimes have more latency than others ESPECIALLY ryzen.
Yeah, I have the Ryzen 5800x3D which has 3D cache, so I'm wondering if this strategy might be less effective for me. But on the other hand, I would think that theoretically, if you are assigning an interrupt to a specific core, that particular interrupt would still have an equal, if not better chance, of a cache hit because it's using the same core, right? I'll read the article you posted and see what I find.
@@TheOneGhost12 ill just repost my comment i made earlier " I put my GPU on core 1 instead of core 0. Cyberpunk 1440p High Settings Ultra RT DLSS Quality, minimum frame-rate in the benchmark consistently stayed at 42-45 min FPS. After the change it is now at 62-64 min FPS. This is incredible, thank you! Using a 3090 and 5800x3d."
Chapeau bas! What a fantastic guide! I always cared about DPC latency, did many of the fixes found online. Debloating system and disabling unnessecary devices gave me a decent boost but I was not aware of this affinity tool that can leverage it even more. Very impressive. I would love to see your guide for boosting FPS in Overwatch 2. My current setup is 9900k @4.5 + 3080, playing 1080p on low and usually reaching around 400 fps (lower in midfights). Thanks!
Warning: setting affinity for the gpu pci express root port can bsod a laptop. The first time this happened, I had used bloatware removal scripts prior, due to which I wasn't able to use system restore or dsim and had to reinstall a fresh copy of Windows
Is this masking per energy profile or kind of "global"? Not everybody is using the PC instantly in High Performance Mode. Does it still work well if you are in energy saving mode? What if the cores are put to sleep? Doesn't it then need additional time to wake them up 1st?
I found a Problem with amd gpu drivers two years ago, where dpc latency Spikes to 27000 sometimes. That occurs when there is more than one monitor is connected. Im in contact with the amd Support for over 20 months and it seems like they will fix it soon.
I actually saw this bug reported A TON in the forums when I was searching for dpc latency spikes and issues related. Kind of interesting how we ran into the same bugs
How u have such low highest report dpc routine I did everything on your videos I still get 200 300 sometimes usually 70 80 is something I'm doing wrong
If we did the CPU affinity/sets technique to reserve cores for gaming and the rest for operating system like in your other video, should we also avoid setting interrupts to those "gaming cores" ?
Please don't use Latency Mon for measuring ISRs and DPC. Latency Mon often doesnt show the true values and forgets to display DPCs for certain device drivers. I would prefer Xperf. Still, nice video!
I had troubles with this problem years ago and I believed it fixed afte a Windows reinstall, now I had this trouble again after a Windows crash and its been a pain, crackling noices when playing games and watching YT videos, now its fixed thanks to this video that I basically found without looking for it, thank you very much!
Hi I have a question. So if I’m setting affinities for my gpu. Do I do it for the pci express root that it is connected to it as well? Or just the gpu itself.
bro i have followed all yoour guides, my dpc latency still spikes to 600 randomly, i am not sure why. i have been trying for over 2 weeks. resinstalled windows and everything
Savitarax please I need your help man, I want to show you a pic of latency moon (I just run it rn), and I have triple numbers on it, so what do I have to do to reduce them under 50?, I need your help man, ty btw for these tutorials
Turn off hyperthreading and turn on prefer maximum performance on NVIDIA control panel so it locks the clocks. I'm at about 35us with a 13600k and 3060 ti
I find it odd. I just have a normal installed windows. However I use the ultimate power config. But I've started watching these videos, and I'm am nowhere near any of the latencies explained here. I ran the Latencymon on my system and had 0 DPC count for CPU 0. Am I doing something wrong, Is this normal for windows what your are showing?
is it normal for average measured interrupt to process latency to go up throughout windows runtime? My averages go up by about .4us after about 2hrs of using the PC. If I restart my PC or put my PC to sleep and wake it up my averages go back to normal as if it was a fresh boot
Depends usually though I tend to notice the opposite. Where over time latency gets better not worse. You probably have a lot of tasks in task scheduler that are delayed.
Do you measure with Latencymon while idle or while playing? And what is expected in either case? Cause I'm using a 5800X3D and 2080 and getting something like 300 microseconds, while you're getting 8... And with just UA-cam open, it goes to like 150... GPU driver, though I do have 4 monitors.
@@3ildcard I am really trying to find one of Savitarax videos with a real world case benchmark, he only shows synthetic benchmarks like latencymons and other similar tools which is useless because i can't notice nanoseconds latency in the processor, the only channel showing latency tests with real life evidence that i found is Tech YES City
sup man I have a modern hybrid cpu (i9-12900k) and Im not sure if I should use the E-Cores or P-Cores for my devices. I would appreciate if you can help me out
For nvlddmkm, is there anything specific you should do for it? Is it putting setting the GPU to a different core than processors, USB, etcetera? If I assign Gpu to core 1 and processors and usb to 2 and 3, does windows then automaticially set the affinities for the remaining devices, is doing this a good solution?
that's the fastest and easiest way. Also disable threaded dpc's. There are a shit ton of articles about it on microsofts website about how to. But those are the easiest ones.
@@Savitarax Without NVIDIA drivers, my DPC latency is fantastic. Do you have any advice to stop DPC latency from nvlddmkm, I set affinity to a core, but naturally the high DPC count was moved to that core; still, DPC was then spread more across more cores. Is there anything I can do about setting affinities to reduce this DPC, or should I just experiment with different NVIDIA driver versions? Edit: With a year old driver, I was able to decrease the NVIDIA DPC latency a good bit, still, is there anything I can do to further reduce NVIDIA DPC?
I had a 9370 and it was best to treat it as a 4 core with hyper threading, putting an interrupt on 2 cores in the same module caused large latency spikes.
Another great video! What do you recommend doing for the nvmes on interrupt ? Select only the ssd or the entire pci bridge chain? Like assign only the nvme so a specific core or the entire chain? that always confused me
If you have enough cores, always do it per device. If you have limited cores do it on entire chains. The idea is that less interrupts on 1 core reduces the chances of spikes.
@@Savitarax 5800x3d here. i already have gpu/mouse on 5...used gpuaffinity..i have only the nvmes left basically. i already have pretty good input latency...trying to squeez the best i can. thanks again
At 01:22 guy looks at LatencyMon window with "CPUs" tab and does not notice that "ISR count" column has "0" in all rows. I mean - clearly LatencyMon is bugged in that regard.
I have set the affinity for tcpip service (among others) and when attempting to disable de affinity mask for that one, now it's missing on the list. Any ideas? I'm stuck at system interrumpts using 12% cpu when idle now 😵
Maybe you can help me once again. My "problematic" driver was Intel USB 3.0 controller, which i've updated to the latest version, tried enabling MSI mode (set it back to off in the end) and changed its affinity to the 2nd physical core. The affinity mask did help, but i was still getting between 80us and 110us. What ended up doing the trick was disabling C-states in the bios. Now, my USB controller driver, nvlddmkm (Nvidia), and dxgkml (DirectX) are always around 30us, and somehow i notice my pc is more snappy and responsive. Do you think i could go below that values?
I’ve actually seen it both ways. But from my experience if you just map the pcie express root port instead of the gpu it does the same thing. So it’s better to do that. But if you want to be extra you can do the gpu inside of the root port as well.
I did this exactly how you have it but I am getting bit stutter on Fortnite I did it with one core I have 8 Cores/ 16 processors You said something about setting mouse into GPU core right?
Depending on how much stress is on each core per game/ per system you can get varying results. That’s why I recommend experimenting. Core 3/7 for me give weird issues for no reason.
@@Savitarax oh, 👌 also I just ran LatencyMon on my desktops PC windows 11 shows One problem may be related to power management, disable CPU throttling settings in Control Panel and BIOS setup. Check for BIOS I’m not sure what this is all about just got myself prebuilt PC from IBUYPOWER two years ago everything works great just having issues with system response time I want to have smooth gaming experience with less stutter because let’s be honest low latency must for competitive gaming my PC motherboard Asrock z590c/ac graphic card RTX 3070 Intel Core i7-11700K someone told me upgrade to a better motherboard. I’m still debating on it…. Any tips would be greatly appreciate it 🙏I forgot to mention LatencyMon says Wdf01000.sys 3347ms latency the bar highlighted RED
@@syedm.sameer2078 Not really, I've tried a lot of things. Tried Windows 7, 8.1, 10 and 11 on my system, 10 worked the best in terms of latency surprisingly. I know at least with my laptop, it's a Dell issue and there's nothing that can be done, no idea about yours though.
@@syedm.sameer2078 yeah I've had the issue on both my Precision 7530 and 3520. At least on the 7530 Windows 10 was best, but in your case that may be different. Btw I let windows update handle the drivers and only installed the nvidia driver manually.
ntoskrnl.exe, should i bind this one to one core or leave it on all cores? And if i should, how do i do it? What should i look for in interrupt affinitypolicy configuration tool?
Single core offers the lowest latency, ideally you just “set mask” and select a core. You want to have few spikes and the ones you do have are infrequent.
I changed my intel usb root hub to a single core, and after restarting none of my usb devices were functioning. I had to force boot into recovery tools, then a system restore point to enable functionality. Any suggestions?
Interesting. We’re you happening to be binding that to a thread and not a core. Cause interrupts work regardless of whichever core. But threads are not cores so binding an interrupt won’t work and may. In your case. Break the device
Well yah cause the driver is handling way more than just input lag and doesn’t have the actual Input thread going through it. That’s CSRSS.EXE’s job. It’s handling everything else on that entire system. The main thing about DPC latency spikes that are useful is they tell you how overloaded a process is in a queue. If the nvidia driver has a dpc latency spike in the thousands of microseconds. Yah it might not cause input lag. But it’ll cause a crap ton else like audio cutouts,stutters,other issues. DPC latency is a great indicator of generalized stability not necessarily speed. Hence why you can get stupidly low averages like .4us. But still have input lag/stutters/FPS issues etc. I do like Fr33thys stuffs and a lot of other UA-camrs. This is also why DPC latency isn’t talked about nearly as much as FPS/Frametimes etc. cause it’s not as indicative of the others but it still matters.
This is pretty much snake oil on a modern machine. This only helps old core architectures or poor quality RAM, and basic hardware inefficiencies etc. Or you're an audio engineer. Other than that? Modern Windows and hardware properly schedule everything, especially if you're already doing basic core affinities or CPU and/or I/O priorities.
it's mostly cause it's a driver that handles so many other parts of the system. It's one that has a lot going on so if it has unoptimized parts in it. It can cause tons of issues down the line.
Leave it to gamers to solve these issues! Im trying to optimize my pc for audio, and this was helpful
I cannot believe this worked. HOURS I SPENT. DAYS I SPENT. WEEKS I SPENT. Trying to figure out why wdf01000 and nvidia drivers were causing latency on Latencymon. All I did was move the Geforce to core 1 like you recommended with the tool and FINALLY it is perfect. I cannot believe this worked. I just can't believe it.
Salut comment avez vous fait je ne comprends pas bien la vidéo merci j’ai le même soucis
so what was "wdf01000" ?
i recon it peaks when i move my mouse..what was the solution?
Also same problem wdf01000.sys high latency always so u just set the gpu driver to core 1?that's it?
Bro to solve wdf01000
Which file did you moved to another core?
If you were brought here trying to get the Apollo work with Windows, don’t bother.
your videos have been like a school classe for me man. the channell i ve always wished for haha frfr hats off bruh
Happy to help out man 😃, I like to give the information cause for me it took so long to get an accurate answer that actually worked for my problems.
The Affinity tool will only affect devices that have an IRQ, you can check this by sorting device manager to Resources by type.
Setting affinity to a mouse will do nothing, you will need to set the affinity on the usb controller the mouse is using.
If a device uses MSI-X (Most likely your storage controller) It should be set to Spread messages across processors.
The biggest IRQ tweak is setting MSI mode on devices that support it, there are a few tools that will help with that and tell you if it's supported.
Finally don't set affinity on a logical core (hyper threaded) it will just cause issues.
how do i know which cores hyper threaded on my pc. i’m getting high latency on latency mon from wdf and some other things and i want to use this method. i’ll on a fresh windows 10 install on new asus rog strix 18 with a 4080. hyper thread is enabled in bios but i have efficient and performance cores or something so i don’t know what i need to do for this method
0,2,4,8 etc are usually real cores. 1,3,5,7 etc are usually hyperthreads.
You want to bind the interrupts to the performance cores as they will be used for running games etc and you want the interrupts to be on the same cores as the game thread.
@@slowchop
@@TheWomble001 so bind the interrupts to real cores not the hyper thread ones?
@@TheWomble001 and do i need to disable hyper threading? i looked it up i have 8 performance cores and 16 efficiency cores.
Yes put the interrupts on real cores.
@@slowchop
Thank you bro keep pumping out these videos
I put my GPU on core 1 instead of core 0. Cyberpunk 1440p High Settings Ultra RT DLSS Quality, minimum frame-rate in the benchmark consistently stayed at 42-45 min FPS. After the change it is now at 62-64 min FPS. This is incredible, thank you! Using a 3090 and 5800x3d.
HOW DO I PUT IT ON A CERTAIN CORE? any tutorials?
dont yo use process lasso? you can just exclude cpu 0 there
INCREDIBLY USEFUL/IMPORTANT ARTICLE REGARDING DPC LATENCY AND SAME CORES VS MULTIPLE CORES INTERRUPTS
(helpdesk.flexradio What-are-DPCs-and-why-do-they-matter-)
In this article he mentions how running interrupts across multiple cores comes with some disadvantages and this can be true in some instances.
However, depending on the system and how it handles them can make this statement false.
This is due to the fact that 1 core handling 100 different interrupt drivers is simply not feasible. Because windows handles interrupts and processes and background tasks ALL ON CORE 0 or 1 SINGLE CORE.
Core to core latency can be an issue as he mentions here, but not as badly as a latency penalty of a buffered core.
The main advantage to having core interrupts ALL being serviced on the same core is cache hits and misses.
If your cpu notices that all your interrupts from 1 service are handled on 1 core, it will automatically have a higher chance of getting a cache hit which is significantly faster than having to fetch the info from ram. This is the reason why there is a give and take tradeoff when it comes to doing stuff like this and that's why I strongly encourage experimentation and testing for yourself.
Architectures sometimes have more latency than others ESPECIALLY ryzen.
Yeah, I have the Ryzen 5800x3D which has 3D cache, so I'm wondering if this strategy might be less effective for me. But on the other hand, I would think that theoretically, if you are assigning an interrupt to a specific core, that particular interrupt would still have an equal, if not better chance, of a cache hit because it's using the same core, right? I'll read the article you posted and see what I find.
@@lofigamervibessame you find out anything????
@@TheOneGhost12 ill just repost my comment i made earlier " I put my GPU on core 1 instead of core 0. Cyberpunk 1440p High Settings Ultra RT DLSS Quality, minimum frame-rate in the benchmark consistently stayed at 42-45 min FPS. After the change it is now at 62-64 min FPS. This is incredible, thank you! Using a 3090 and 5800x3d."
@@germanmade1219have u tried setting it to other core and if the result is similar?
Chapeau bas! What a fantastic guide! I always cared about DPC latency, did many of the fixes found online. Debloating system and disabling unnessecary devices gave me a decent boost but I was not aware of this affinity tool that can leverage it even more. Very impressive. I would love to see your guide for boosting FPS in Overwatch 2. My current setup is 9900k @4.5 + 3080, playing 1080p on low and usually reaching around 400 fps (lower in midfights). Thanks!
you should overclock it to 5ghz
@@agsechogd6406 I did it a few months ago already :) but thanks!
I tried this for my mouse and it felt snappier and felt like a faster response time to me thank you.
Nice! On which core did you put your gpu , if you dont mind me asking fam! Thanks in advance
your latency is incredible :0
You rock solid ! thanks for that video first time i saw such details not the same BS : "update your driver & disable wifi"
Literally half the reason I make these guides is because everyone legit had nothing new to say or explain. It’s always random ass forum posts XD
@@Savitarax True mate really apreciate thank you ...
Man that's a whole lot of work to fix my latency issues. They really need to do something about the interrupt scheduler.
Warning: setting affinity for the gpu pci express root port can bsod a laptop. The first time this happened, I had used bloatware removal scripts prior, due to which I wasn't able to use system restore or dsim and had to reinstall a fresh copy of Windows
Pci doesn't matter you need to set GPU affinity that won't cause anything
So this thing doesn't work on laptops ?
Is this masking per energy profile or kind of "global"?
Not everybody is using the PC instantly in High Performance Mode.
Does it still work well if you are in energy saving mode?
What if the cores are put to sleep? Doesn't it then need additional time to wake them up 1st?
I found a Problem with amd gpu drivers two years ago, where dpc latency Spikes to 27000 sometimes. That occurs when there is more than one monitor is connected. Im in contact with the amd Support for over 20 months and it seems like they will fix it soon.
I actually saw this bug reported A TON in the forums when I was searching for dpc latency spikes and issues related.
Kind of interesting how we ran into the same bugs
Just came across your video. When I run latency mon. It keeps saying I'm having issues with APCI. How would I go about fixing this??
Im gonna try set every driver off of core 0, that way I can free up a core so OW runs on all 4 of my physical cores.
Did it work well?
How u have such low highest report dpc routine I did everything on your videos I still get 200 300 sometimes usually 70 80 is something I'm doing wrong
Thanks Brotha This Helped Alot
If we did the CPU affinity/sets technique to reserve cores for gaming and the rest for operating system like in your other video, should we also avoid setting interrupts to those "gaming cores" ?
So it’s okay to have interrupts on game cores because the os handles it differently.
Stacking interrupts, is the main issue.
@@Savitarax okay thanks :)
What a fantastic guide
Hey, man keep of the good work I hope your doing well friend!
Amazing info, the best solution after lurking arround in internet for weeks. Thank you!
😃happy to help however I can.
Please don't use Latency Mon for measuring ISRs and DPC. Latency Mon often doesnt show the true values and forgets to display DPCs for certain device drivers. I would prefer Xperf. Still, nice video!
I had troubles with this problem years ago and I believed it fixed afte a Windows reinstall, now I had this trouble again after a Windows crash and its been a pain, crackling noices when playing games and watching YT videos, now its fixed thanks to this video that I basically found without looking for it, thank you very much!
Hi I have a question. So if I’m setting affinities for my gpu. Do I do it for the pci express root that it is connected to it as well? Or just the gpu itself.
i notice u disable something in device manager like
numeric data processor or smbus and etc
does it benefit something?
can anyone help me with affinity. my gpu going 1500-2000ms
i have noticed you disable numeric data processor why?
bro i have followed all yoour guides, my dpc latency still spikes to 600 randomly, i am not sure why. i have been trying for over 2 weeks. resinstalled windows and everything
saame its so annoying, i have done so many latency configurations and it stilll randomly spikes, have you found a fix
Does it matter what cores you put certain devices on? For example GPU needs to be on a certain number core?
so i have a i36100 do u think i can do this? i only have 2cores 4 threads. love your videos btw learning a lot.
Savitarax please I need your help man, I want to show you a pic of latency moon (I just run it rn), and I have triple numbers on it, so what do I have to do to reduce them under 50?, I need your help man, ty btw for these tutorials
Turn off hyperthreading and turn on prefer maximum performance on NVIDIA control panel so it locks the clocks. I'm at about 35us with a 13600k and 3060 ti
I find it odd. I just have a normal installed windows. However I use the ultimate power config. But I've started watching these videos, and I'm am nowhere near any of the latencies explained here. I ran the Latencymon on my system and had 0 DPC count for CPU 0. Am I doing something wrong, Is this normal for windows what your are showing?
I didnt understand a single thing the only thing i ever did in interrupt-affinity as set the controller to my mouse to all the even cores
is it normal for average measured interrupt to process latency to go up throughout windows runtime? My averages go up by about .4us after about 2hrs of using the PC. If I restart my PC or put my PC to sleep and wake it up my averages go back to normal as if it was a fresh boot
Depends usually though I tend to notice the opposite. Where over time latency gets better not worse. You probably have a lot of tasks in task scheduler that are delayed.
HOW DO I Get this mines always high especially the nvidia portion
Do you measure with Latencymon while idle or while playing? And what is expected in either case? Cause I'm using a 5800X3D and 2080 and getting something like 300 microseconds, while you're getting 8... And with just UA-cam open, it goes to like 150... GPU driver, though I do have 4 monitors.
Get a custom OS, Look at his custom OS showdown video
@@dioklepper6367Yup.
ReviOS. Try it. Best latency you can get for gaming.
@@3ildcard I am really trying to find one of Savitarax videos with a real world case benchmark, he only shows synthetic benchmarks like latencymons and other similar tools which is useless because i can't notice nanoseconds latency in the processor, the only channel showing latency tests with real life evidence that i found is Tech YES City
You think it is better to not let the game on core 0 as well in lasso? or does this better work if together.
Can soneone post download link to the interrupt affinity policy program. Pls I could not find it
sup man I have a modern hybrid cpu (i9-12900k) and Im not sure if I should use the E-Cores or P-Cores for my devices. I would appreciate if you can help me out
usse p cores for high perfamnce sstuff
@@Mstfamstr gpu = p-core? What about an audio interface, ethernet-adapter and mouse?
same question, what about usb devices? should it be on e-core?
For nvlddmkm, is there anything specific you should do for it? Is it putting setting the GPU to a different core than processors, USB, etcetera? If I assign Gpu to core 1 and processors and usb to 2 and 3, does windows then automaticially set the affinities for the remaining devices, is doing this a good solution?
that's the fastest and easiest way. Also disable threaded dpc's.
There are a shit ton of articles about it on microsofts website about how to. But those are the easiest ones.
@@Savitarax Without NVIDIA drivers, my DPC latency is fantastic. Do you have any advice to stop DPC latency from nvlddmkm, I set affinity to a core, but naturally the high DPC count was moved to that core; still, DPC was then spread more across more cores. Is there anything I can do about setting affinities to reduce this DPC, or should I just experiment with different NVIDIA driver versions? Edit: With a year old driver, I was able to decrease the NVIDIA DPC latency a good bit, still, is there anything I can do to further reduce NVIDIA DPC?
So, having an FX cpu can be beneficial because of amount of cores it has and potential to spread interrupts?
I had a 9370 and it was best to treat it as a 4 core with hyper threading, putting an interrupt on 2 cores in the same module caused large latency spikes.
Hi does this apply to windows 11 ?
ok and how the fk do u set it to a certain core?
Another great video! What do you recommend doing for the nvmes on interrupt ? Select only the ssd or the entire pci bridge chain? Like assign only the nvme so a specific core or the entire chain? that always confused me
If you have enough cores, always do it per device.
If you have limited cores do it on entire chains.
The idea is that less interrupts on 1 core reduces the chances of spikes.
@@Savitarax 5800x3d here. i already have gpu/mouse on 5...used gpuaffinity..i have only the nvmes left basically. i already have pretty good input latency...trying to squeez the best i can. thanks again
I turned off hpet from windows and enabled bios. Do you think I should choose single core or dual core while doing these steps?
At 01:22 guy looks at LatencyMon window with "CPUs" tab and does not notice that "ISR count" column has "0" in all rows. I mean - clearly LatencyMon is bugged in that regard.
I have set the affinity for tcpip service (among others) and when attempting to disable de affinity mask for that one, now it's missing on the list. Any ideas?
I'm stuck at system interrumpts using 12% cpu when idle now 😵
Once you delete the mask you have to reenable the affinity.
@@Savitarax Thanks for the fast response. Gonna try
Maybe you can help me once again.
My "problematic" driver was Intel USB 3.0 controller, which i've updated to the latest version, tried enabling MSI mode (set it back to off in the end) and changed its affinity to the 2nd physical core. The affinity mask did help, but i was still getting between 80us and 110us. What ended up doing the trick was disabling C-states in the bios.
Now, my USB controller driver, nvlddmkm (Nvidia), and dxgkml (DirectX) are always around 30us, and somehow i notice my pc is more snappy and responsive.
Do you think i could go below that values?
If I have a spare core that's being unused, would it be good to assign a core to the gpu and then assign another core to the generic pnp monitor?
Nah generic monitor is just controlled by your GPU.
If I reinstall a driver, lets say Nvidia drivers, do I have to re-do the interrupt policy tweaks again?
I’ve actually seen it both ways. But from my experience if you just map the pcie express root port instead of the gpu it does the same thing.
So it’s better to do that.
But if you want to be extra you can do the gpu inside of the root port as well.
I did this exactly how you have it but I am getting bit stutter on Fortnite I did it with one core I have 8 Cores/ 16 processors You said something about setting mouse into GPU core right?
Depending on how much stress is on each core per game/ per system you can get varying results.
That’s why I recommend experimenting.
Core 3/7 for me give weird issues for no reason.
@@Savitarax oh, 👌 also I just ran LatencyMon on my desktops PC windows 11 shows One problem may be related to power management, disable CPU throttling settings in Control Panel and BIOS setup. Check for BIOS I’m not sure what this is all about just got myself prebuilt PC from IBUYPOWER two years ago everything works great just having issues with system response time I want to have smooth gaming experience with less stutter because let’s be honest low latency must for competitive gaming my PC motherboard Asrock z590c/ac graphic card RTX 3070 Intel Core i7-11700K someone told me upgrade to a better motherboard. I’m still debating on it…. Any tips would be greatly appreciate it 🙏I forgot to mention LatencyMon says Wdf01000.sys 3347ms latency the bar highlighted RED
You prefer the Windows 10 over Windows 11 ?
Win 10, win11 is still missing so many features and has tons of bugs
Hey savitrax my high reported dpc was from ACPI.SYS which driver is that?
It manages power management, specifically laptops tend to have high latency with that driver as it handles power management iirc.
so do you know how can i reduce the dpc latency coming from those drivers@@whitebeartigtig
@@syedm.sameer2078 Not really, I've tried a lot of things. Tried Windows 7, 8.1, 10 and 11 on my system, 10 worked the best in terms of latency surprisingly. I know at least with my laptop, it's a Dell issue and there's nothing that can be done, no idea about yours though.
@@whitebeartigtig yup I have a latitude dell 5490 , Windows 10 I should try next?
@@syedm.sameer2078 yeah I've had the issue on both my Precision 7530 and 3520. At least on the 7530 Windows 10 was best, but in your case that may be different. Btw I let windows update handle the drivers and only installed the nvidia driver manually.
What’s dcp
my highest reported DPC is Nvidia windows kernel mode driver what should i do please ?
Move it to a core that isn’t busy
you mean the gpu right ? can i put my sata to core 0 ?@@Savitarax
it fixed nvidia now its Wdf01000.sys Kernel Mode Driver Framework Runtime what is that ? please answer
@@idmas991 USB ports.
So try moving the “XHCI host controllers” around
What driver is associated with Nvidia windows kernel mode?
ntoskrnl.exe, should i bind this one to one core or leave it on all cores? And if i should, how do i do it? What should i look for in interrupt affinitypolicy configuration tool?
Single core offers the lowest latency, ideally you just “set mask” and select a core.
You want to have few spikes and the ones you do have are infrequent.
@@Savitarax ok, thank you, i'm just having troubles identifying the device in device manager thats responsible for ntoskrnl.exe driver
@@aabb-sx4js that is mostly the usb hub for NTOS
@@Savitarax where would i find that
Joelle Brook
спасибо, ты мне помог
@Savitarax how do you choose which core to choose for which task on ryzen 7 7800x3d?
Legros Keys
I changed my intel usb root hub to a single core, and after restarting none of my usb devices were functioning. I had to force boot into recovery tools, then a system restore point to enable functionality. Any suggestions?
Interesting. We’re you happening to be binding that to a thread and not a core. Cause interrupts work regardless of whichever core. But threads are not cores so binding an interrupt won’t work and may. In your case. Break the device
@@Savitarax I'm using a 9700k, so there is no hyperthreading. Only cores. Core 07 should represent the 8th core.
I am pretty sure FR33THY debunked the whole driver latency thing. It showed no difference in input lag via his tool.
Well yah cause the driver is handling way more than just input lag and doesn’t have the actual Input thread going through it.
That’s CSRSS.EXE’s job.
It’s handling everything else on that entire system.
The main thing about DPC latency spikes that are useful is they tell you how overloaded a process is in a queue.
If the nvidia driver has a dpc latency spike in the thousands of microseconds.
Yah it might not cause input lag. But it’ll cause a crap ton else like audio cutouts,stutters,other issues.
DPC latency is a great indicator of generalized stability not necessarily speed.
Hence why you can get stupidly low averages like .4us.
But still have input lag/stutters/FPS issues etc.
I do like Fr33thys stuffs and a lot of other UA-camrs.
This is also why DPC latency isn’t talked about nearly as much as FPS/Frametimes etc.
cause it’s not as indicative of the others but it still matters.
@@Savitarax FR33THY's knowledge is highly limited.
836 Elsie Center
Im hungry for somethin New in Mouse/keyboard inputlag World , Give it to me🤪
Oceane Trafficway
This is pretty much snake oil on a modern machine.
This only helps old core architectures or poor quality RAM, and basic hardware inefficiencies etc.
Or you're an audio engineer.
Other than that? Modern Windows and hardware properly schedule everything, especially if you're already doing basic core affinities or CPU and/or I/O priorities.
My CPU 0 has a dpc count of 140656 with CPU 0 with dxgkrnl.sys having the highest execution time. what do you recommend I do?
Lol not have your Gpu on core 0 is a start and then your usb devices like mouse on another core.
Can anyone tell me what dxgkrnl.sys dirextx graphics kernel is so highest reported dpc routine??
i need to know how to stop it
win32priorityseparation 28 or 2a for lowest input delay
just put 2a and forget about it, other things help more
@@inputvol.280 tq bro
Why does NVLDDMKM.SYS have so much DPC latency?
it's mostly cause it's a driver that handles so many other parts of the system.
It's one that has a lot going on so if it has unoptimized parts in it. It can cause tons of issues down the line.
@@Savitarax any fixes?
@@ramsaybolton9151 did you found anything usefull?
@@Savitarax any fix please ?
6762 Aaron Turnpike
9455 Rylee Knolls
can you help me figure it out nvlddmkm.sys always pop out, spike so high than any system. I have 11 CPU, My Problem only about that :(
Thank you
83934 Aliya Overpass