no they aren't. individual computers are. computers bogged down by jank drivers and with too much crap on the system bus.
Well it's back again and indeed repeatedly the highest reported DPC routine execution time is by ntoskrnl.exe What can be done about that? Yeah I read track down the culprit driver via Windows Performance Analyzer but what can I track down if ntoskrnl.exe is taking too long to execute DPC routines? The tracking can be done via LatencyMon directly just as well, as far as I can tell: (above is just for reference, no issues there but ntoskrnl.exe will hit between 900 ... 1200+ consistently) Is any recent KB update known to have such issues? Btw I also tried enabling HAGS and it seemed to make things worse, interrupt to process latency was consistently higher. Having said all that I do like having every PCI device in MSI mode:
the major difference between WPA and LatencyMon is that WPA can be expanded right down to the function taking too long within the driver. just to say, thats two hundred fifty eight <point> seven seven zero, not 250 thousand, which would have thrown a bsod by the point it reached 100k, the kernel might have spun up a core from parking or sleep state to perform a task. I can't say yay or nay, i haven't seen anything that would indicate as such but system vs system could chang everything.
All right, I'll try WPA too. Yeah I know, and that image too was just for reference. 258 is perfectly fine as a result. It's the constant and near-instant spikes to 900+ that worry me (in combination with actually having sound and input issues, even if sporadically).
Why would this show up for me? When I search the kb number it says that it's a microcode update for intel cpu's, or is it also valid for ryzen? https://support.microsoft.com/en-us...and-20h2-9f8641b3-bd8e-6fe7-ab4d-4603847c4030 Out of curiosity are you using the audio drivers that are made for your motherboard or are you using "newer" audio drivers obtained from elsewhere?
I have disabled on board audio from BIOS. Using Creative's newest driver for Sound BlasterX AE-5. Btw was wondering the same about that KB update but I guess all systems get them.
I don't know they have certainly bad reputation. It is recurring thing on the internet. While I did not heard DPC latency complains about AMD. AMD users saying never had problem with DPC latency before switching to Nvidia. https://www.overclock.net/threads/nvidia-rtx-dpc-latency-fixed.1736362/ On sound production forums they also don't like Nvidias. But I found only forums posts, don't know if anyone did tests about this, to shed more light on issue. So I have only anecdotes... Also Nvidia has crappy drivers lately!
there are as many amd boards with dpc issues as there are as intel board with dpc issues there aren't as many radeon gpu's with dpc issues because there are more nvidia boards in the user market (and the amd driver doesn't do quite as much with multithreaded cpu's), DPC is a system wide thing to optimize for, and most of the cases i have come across have been hardware level issues with a component that affects the pcie bus and results in error corrections, or signalling misbehaviors such as where a pcie slot actually has defective lanes or the board has just been jammed with so many features and left turned on that putting other devices in suffer from the already saturated queues. other times theres been a out of date vendor utility driver that exploited a flaw in windows that got patched and broke the driver. In the past there was a range of bulldozer centric motherboards from a vendor that had horrendous dpc without any display driver involvement because they were using an interrupt based digital vrm that was rapidly firing interrupts to micromanage the voltage, you could get a utility that completely disabled that but it wasn't heavily publicized. It's been very rare where a driver actually inherently had a user agnostic bug that resulted in DPC latencies operating above 500us, and its generally normal for a multicore cpu with C states enabled to run the nvidia driver in legacy irq's at 150-250us. People get their knickers in a twist when latencymon shows that any driver suddenly spiked, without factoring in what the driver was doing at the time. any time a device changes power state it'll jump (blocking interrupt). The author has gone out of his way to point out that its only intended to be used on an idle system (but good luck with that on windows 10)
Make sure you are not signing in online with a MS account for Windows 10. Instead use a local account. MS machine learning is broken.
Just got KB5001567 via optional but chose not to install... anyone else get this today? Looked it up on MS Update Catalog and there's zero info on what it even is: https://www.catalog.update.microsoft.com/Search.aspx?q=KB5001567 edit: nevermind, support articles just popped up
KB5001567 is an "out-of-band" update that fixes a blue screen problem with the previous KB5000802 update when printing to certain printers. that's what happens when M$ fired or got rid of their old Windows OS QA team
'" I serisouly dont think that have 1, I probably only one that that that though. Q/A is probably 2 people in room in there MS Dungeon/Basement
On build 21H1 19043.899 and this update seems to give a newer AMD driver. Version 27.20.14537.1001 Still no HAGS support for 6000 series though.
Does HAGS even matter? last i heard it cause problems being on for lot people like crash. I might be on 20h2 but it was disable by upon installation and still is have and probably wont enable it manual either.
Not too sure, I did have problems with it on my old 1080Ti but what from what I have heard/read online is that it seems to work better with AMD hardware. Some users of older AMD cards (500 and 5000 series) who have managed to get it working have reported gains.
NVIDIA boards wait what??? Yeah it is system-wide, other bottlenecks can worsen DPC latency issues sure! Ye it varies per motherboard greatly, I got ASUS Z390-i gaming on which I have maximum interrupt to process latency 100us as advertised. But I had RTX 2070 on the same mobo and I had latency spikes to 600us. They destroyed it in RMA, so I had to order new RTX 3070. Now I have 50-70us max from multiple 30 minute long tests, nothing is running on a background! Also on my old computer I had same spikes to 600us, which had cheap ass motherboard, but it had actually relatively low user to process latency like 250max. What components could affect PCI-E bus except GPU? If you don't use any other PCI-E cards? PCI-E cards in general can cause huge DPC latency!!! I don't think you can ascribe this issue only to defective lanes, or hardware level issues. Also you didn't mention that: main source of DPC latency in general are badly coded drivers. And NVIDIA drivers are hot pile of steaming crap lately, plus no one really cares about DPC latency in the industry. Who knows it could be NVIDIA drivers, I wouldn't certainly reject this idea that it can be crappy drivers... Yeah many motherboards still have DPC latency issues, take ASUS hero 500$ motherboards for example, they suffer even from 1ms+ DPC latency with no fix! DPC latency is underestimated... Devs really don't care much unless it is completely overboard and sometimes even then... Yeah latencymon isn't perfect, because Windows is working system and you can't really test it without Windows... Even with disabled C states, ultimate power profile, disable idle saver, no turbo, no speed step, bclk 100 + everything closed, something may happen, which could cause spike. I also measured same spikes using WPR. It would be interesting to analyze what it was doing at that time. I don't know if during latencymon test, GPU: frequency/voltage is stable. But I don't have any spikes anymore. So I can't test it anyways... Tho this is the thing: now I have 50-70us no matter what, I did multiple long 30 minute tests and no spikes anymore, seems uncanny, if it was freq/voltage spike. Also it can't be faulty PCI-E lane then. Dunno still... But interesting post! Hey BTW did you heard about these advanced tweaks for DPC latency? ctrl+ f to: Tell Windows to stop tolerating high DPC/ISR latencies. Source: https://sites.google.com/view/melodystweaks/basictweaks Google can't even find these registry entries and what they do, LOL basic tweaks...
Yes it does. It will be the basis for a lot of things in the future for the Windows graphics subsystem. It's not really a performance upgrade (that might come as a side effect in some scenarios). AMD should support it, and it could support it back to the 7970 if they wanted to.