Another look at HPET High Precision Event Timer

I've found what was causing Lag input for me. could fix many micro stuttering or freezes for people if you know where to look. It happened also for my Intel 3770K. the problem is within the HT or SMT.

https://www.resplendence.com/latencymon_cpustalls

After testing again.

The most useful lowered my Lag input was

EHCI Hand-off - Disabled
XHCI Hand-off - Disabled

I don't have XHCI mode on my Gigabyte X399 aurous gaming 7 mobo. so USB 3.0 still active.

But also I ran a test using In-Depth Latency Tests comes with LatencyMon and I've found out my Intel 3770K and AMD 1950X Both has a serious issue with High-IRQ SMI request. and this issue comes using HT or SMT, disabling them fix the issue and fix my mouse lag input. The real fix was Hot-Fix they released after the first CPU and you needed to RMA it. So CPUs have serious issues before anything else.

You can see

 

I see no point in tests without hardware interrupts - we do not game without interrupts from video-, audio-, network-, USB- controllers.
I made a test - 3 times without hyperthreading, 3 times with - but namely with enabled hardware interrupts.
I toggled HT in BIOS, booted launched IDLT and browser, started the monitoring in IDLT, opened youtube site and started video playback, stopped the monitoring after 1 minute.

Code:

1.
REPORT
In Depth Latency Tests has been analyzing your system for  0:01:08  (h:mm:ss) on all processors in the system.
PER CPU MEASURINGS
Highest latency on CPU 0:      598,0 µs
Highest latency on CPU 1:      4,60 µs
Highest latency on CPU 2:      4,20 µs
Highest latency on CPU 3:      4,60 µs
Highest latency on CPU 4:      4,40 µs
Highest latency on CPU 5:      4,40 µs
2.
REPORT
In Depth Latency Tests has been analyzing your system for  0:01:11  (h:mm:ss) on all processors in the system.
PER CPU MEASURINGS
Highest latency on CPU 0:      357,30 µs
Highest latency on CPU 1:      139,70 µs
Highest latency on CPU 2:      128,60 µs
Highest latency on CPU 3:      139,50 µs
Highest latency on CPU 4:      140,50 µs
Highest latency on CPU 5:      139,10 µs
Highest latency on CPU 6:      112,0 µs
Highest latency on CPU 7:      115,60 µs
Highest latency on CPU 8:      135,60 µs
Highest latency on CPU 9:      125,10 µs
Highest latency on CPU 10:      137,70 µs
Highest latency on CPU 11:      118,60 µs
3.
REPORT
In Depth Latency Tests has been analyzing your system for  0:01:14  (h:mm:ss) on all processors in the system.
PER CPU MEASURINGS
Highest latency on CPU 0:      453,90 µs
Highest latency on CPU 1:      76,90 µs
Highest latency on CPU 2:      76,90 µs
Highest latency on CPU 3:      76,80 µs
Highest latency on CPU 4:      76,80 µs
Highest latency on CPU 5:      4,40 µs
4.
REPORT
In Depth Latency Tests has been analyzing your system for  0:01:02  (h:mm:ss) on all processors in the system.
PER CPU MEASURINGS
Highest latency on CPU 0:      474,90 µs
Highest latency on CPU 1:      144,20 µs
Highest latency on CPU 2:      137,0 µs
Highest latency on CPU 3:      145,80 µs
Highest latency on CPU 4:      144,60 µs
Highest latency on CPU 5:      133,60 µs
Highest latency on CPU 6:      140,60 µs
Highest latency on CPU 7:      143,70 µs
Highest latency on CPU 8:      124,20 µs
Highest latency on CPU 9:      118,90 µs
Highest latency on CPU 10:      110,50 µs
Highest latency on CPU 11:      135,0 µs
5.
REPORT
In Depth Latency Tests has been analyzing your system for  0:01:03  (h:mm:ss) on all processors in the system.
PER CPU MEASURINGS
Highest latency on CPU 0:      292,30 µs
Highest latency on CPU 1:      4,70 µs
Highest latency on CPU 2:      4,90 µs
Highest latency on CPU 3:      4,80 µs
Highest latency on CPU 4:      5,0 µs
Highest latency on CPU 5:      4,20 µs
6.
REPORT
In Depth Latency Tests has been analyzing your system for  0:01:06  (h:mm:ss) on all processors in the system.
PER CPU MEASURINGS
Highest latency on CPU 0:      482,0 µs
Highest latency on CPU 1:      140,60 µs
Highest latency on CPU 2:      145,10 µs
Highest latency on CPU 3:      135,20 µs
Highest latency on CPU 4:      135,70 µs
Highest latency on CPU 5:      140,70 µs
Highest latency on CPU 6:      117,80 µs
Highest latency on CPU 7:      111,40 µs
Highest latency on CPU 8:      130,0 µs
Highest latency on CPU 9:      135,0 µs
Highest latency on CPU 10:      131,10 µs
Highest latency on CPU 11:      114,20 µs

We can see a pattern - CPU 0 is busy with interrupts from hardware. And measured latency for that CPU is pretty much the same in all tests.

 

Thanks, this test showed me I have high latency, it's including USB which had 500 ms or more. after disabling HT, my USB driver was less than 10ms. so it effects all hardware and software.
My USB driver in windows checking with Windows Performance analyzer.
Also, do you have Secure Boot enabled or Disabled with removed keys? because it effect the mouse a lot. but still testing because I had SR-IOV enabled also but it shouldn't enabled if IOMMU disabled.

 

Yes, Secure Boot is enabled.

SR-IOV is relevant only for virtual machines as I take it. Do you run your Windows in virtual machine?

How do you disable IOMMU?

 

HEPT will not affect game performance. The frame rate difference is result from more accurate clock and the actual frame output to the display is still the same. So you can not compare frame rate from pure software with different clock sources. You need FCAT.

 

I run windows normally.

AMD bios have the option to disable IOMMU and there is a driver when installing the AMD chipset IOMMU Null on device manager

 

neither of these should be turned on with a PNP operating system (same as pnp handoffs).

on a PNP OS (and windows especially), the driver always assumes control of the usb controller

XHCI Enabled is required for Windows 7 where third party usb3 drivers was required.

 

Something weird with LatencyMon for me. anyone have idea what could it be? it doesn't use CPU0 at all anymore when I enable SMT/HT. When I don't use SMT/HT it uses CPU0.
How can I fixed it so it will use NODE0 and not NODE1.

It always uses CORE 25,26.27 now the last test it changed to 25,26,30

Why does it happen? Can I manually change it with a program or registry?

 

These are obviously the preferred cores in the system.

 

You can try the setting "Interrupt Steering Settings => Interrupt Steering Mode"

in this utility https://forums.guru3d.com/threads/windows-power-plan-settings-explorer-utility.416058/

 

Hey, thanks

Did you test it? did you see changing for yourself?
I mean would it change randomly somehow? what is the default settings? I just want to make sure I know what to go back to if nothing changes.

 

Spoiler

It seems it's default like normal. does the apply happens instant or do I need to restart the computer? I want to see what's the fastest way to see if it changes.

I tried changing it but I didn't see any changes with Latencymon. Does it change instant?

 

use ryzen master to identify the preferred cores.....

 

What do you mean? The last thing I remember with ryzen is nothing much with settings. just some overclock.

 

Default value for this setting is "Default". So you select value "Processor 0 - Route interrupts to Processor 0", and if nothing changes then you can select "Default".
I have not test it.
And I think you need to reboot after you change this setting, just in case.

 

I tried restarting also. it doesn't change anything. what could this be? without SMT/HT Disabled it goes to Core0 like normal, when SMT/HT Enabled it goes to 24,25,26.

 

hello



hpet off with:

bcdedit /set useplatformtick yes
bcdedit /set disabledynamictick yes
bcdedit /set useplatformclock yes

don't know which one command did that tbh

Mouse is REALLY responsive and after i delete all values, responsiveness stays same but i can play on high fps. Don't know why but it's working only first enabling hpet, then disabling it, then putting 3 commands, and then removing them from bcdedit xd PC restart after each step.

EDIT:

HPET OFF with bcdedit /set useplatformclock yes

Mouse (and csgo) have best responsiveness in this combination but my low fps are lower than they used to ;/

 

No offense, but if you are so obsessed and worried about the "latency" lecture some arbitrary program gives due to HT, then sell your i7 and get an i5 K CPU of the same gen or just buy a new i3 9100F or something like that. After all, if you disable HT on any i7 it just becomes an i5. You are crippling it's performance by half, that under any circumstance and logic is extremely stupid to me beyond words. Also, just call it HT to avoid confusions, SMT is for Ryzen.

 

Intel`s HT is a marketing name for SMT.
Btw, I met two abbreviations - Symmetric Multithreading and Simultaneous Multithreading.

PS Also I have impression that X7007 wrote about Ryzen rig.

 

Face-palm...
Smoke manuals again.
How to disable HPET:
bcdedit /set useplatformclock no
bcdedit /set disabledynamictick yes (optional but needs to be disabled, not HPET)