revisiting hpet bcdedit tweaks: what are your timer bench results and settings?

[:D:D]

would there be any issues from forcing globaltimerrequests on in registry? islc says this feature is only for win 11.

that works for me. so globaltimerrequests seems to work on win 10, too. i get best values around 0.5070 and theres definitely delta changing going on above and below that. as soon as i close dpclat "slept" goes back to 15 again.

Some nice times there, my gear is too old I think.
Shouldn't be any issues with registry except taking up space and not doing anything for W10.
I've heard of islc and lasso but don't use them.
I could only get global settings by using DPCLat driver.
As lowest value wins then maybe apps setting 0.5 destroy the purpose of having a bit more than that unfortunately.

 

[zoulztealer]

Framerate Isn't Good Enough: Latency Pipeline, "Input Lag," Reflex, & Engineering Interview

 

[zoulztealer]

Some nice times there, my gear is too old I think.
Shouldn't be any issues with registry except taking up space and not doing anything for W10.
I've heard of islc and lasso but don't use them.
I could only get global settings by using DPCLat driver.
As lowest value wins then maybe apps setting 0.5 destroy the purpose of having a bit more than that unfortunately.

what i could tell is that its enough to have dpclat merely opened and not running / measuring to get the slept timer decrease.

so to have lowest slept timer i would have to keep open dpclat in background not running but sadly i could not tell any improvement from this.

i also dont get / feel any improvement from nvidia reflex which makes me believe ive more or less already reached bottom of reduction of input latency - on this machine - but theres still much more stuff to go.

this ones also important, i believe it could be responsible for changing the timer dynamically by prioritizing applications:

Multimedia Class Scheduler Service - Win32 apps

The Multimedia Class Scheduler service (MMCSS) enables multimedia applications to ensure that their time-sensitive processing receives prioritized access to CPU resources.

learn.microsoft.com

How does Multimedia Class Scheduler function?

enable hags and add your game / app to high performance preference:

there are several tweaks out there for this. you can tweak every task in the same way. i have maxed everything out / set all other tasks identical to the games task.

some of the most relevant powershell commands, registry locations, etc from dady madus github script (only for advanced users, may contain danger of system damage. if you dont understand the code / are able to read / understand the script nor the commands do not apply these.): (https://github.com/DaddyMadu/Window...217d0d4/win10debloatandgamingtweaks.ps1#L2534):

##########
# Gaming Tweaks Functions
##########

#Game Optimizations Priority Tweaks
#MULTIMEDIA CLASS SCHEDULER PRIORITY TWEAKS
Set-ItemProperty -Path "HKLM:\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Multimedia\SystemProfile\Tasks\Games" -Name "GPU Priority" -Type DWord -Value 8
Set-ItemProperty -Path "HKLM:\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Multimedia\SystemProfile\Tasks\Games" -Name "Priority" -Type DWord -Value 6
Set-ItemProperty -Path "HKLM:\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Multimedia\SystemProfile\Tasks\Games" -Name "Scheduling Category" -Type String -Value "High"
Set-ItemProperty -Path "HKLM:\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Multimedia\SystemProfile\Tasks\Games" -Name "SFIO Priority" -Type String -Value "High"
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\PriorityControl" -Name "IRQ8Priority" -Type DWord -Value 1
#CSRSS.EXE REALTIME PRIORITY TWEAK / REDUCES INPUT LATENCY, IMPROVE SYSTEM RESPONSIVNESS
reg ADD "HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image File Execution Options\csrss.exe\PerfOptions" /v CpuPriorityClass /t REG_DWORD /d 4 /f | Out-Null
reg ADD "HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image File Execution Options\csrss.exe\PerfOptions" /v IoPriority /t REG_DWORD /d 3 /f | Out-Null
#DISABLE FILE CREATION BEHAVIOUR / REDUCE DISK OVERHEAD
fsutil behavior set disable8dot3 1
fsutil behavior set disablelastaccess 1

# [...] Fullscreen Optimizations
#DISABLE WINDOWS GAME VIDEO RECORDER
Set-ItemProperty -Path "HKCU:\System\GameConfigStore" -Name "GameDVR_FSEBehaviorMode" -Type DWord -Value 2
Set-ItemProperty -Path "HKCU:\System\GameConfigStore" -Name "GameDVR_HonorUserFSEBehaviorMode" -Type DWord -Value 1
Set-ItemProperty -Path "HKCU:\System\GameConfigStore" -Name "GameDVR_FSEBehavior" -Type DWord -Value 2
Set-ItemProperty -Path "HKCU:\System\GameConfigStore" -Name "GameDVR_DXGIHonorFSEWindowsCompatible" -Type DWord -Value 1
Set-ItemProperty -Path "HKCU:\System\GameConfigStore" -Name "GameDVR_EFSEFeatureFlags" -Type DWord -Value 0
Set-ItemProperty -Path "HKCU:\System\GameConfigStore" -Name "GameDVR_DSEBehavior" -Type DWord -Value 2
Set-ItemProperty -Path "HKCU:\SOFTWARE\Microsoft\Windows\CurrentVersion\GameDVR" -Name "AppCaptureEnabled" -Type DWord -Value 0
Set-ItemProperty -Path "HKCU:\Software\Microsoft\DirectX\GraphicsSettings" -Name "SwapEffectUpgradeCache" -Type DWord -Value 1
Set-ItemProperty -Path "HKCU:\Software\Microsoft\DirectX\UserGpuPreferences" -Name "DirectXUserGlobalSettings" -Type String -Value 'SwapEffectUpgradeEnable=1;'
#DISABLE GPU MULTI PLANE OVERLAY / MPO
Set-ItemProperty -Path "HKLM:\SOFTWARE\Microsoft\Windows\Dwm" -Name "OverlayTestMode" -Type DWord -Value 5
Disable-MMAgent -MemoryCompression | Out-Null

disable memory compression:
#REDUCES CPU & MEMORY OVERHEAD

Enable or Disable Memory Compression in Windows 10 and Windows 11 Tutorial

This tutorial will show you how to enable or disable Memory Compression in Windows 10 and Windows 11. Memory Compression allows Windows 10 and Windows 11 to make better use of available memory on all systems by reducing physical memory utilization, resulting in fewer pagefile IO operations and...

www.elevenforum.com

#Enable Hardware-accelerated GPU scheduling
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\GraphicsDrivers" -Name "HwSchMode" -Type DWord -Value 2

#Decrease mouse and keyboard buffer sizes.
#REDUCES MOUSE & KB INPUT LATENCY
New-Item -Path "HKLM:\SYSTEM\CurrentControlSet\Services\mouclass\Parameters" | Out-Null -ErrorAction SilentlyContinue
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Services\mouclass\Parameters" -Name "MouseDataQueueSize" -Type DWord -Value 0x00000020
New-Item -Path "HKLM:\SYSTEM\CurrentControlSet\Services\kbdclass\Parameters" | Out-Null -ErrorAction SilentlyContinue
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Services\kbdclass\Parameters" -Name "KeyboardDataQueueSize" -Type DWord -Value 0x00000020
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Services\kbdclass\Parameters" -Name "KeyboardDataQueueSize" -Type DWord -Value 0x00000020

#Tell Windows to stop tolerating high DPC/ISR latencies.
#EFFECTS UNKNOWN
New-Item -Path "HKLM:\SYSTEM\CurrentControlSet\Control\Power" | Out-Null -ErrorAction SilentlyContinue
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\Power" -Name "ExitLatency" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\Power" -Name "ExitLatencyCheckEnabled" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\Power" -Name "Latency" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\Power" -Name "LatencyToleranceDefault" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\Power" -Name "LatencyToleranceFSVP" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\Power" -Name "LatencyTolerancePerfOverride" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\Power" -Name "LatencyToleranceScreenOffIR" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\Power" -Name "LatencyToleranceVSyncEnabled" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\Power" -Name "RtlCapabilityCheckLatency" -Type DWord -Value 1
New-Item -Path "HKLM:\SYSTEM\CurrentControlSet\Control\GraphicsDrivers\Power" | Out-Null -ErrorAction SilentlyContinue
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\GraphicsDrivers\Power" -Name "Latency" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\GraphicsDrivers\Power" -Name "MaxIAverageGraphicsLatencyInOneBucket" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\GraphicsDrivers\Power" -Name "MonitorLatencyTolerance" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\GraphicsDrivers\Power" -Name "MonitorRefreshLatencyTolerance" -Type DWord -Value 1
Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\GraphicsDrivers\Power" -Name "TransitionLatency" -Type DWord -Value 1

also markcs mouse fix is very important to apply for gaming:

The MarkC Windows 10 + 8.1 + 8 + 7 Mouse Acceleration Fix

The 'MarkC Windows 10 + 8.1 + 8 + 7 Mouse Acceleration Fix' is available for download from this link: DO NOT USE THIS: MarkC_Windows_10+8....

donewmouseaccel.blogspot.com

 

[:D:D]

Wow, you've done a lot of testing. Thanks for the sleep(1) test. Even if the result was disappointing it's still good to know.

 

[zoulztealer]

Wow, you've done a lot of testing. Thanks for the sleep(1) test. Even if the result was disappointing it's still good to know.

i didnt test much as it would have to be done like testing for max fps, avg lows etc although my 3dmark score seems insanely good but rather i was trying to get the best responsiveness and input latency out of it since im always gaming with vsync 60 and theres virtually no difference between vsync on or off.

its great to see people appreciate all of this, thank you.

 

[zoulztealer]

ive received a message now from the creator of the timerbench app which is very useful to know:

Only useplatformclock is used to enable/disable HPET. And it's only done when you want to switch between TSC and HPET.

TimerBench is not a tweak software. It provides two benchmarks to determine if your timer settings are bottlenecking your game performance.

So you can try all of the toggles above and check if they have any influence.

my question was:

how does your app relate to what is defined and listed in bcdedit.

the commands in question are:

tscsyncpolicy
useplatformclock
useplatformtick
disabledynamictick

does your app read or change these settings to determine or define whether hpet is enabled or disabled?

when clicking on the button in the app to disable or enable hpet what does your app exactly do or change, what methods or commands are exactly issued?

from melodys tweaks there are several combos for different timer compositions:

https://sites.google.com/view/melod...rs-hpet-tsc-pmt

TSC + TSC without desync: bcdedit /deletevalue useplatformclock - bcdedit /deletevalue useplatformtick - make sure HPET is enabled in BIOS

TSC + RTC: bcdedit /deletevalue useplatformclock - bcdedit /set useplatformtick Yes

HPET + RTC: bcdedit /set useplatformclock Yes - bcdedit /set useplatformtick Yes - make sure HPET is enabled in BIOS

PMT + RTC: bcdedit /set useplatformclock Yes - bcdedit /set useplatformtick Yes - make sure HPET is disabled in BIOS

so in order to use your app would i have to delete all bcdedits first or can i just run your app and toggle hpet on / off while leaving bcdedit entries untouched?

i imagine it would probably be best to delete the entries first before running your app or making changes with it but in my research i have also found that the same settings that define the hpet status are hidden from and inacessible to the user in another location.

you can read about it here:

The HPET bug: What it is and what it isn't

Can you tell me know how upload results ?

www.overclockers.at

 

[zoulztealer]

some benches i did recently and theres are all first run and single run.

setting timer to 0.507 and changing the dword "Background Only" from true to false in multimedia class scheduler task for games, display post process and window manager seemed to have had some measureable improvements in war thunder bench and min fps went from 56 / 57fps to solid 60 that is with vsync on and fps fractured at 59.97fps with rtss. so theres definitely more pressure to keep up fps now and this seems to me to be some kind of a new effect, too. the other tasks are assigned all background only true; audio, capture, distribution, playback, pro audio, etc

i used to keep all categories of tasks including games in background only before because i was expecting to see some improvement from this and there is, but it seems now its better to toggle it / set to false (allow task to change threads on window focus; this seems to be probably important to assign foreground priority boosts)

Background Only

REG_SZ

Indicates whether this is a background task (no user interface). The threads of a background task do not change because of a change in window focus. This value can be set to True or False.

Multimedia Class Scheduler Service - Win32 apps

The Multimedia Class Scheduler service (MMCSS) enables multimedia applications to ensure that their time-sensitive processing receives prioritized access to CPU resources.

learn.microsoft.com

id like to receive some data from your pcs for the same 3dmark benches i did and post them here if you like to. im a bit surprised i get so good results in the firestrike dx11 bench but the others are more average still not bad i believe for the hardware.

I scored 29 065 in Fire Strike

AMD Ryzen 7 5700X, NVIDIA GeForce RTX 2080 Ti x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

I scored 14 040 in Time Spy

AMD Ryzen 7 5700X, NVIDIA GeForce RTX 2080 Ti x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

I scored 14 350 in Steel Nomad Light

AMD Ryzen 7 5700X, NVIDIA GeForce RTX 2080 Ti x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

I scored 3 463 in Steel Nomad

AMD Ryzen 7 5700X, NVIDIA GeForce RTX 2080 Ti x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

I scored 54 872 in Night Raid

AMD Ryzen 7 5700X, NVIDIA GeForce RTX 2080 Ti x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

i also believe frametimes in timerbench have become better, i also get better results in windowed mode, ie. higher gpu load / utilization, higher fps and better max frametimes. but these are also first and single run. may get better values with caching over time, fullscreen was first uncached run which usually explains shot high bursts of max frametimes when theres stuttering when theres loading from the disk / caching going on, windowed was second 1 x time cached run.

i ran timerbench quickly on my other r7 1700 & rx 570 b450 it was running on itsc by default and had none of the bcdedit entries and scored like 100-120fps. so thought of as in fps this would mean this pc is roughly 4 times faster than the other one. the rx570 also got way better gpu load in windowed than in fullscreen.

there are visible differences in the frametime graphs that were better to be analyzed if you could zoom in because the windowed frametimes seem to be lower in general but more jittery or spikey while the fullscreen frametimes seem to be genrally higher but more flat or something. i can clearly tell that exclusive fullscreen is absolutely smooth and usually the best choice in games in terms of fps and input latency while borderless / windowed modes tend to be a good choice, too - with some exceptions where it can be worse than exclusive fullscreen but windowed modes usually run super well, too - and mind there are 3 or 4 different modes of fullscreen modes and input latency varies greatly from game to game and how sync modes and triple buffering have been implemented.

i have to mention here again doom eternals graphic option / function called "present from compute" which brings in a frame from an asychnronous que; this combined with real triple buffering - how they implemented it - make the game utterly perfect perfect smooth and absolute lowest input latency with vsync on. doom eternal runs incredibly well and is incredibly well optimized.

i have no idea what the difference between borderless window and windowed fullscreen is and what it would mean in terms of fps, stability and compatibility - but as far as i can tell windowed mode in general / resizeable window seems to run very well for me especially when your alt tabbing all the time and it solves a lot of overlay issues. ive recently reinstalled gpu driver with multi plane overlay on and its better than off i believe.

 

[zoulztealer]

the wintimertester app in the reddit thread i have checked to be safe.

https://www.reddit.com/r/leagueoflegends/comments/284ctp
afaik youd want a counter without syncing which avoids several issues:

TSC+LAPICs Low performance (slow timers + syncing) = 2.76MHz

LAPICs low performance (slow timer - no syncing) = 3.5Mhz

TSC+HPET medium performance (slow and fast timer + syncing) = 3.8Mhz

HPET high performance (fast timer - no syncing) = 14.3MHz

read the discussion with the microsoft engineer below the tweak hound article:

Timer Tweaks Benchmarked - TweakHound

Optimize your computing experience, Performance, Tweak, Tweaker, Tweaking Windows, Tweak Guides, Tweakguides, Privacy, Security, Computer How-to, fix computer, Computer news and information, Windows Xp, Windows 7, Windows 10, Linux

www.tweakhound.com

bcdedit /set useplatformclock true >>> don't do this cmd edit yet! :: No Man's Sky General Discussion

Some players are editing their registry to solve system clock-related performance and crash issues, see thread: https://steamcommunity.com/app/275850/discussions/0/1643171537291925442/ However, if you go to the source on this (or one of the sources), the cmd edit may resolve the issue for...

steamcommunity.com

You may recall that traditionally Windows OSs function on the assumption / use of a Single Clock Domain for a given server, however with the ability of servers to be physically “scaled” (connected together) to create a larger “multinode” server (IE – 2, 3, 4, 5, 6, 7 or 8 node), we have a problem where each server has it’s own local “clock”. This creates a “multi-clock” domain which in and of itself is not bad, however, the “clocks” are Not synchronized across all nodes (unless Hardware clock synchronization is implemented which is very difficult/involved to implement) therefore there can be clock skew / drift between nodes & processors (other than Node 0) which can lead to thread scheduling and timing issues which at minimum can lead to performance problems in addition to other strange & bizarre behavior (poor network & disk performance, hang conditions, etc. etc. etc.).

The problem is encountered within a multi-clock domain beginning in 2008 R2 when the TSC was re-introduced as the default Clock (as mentioned above) Vs the use of the HPET (or Power Management (ACPI / PMclock)) Clock that prior OS versions used. The TSC is very fast and reliable but in using the TSC as the default Source Clock, the OS assumes a Single clock source. Depending upon which node the RDTSC thread executes, clock synchronization may / can become skewed resulting in thread scheduling problems and issues such as is mentioned above.

[..]

This said, it is possible that depending upon the era of the HW platform (desktop, laptop, workstation, etc) that there may be some stability issues of some systems running Windows (due to HW / FW and OS (TSC) implementations) being within a transitional period of time where a system may have better stability when using a system’s default (or configured platform clock within BIOS/UEFI) which is not the TSC (HPET, ACPI/PM_Timer) so stability is also a performance consideration.

In short, the TSC provides the lowest overhead which can and does translate to lower overhead / latency generally resulting in better performance and therefore the use of the TSC for Windows and later should remain as the default clock source unless known issues or problems are encountered.​

Acquiring high-resolution time stamps - Win32 apps

Windows provides APIs that you can use to acquire high-resolution time stamps or measure time intervals.

learn.microsoft.com

Should I use QPC or call the RDTSC/RDTSCP instructions directly?

To avoid incorrectness and portability issues, we strongly encourage you to use QPC instead of using the TSC register or the RDTSC or RDTSCP processor instructions.

Is QPC accuracy affected by processor frequency changes caused by power management or Turbo Boost technology?

No. If the processor has an invariant TSC, the QPC is not affected by these sort of changes. If the processor doesn't have an invariant TSC, QPC will revert to a platform hardware timer that won't be affected by processor frequency changes or Turbo Boost technology.

Does QPC reliably work on multi-processor systems, multi-core system, and systems with hyper-threading?

Yes.​

Which processors have non-invariant TSCs? How can I check if my system has a non-invariant TSC?

You don't need to perform this check yourself. Windows operating systems perform several checks at system initialization to determine if the TSC is suitable as a basis for QPC. However, for reference purposes, you can determine whether your processor has an invariant TSC by using one of these:

the Coreinfo.exe utility from Windows Sysinternals
checking the values returned by the CPUID instruction pertaining to the TSC characteristics
the processor manufacturer’s documentation
The following shows the TSC-INVARIANT info that is provided by the Windows Sysinternals Coreinfo.exe utility (www.sysinternals.com). An asterisk means "True".

Coreinfo - Sysinternals

Coreinfo is a command-line utility that shows you the mapping between logical processors and the physical processor.

learn.microsoft.com

 

[Macro Device]

HPET was disabled by default. Timerbench showed significant loss of performance:

And Cyberpunk 2077 just outright became UNPLAYABLE. 148 FPS sans HPET, 3 FPS with HPET. I wish I were kidding.

 

[zoulztealer]

HPET was disabled by default. Timerbench showed significant loss of performance:
View attachment 366221
View attachment 366222

And Cyberpunk 2077 just outright became UNPLAYABLE. 148 FPS sans HPET, 3 FPS with HPET. I wish I were kidding.

awesome, thanks! strangely there are no values on gpu load. can you try another bench with fullscreen or post the gpu load graph from your results?

 

[Macro Device]

GPU is basically unable to initialise properly. It must clock as high as ~2650 MHz, yet it crawls at 15...495 MHz and jumps 0...20 percent load, usually around 0. CPU load is stupid high and Task Manager can't tell me why, will later take a peek with more advanced software.

I suspect it's AMD driver.

 

[zoulztealer]

View attachment 366226

GPU is basically unable to initialise properly. It must clock as high as ~2650 MHz, yet it crawls at 15...495 MHz and jumps 0...20 percent load, usually around 0. CPU load is stupid high and Task Manager can't tell me why, will later take a peek with more advanced software.

I suspect it's AMD driver.

very strange. still love to see your timerbench graph for the gpu load if you have em.

is hpet enabled or disabled in your bios and are you running the latest firmware?

 

[Macro Device]

is hpet enabled or disabled in your bios

No idea. There is no such toggle to begin with. Latest BIOS for my mobo, yes.

still love to see your timerbench graph for the gpu load if you have em.

Lemme re-enable HPET once again...

It's insane.

MSi Afterburner overlay showed 30...75% GPU load. Sans HPET, 100% sharp all the time.

Will re-test next hour after complete driver reinstall, need to go shopping.

So, the complete GPU driver sweep didn't do a thing. 0 difference between before and after.
Enabling/disabling MPO didn't affect the behaviour.
Changing display refresh rate didn't alter anything.

With HPET enabled, the PC behaves like the OS is installed on a very slow HDD and there's viri going on + the GPU is 70% paralysed. Sometimes it feels like it's gonna BSoD right away.

NO IDEA WHAT'S GOING ON.

 

[zoulztealer]

No idea. There is no such toggle to begin with. Latest BIOS for my mobo, yes.

Lemme re-enable HPET once again...

It's insane.

View attachment 366235

MSi Afterburner overlay showed 30...75% GPU load. Sans HPET, 100% sharp all the time.

Will re-test next hour after complete driver reinstall, need to go shopping.

So, the complete GPU driver sweep didn't do a thing. 0 difference between before and after.
Enabling/disabling MPO didn't affect the behaviour.
Changing display refresh rate didn't alter anything.

With HPET enabled, the PC behaves like the OS is installed on a very slow HDD and there's viri going on + the GPU is 70% paralysed. Sometimes it feels like it's gonna BSoD right away.

NO IDEA WHAT'S GOING ON.

tyvm for the amendments.

yes, that can indeed happen. you shouldnt do any more testing with this to prevent the pc from getting stick in the bad performance state since this could happen when you apply bcdedit tweaks.

so my recommendation is do not apply any bcdedit tweaks and leave everything as it is with itsc.

 

[Macro Device]

you shouldnt do any more testing

I got a mustdie USB stick chilling around ready for any unexpected outcome and I don't have anything else to do as of now since I'm waiting for my work partners to come up with something and pushing them doesn't seem to work so I'm happily screwing around with my PC. Especially considering it's 3 in the morning and they don't show up before 12 pm anyway.

But yeah, I'd rather not. Any idea why it does misbehave like this?

 

[zoulztealer]

I got a mustdie USB stick chilling around ready for any unexpected outcome and I don't have anything else to do as of now since I'm waiting for my work partners to come up with something and pushing them doesn't seem to work so I'm happily screwing around with my PC. Especially considering it's 3 in the morning and they don't show up before 12 pm anyway.

But yeah, I'd rather not. Any idea why it does misbehave like this?

hehe, cool! its always good to have your windows ready to be shredded any time when your a gamer or similar user and its reinstalled quite fast; cup of coffee and 30m your good with ssd and usb 3.0

all i can imagine is there are issues with syncing the different timers or certain bios settings that pertain to gpu, cpu, power saving, power plan etc

but thats actually why i made this thread and wanted to investigate and i just got another report from elsewhere where one guy botched his windows with the bcdedit tweaks which needs to be reinstalled probably since theres no other solution to reverse this.

im very surprised im getting such good values and behaviour with hpet forced on in all aspects and im afraid to make a change because i cant risk this windows to be tossed.

have you tried applying different msi(x) irq policies with msiutilv3 or ms affinity tool?

what do you get from dpclat and latencymon?

you may also try to "repair windows" with sfc /scannow or DISM /Online /Cleanup-Image /RestoreHealth and see if that does anything.

and also tried processlasso or islc / custom timer resolution & global timer resolution tweak?

 

[Macro Device]

msi(x) irq policies

dpclat

islc?

Have no idea what these are.

ms affinity tool?

latencymon?

processlasso

Never tried them.

If that's of any importance timers check out regardless of the state of HPET. 100% in sync, all three show 11:33:05.

HPET off.

 

[zoulztealer]

its a very good sign if all your timers are 100% in sync all the time. you tested with cpuz? if you can post a screen of that. my timers seemed to be in sync too on first galance but when i made a screenshot with the print screen key it turned out they were slight off by a very low value.

dpclat

DPC Latency Checker

Testet die verfügbaren Ressourcen des Rechners zum fehlerfreien Verarbeiten von Echtzeit-Datenströmen.

www.heise.de

islc

Intelligent standby list cleaner (ISLC) v1.0.3.3 Released - Wagnardsoft Forum

go ahead and make some diagnosis with dpclat and latencymon. they dont make any changes just read results.

values look good on the latencymon you posted with hpet off.

msiutilv3

Windows: Line-Based vs. Message Signaled-Based Interrupts. MSI tool.

... or another attempt to improve latencies Little bit of theory: ***** From "Windows Internals" by Mark Russinovich, David A. Solomon, Alex...

forums.guru3d.com

try switching all your devices or devices important to you into msi mode.

 

[Macro Device]

Sans HPET once again.

Switched everything to the MSI mode. Now we reboot with HPET enabled and see if anything changes.

It's all over the place now with HPET enabled.

 

[zoulztealer]

View attachment 366248

Sans HPET once again.

Switched everything to the MSI mode. Now we reboot with HPET enabled and see if anything changes.

ive expected it to look like this which isnt necessarily bad. there was another case in this thread a few pages before that had the same look and we managed to improve it somewhat. what helped in this case most of all was to set disabledynamictick enabled. but as i said be very careful with the bcdedit tweaks although disabledynamictick shouldnt to anything irreversible but platformclock and platformtick will definitely.

the hpet on effects are disturbing but very intresting.

 

[Macro Device]

Nah, I don't wanna fix what's not broken anymore. Was perfectly fine with HPET off, thought maybe it can gimme something interesting. Fruits are there but they ain't tasty.

If I can do anything useful for this thread without enabling HPET again let me know.

 

[zoulztealer]

Nah, I don't wanna fix what's not broken anymore. Was perfectly fine with HPET off, thought maybe it can gimme something interesting. Fruits are there but they ain't tasty.

If I can do anything useful for this thread without enabling HPET again let me know.

you could try and check if you get improvements in either the timer bench and / or 3dmark benches (preferably fire strike dx11) with islc and process lasso.

from inside islc set custom timer res to 0.507 and toggle globaltimerrequests and see if you get improvements from that.

try different power plans.

 

[Macro Device]

globaltimerrequests

Shows as Windows 11 only and I'm on W10. Will it blow me up?

process lasso

Do I need to do anything in this software or does it just do its thing on its own?

 

[zoulztealer]

Shows as Windows 11 only and I'm on W10. Will it blow me up?

Do I need to do anything in this software or does it just do its thing on its own?

im on win10 and have it enabled. it shouldnt do any harm. weve found that it becomes only really active when dpclat is opened in the background. you can leave that one out if you want.

heres an explanation on it:

try to check if you get significant improvements in timerbench or 3dmark between 0.5 and 0.507 timer res in islc.

in process lasso the beauty is you dont need to do or config anything just install it with all default options dont need to set anything just be sure you have bitsum highest performance power plan active.

oh yeah, and disable windows game mode if you havent thats quite important.

 

[Macro Device]

0.5 to the left, 0.507 to the right. Vanilla test in the centre. Windowed/fullscreen seems to have zero difference. Apparently these tools only do harm to the performance (or don't impact it at all, idk). Gaming tests didn't reveal any difference. CP2077 shows exactly the same FPS in all scenarios. 1% and 0.1% lows stay where they were.