In Depth latency test ryzen latency(help)

So it seems that windows is loading a driver or system process that is causing high IRQL, i did a test, when windows is starting i just fast double click on In depth test latency and see what i got:

 

Interesting, I'll try that too. I suppose any Windows process running might cause higher values if it's using the CPU at the moment.

If you look at CPU usage you will see slight usage almost constantly even if your system is idling. Windows performs all sorts of crap in the background.

 

Yeah, i think probably cause svchost.exe services windows has many of them, something regardless to system services interfering with the IRQL cpu...
I made some test like disabling all extra hard drives, optical, usb, audio, leaving only the necessary, but my conclusion is something regardless with the services of windows

Now i asked in sysnative forums to track the issue with Windows Performance Tool, lets see what we got. (Cause i don't really properly know how to measure or track issues on performance tool)

Hey Asiju a question do you run Liquid Cooling on that Ryzen 5 1600? or stock fan cooler? i was wondering if you can do some test in Aida extreme 64, only with the load default settings and just Check the FPU and ran i dont know 30 minutes or 1 hour, to see in aida64 statistic in the same stability test, how raise your cpu diode temps?

Cause mine reached to 70 o.o on default.. .cpu running like 3.4 ghz some cores up to 3.6 cause performance boost, im trying to diagnose my pc comparing temps. I use only stock cooler ryzen 5 1600, and case black edition v9, just with the front chasis fan, and one in the top of the chasis and thats all. room temperature 20C

the aida test is not necessary cause yesterday was having some instability test while running 3.7 cpu and rams on xmp 2.936. so runnign all on default each category which are recommended.

 

Air cooling but not stock, Thermalright Macho 120.

Running all cores oc'd to 3.9 GHz w/ 1.312 volts.
I haven't done extended temp testing but a few consecutive runs of Cinebench show temps of about 70 ... 75 C I think.

I can retry some benchmarks at some point.

 

Nice overclock there, would be great to see your temps in a future, so take your time, dont worry, right now i'm getting ramdom blue screen with
IRQL NOT LESS OR EQUAL. which is is weird, probably something with windows 10

 

@AsiJu
I don't know if this smart or not, but by downgrading drastically my bios from 5.00 to 3.40 the problem persist then just when to the lowest Bios which is 2.50

And here are the result:
PS: also funny thing that with this bios appear C-6 state which i disabled,but idk, i always disable hpet-amd cool quiet, core performance boost, c-states, but this is the results now: windows running more than 20minut, and test only 30 seconds even with a game open

FALSE ALARM: it was cause when open a game and minimizing the cores wen all to 6 us latency probably is something with core parking or timer resolution ill check.

 

@AsiJu

I found the culprit of our issues all point to a bug that was fixed after week 24 of people that bought Ryzen first generation cpu back in 2017. mine was from may 5 2017 which sadly didn't came with the hot-fix the problem was segmentation by high cpu usage and heavy compilation codes..
The only fix was to RMA the cpu, but as i state that i live in Venezuela, then i had to find another alternative which was to disable in Bios the opcache control + SMT and finally problem is solved

Also i found that the latency will be triggered when installing any nvidia drivers, but once i disabled opcache+smt finally all latencies goes 0.00232 or 1 us to 5 or 8 maximum. The guy in amd with the threadripper probably he bough his cpu after the week 24 and thats why he doesnt have the problem

The sad part is that i have to sacrifice my 6 threads, now i have only 6

but anyway just letting u know, cheers.

 

This sounds really odd, since the fault only appears with high CPU load and you are measuring latency when CPU is idle.

 

Actually the cpu isn't on idle is pushing at 99.9% usage when doing the in depth test latency at high LVL IRQ it scaled to 99 usage cpu
The weird part is that i was playing right now counter strike 1.6 and a server ban cause of Speed hacks ._.

Do this opcache setting on my bios do something tricky lol ? xD!
I'm so unlucky the ryzen first generation after june was fixed, but mine was from may 5 2017 :/

Also i noticed that when using opcache disable, and running normal latency test, the Nvidia drivers will behave crazy at Latency lol so i fixed the interrupts which is more important for CPU at High IRQL but on latency mon the second utility seems to have more latency the kernel nvidia lolol...

 

Do you have any source information about that bug? Links?

Sounds quite odd to me as well and this is the first time I've heard of such a thing. Not saying it isn't there but strange no one has mentioned it anywhere.

@Hilbert Hagedoorn check out the above, have you ever heard of anything like this? Ie. a bug in first gen Ryzens manufactured before week 24 in 2017.

 

Yeah the problem is very real i had to turn SMT to avoid problems my start menu gets corrupted and some functions sometimes till yesterday that i found the culprit, i tried to contact amd for alternatives rather than RMA cause im from Venezuela

Here some interesting links about Segfault- bug, also this problems happens not only in Linux, Windows too
Problem is that compiling or just stressing the system the cpu with High IRQL interrupts will cause the issues

https://www.reddit.com/r/Amd/comments/72ajpw/ryzen_segfault_bug_and_actual_batches/
https://www.extremetech.com/computing/254750-amd-replaces-ryzen-cpus-users-affected-rare-linux-bug

 

The later article says Linux and R7 series only with very specific workloads. Reddit discusses all R series but still a rare bug.

I'm fairly certain this isn't the cause of high latency and as is written in Extremetech's article similar issues exist knowingly in for example Intel's sixth-gen processor family.

Disabling opcache and SMT was given as workaround as well as updating BIOS / AGESA. A BIOS or AGESA update should be able to fix it.

I certainly won't disable SMT for this and lose 6 threads because of a bug that may be there and affects only certain workloads (and has no palpaple effect to every day use).

Also worth noting In_Depth latency test records highest delay, not current or average.
Any glitch anywhere may cause a temporarily high value. IMO current or average measurement would be a lot more meaningful.

I'll retest myself with opcache disabled for the hell of it but as far as I'm concerned this is a non issue (and whatever lag you have with your system is caused by something else.
Case in point I have high IRQL latency too but my system is responsive and lag free).

 

Hi
First I want to thank you for hard work and information about SMT problem.
Recently I bought Rizen r7 1800x and it was drive me crazy because of micro stuttering in PUBG and random noise in headset.
I checked my latency with SMT and with out and everything become clear like day and night.
I found this info about some experimental software:
https://community.amd.com/thread/215217
Im testing it but not too much sukcess for the moment.
Now Im trying to make decision use with SMT off or return it.

 

Never had any stuttering in PUBG video or audio on 1800x or 2700x. I know it doesn't help you much, but maybe just maybe the problem is somewhere else.

 

I'm looking for solution for last 3 weeks its maybe my msi b350 gamer plus.
But with all my tests its coming back to latency issues.
I'm now reinstall windows 3 time. Have uefi image on usb stick fallow guideline of which way is best to install drivers.
I got get ride sound stutter and manage get better mouse response but but still in pubg even with frames above 144 I got not fluid picture.

 

As a last option try installing Process Lasso.
https://forums.guru3d.com/threads/t...-tweaking-cooling.412876/page-98#post-5585949
Helped solve performance issues on TR 2990WX by excluding core 0. Looks like a windows scheduler anomaly, since it's a non issue on Linux.

 

Tks but I think I found issues. I run memory test and it gives me errors I was try change speed but still give me errors.
That first time ever I have problem with memory. So how I will know its memory issues not motherboard?

I think reason is one of my memory module give me errors I dont get blue screens tho.

Unfortunately after getting new set of sticks latency problem persist.
I think is motherboard issue or processor.
As at the moment I remove any extra hdd sound card. I did I think all the tweks from google to fix it windows reinstal multiple times and almost aways comeback storeport issue.

Update:
I got new memory and new motherboard and problem stayed same. What I found is when I install amd ide ata/atapi drivers latency issues coming back when I dont install its looking fine no latency issues for now.

 

Hey guys sorry for abandon this thread, also the guru3d never notify about new comments, my bad.

Also Erouz don't worry we are all in the same boat it seems.

By the way I ended up doing a RMA of my Ryzen 5 1600, but, sadly the problem persist with the high IRQL , idk if most b350 or some x370 amd motherboard have this issues, cause i tried many bios, also ASROCK developers created a custom bios for me 4.92B cause my motherboard has issues, such PCI-Firmware issues by using bcdedit /set usepcifirmwaresettings Yes <-- this allow firmware to do the assignation of PCI+IRQ instead of windows, and for me when letting the firmware do the distribution or thing, that will disable my whole audio with an error or not enough resources, already confirmed with Asrock, so cause i had also Audio stuttering in all windows versions 7-8-10 etc,etc.

Also i have another issue that is HPET-BUG not being disabled at all, even by forcing off in Bios, it will keep enable the driver under device manager system category (high event precision timer) and of course even when Asrock made that bios 4.92 they manage to disable the driver, but the current firmware was bugged i mean was actually corrupted or damaged, cause windows still show enable the HPT in ACPI table in the software of RW-RW utility made for developers of motherboards.

So yeah idk this problems are very serious not everyone has high IRQL<-- and this not normal, i asked to my friend from italy the girl that work in Google she have like 7 pc , and she doesnt have the problem with hpet, nor pci firmware, but she used to have 2 pc with issues of firmware and she did RMA and everything perfect.


So yup people as sensitive as me can realize the problems of input lag visual lag, also the dirty electricity is terrible, right now im just hoping for Asrock to sent me a new motherboard, and also to get a power-conditioner ^_^

 

So what could be his low latency ? I also have 1950x with Gigabyte x399 F3G . I always get 120+ on some Cores , not all of them, some Cores stays 5-8 .. what could be the issue ?

it is said the nvidia driver causing this big jump....

https://forums.geforce.com/default/...-high-interrupt-latency-with-hyper-threading/

 

Does anyone still has updates about this? I have the 1950x and have the SMT issue.