Timer Benchmarking tool

[soul4kills]

I got AI to make this timer benchmarking tool. I made it to replace ISLC & DPC Latency Checker. And it pretty much does just that even better and a lot more light weight.

Helps to maintain the lowest possible timer when you're gaming.

Polling rate (Adjustable): 1000ms - 1ms

Sample Range limit/cap (Adjustable): 500 - 50
Lowering it gives you more a more live view of metrics. Higher gives you more historical metrics.

Timer Resolution: Can use up and down arrows to adjust by the millionths or type it in manually. Starts up at 0.5001, which is the best i've found for my machine.

2 Views. A progressive bar graph, and a histogram graph. Giving a real clear view of how your kernels timer resolution is behaving.

Theres also a benchmark tool which is supposed to find the lowest standard deviation by sampling 1000 measurements @ ever 1ms. It steps your timer resolution starting at 0.5000 - 0.5010. Then locks in the one with the lowest deviation, meaning it's the most stable. Still on the fence if it works well or not, but I'll just keep it.

It has to be running on the same desktop as the game if you have a dual monitor or virtual desktop setup. It's a quirk of windows 11 dynamic timer resolution. It doesn't do anything to the machine, just takes measurements. When you exit your timer will go back to defaults.

Enjoy.

Update:
Newest file is here.
viewtopic.php?t=15427&sid=1e12118b95941 ... 8e#p124054

[soul4kills]

Updated the tool.

Adjusting timer: Pressing down past 5000 uses windows defaults.
Windows 11 Dynamic Timer: Added a focused mode where this isn't a problem anymore. Starts up respecting the dynamic timer.
Jitter Distribution: Added a new graphic just for fun to see the deviations. Can be toggled on/off with "j"
Conhost: Starts up opened in windows new terminal but it has a bit of overhead so added an option to open the script in conhost to minimize overhead. It's less pretty.
Benchmarking: The original benchmarking worked fine, just the numbers were calculated wrong, but it still resulted in the right tier of results. Fixed that.
The Graphs: Fixed the graphs, they were not aligned properly with their buckets. Also moved percentages to the tips of the histogram graphs.

I think that's it.

Edit: Be careful with the sample rate. If you're bench-marking or stress testing, using 1ms rate is fine, but if you're in a game, it's better to have it at 1000ms or 500ms, if you're actively playing a game and are not stress testing.

---

Why did i make it?
Saw another tool like it but it was paywalled so i built my own and am sharing it for free. It's a lot better than the paywalled version.
Also to replace 2 tools I use, ISLC & DPC Latency Checker, to help with maintaining a low timer resolution.

Does it work?
I believe it does. My view is this. This tool just like ISLC tells your kernel, hey be ready for requests at this interval. Basically it. Normally games only request a timer of 1ms. This requests 0.500ms. So if you understand logic, you understand how this can be advantageous. To put it into better context, 2 seconds of cpu time becomes 1 second of cpu time.

How do I know it works for real though?
Well I play a game with it on and off. And i feel the difference in response time from my inputs. Also it tightens up your frame times as well.

How else can it be useful?
Even if you don't believe the above. If you have bad input latency, this tool can help you see it in real time. A little better than DPC Latency Checker and Latency Mon. In that it has the graphs lacking in the former tools to better visualize the problem. Not made to replace it in this sense, but to complement them.

Should you use it?
Do what you want.

[soul4kills]

Another Update:

-Added ability to turn off/on GlobalTimerResolutionRequests, and it will also read it's state.
-Added elevation detection for the tool. You need to be admin level to change GlobalTimerResolutionRequests. It also requires a restart if changed.
-Added additional deviation methods, raw/trimmed/winsor, so outliers don't skew your deviation results.
-Coupled the Sample amount to the benchmarking tool. If it's at 100, your benchmark will take test for 100 samples each. If it's at 500, your benchmark will test at 500 samples each.
-Cleaned up the interface a bit.
Made Focused mode default during startup.
-Cleaned up the jitter graph. The tighter the inside wiskers are the more stable your timer is. The red outside wiskers just represent the max/min outliers. So they'll always be big.

Video of a live demonstration.
https://imgur.com/a/UyDrOCP

The top app you see is my first iteration of this tool. It does not hold a timer request. It only probes the kernel and measures how fast it responds. So it's not connected in any way to the second app. Watch it's measurements when I start up my updated timer bench tool. The measurements get better right away. The first app represents your game. The the tighter the measurements the faster the response time.

Then throughout the video, in the updated tool, I will cycle the "[T]imer Res." from windows defaults "10000us" to "5001us". You can see it's relationship to the kernel and the timer resolution and the other app and how it can positively impact your game.

Be mindful of the sample rate, use only 1000ms - 250ms only when gaming. Any lower than that adds overhead to your kernel. Use only lower values if youre benchmarking or stress testing things. The sample quantity doesn't matter, 500 means more data to smoothing out the numbers and the histogram. 100 gives you better view of what's currently going on and will jump around a lot. Make sure you have GlobalTimerResolutionRequests set to on if you're on windows 11. High Performance power plan, and in focused mode, which is it's default state now.

GLHF

[mocha__]

the thing im starting to question is; any little advantage or milliseconds you give yourself are a lot of the times detrimental in the multiplayer gaming world today. the lag compensation and algorithms all these mp games use to make things fair actually seem to be a detriment on tweaked systems. aside from 'measurable' performance benefits, more times than not, you will encounter the issues i mentioned, leading you to tweak more or less.
call me an idiot but i think the state of mp games, competitive mp games, dont want people having too much of an advantage. its not about raw skill as much as it used to be. theres so many variables in games where they try so hard to compensate. theres people out there far worse than you at any said game but they will defy logic, might place better than you or average better than you, and make you question your own skill, or your setup.
i think that plays more of a part than anything else.

[soul4kills]

the thing im starting to question is; any little advantage or milliseconds you give yourself are a lot of the times detrimental in the multiplayer gaming world today. the lag compensation and algorithms all these mp games use to make things fair actually seem to be a detriment on tweaked systems. aside from 'measurable' performance benefits, more times than not, you will encounter the issues i mentioned, leading you to tweak more or less.
call me an idiot but i think the state of mp games, competitive mp games, dont want people having too much of an advantage. its not about raw skill as much as it used to be. theres so many variables in games where they try so hard to compensate. theres people out there far worse than you at any said game but they will defy logic, might place better than you or average better than you, and make you question your own skill, or your setup.
i think that plays more of a part than anything else.

You're wrong but you're also right at the same time. Because the same unskilled person is subject to the same variables as you. So skill does still matter.

Edit: A good example of what you're talking about when developers design to close the skill gap. Modern warfares slide canceling is a good example. It gave people who knew how to do it a huge advantage in fights. Then when the next MW iteration came out they took out slide canceling completely. If you started to lose to "unskilled" players after this change, could you really say you were any good at all? In this situation, you gotta ask yourself were you only good at slide canceling? A skilled player would be unaffected by it.

But my tool isn't about gaining an advantage. You won't get that advantage if you're suffering from input lag due to IO overhead while using my tool. It might help a little but it's more to help people identify if they're suffering from input lag due to kernel IO overhead. The graphs help a lot to make it obvious and help identify if a tweak worked or not. But it can also give you an a slight edge as well once you've solved all the latency issues. My game feels a lot smoother while using the tool vs when not using it.

[mocha__]

the thing im starting to question is; any little advantage or milliseconds you give yourself are a lot of the times detrimental in the multiplayer gaming world today. the lag compensation and algorithms all these mp games use to make things fair actually seem to be a detriment on tweaked systems. aside from 'measurable' performance benefits, more times than not, you will encounter the issues i mentioned, leading you to tweak more or less.
call me an idiot but i think the state of mp games, competitive mp games, dont want people having too much of an advantage. its not about raw skill as much as it used to be. theres so many variables in games where they try so hard to compensate. theres people out there far worse than you at any said game but they will defy logic, might place better than you or average better than you, and make you question your own skill, or your setup.
i think that plays more of a part than anything else.

You're wrong but you're also right at the same time. Because the same unskilled person is subject to the same variables as you. So skill does still matter.

Edit: A good example of what you're talking about when developers design to close the skill gap. Modern warfares slide canceling is a good example. It gave people who knew how to do it a huge advantage in fights. Then when the next MW iteration came out they took out slide canceling completely. If you started to lose to "unskilled" players after this change, could you really say you were any good at all? In this situation, you gotta ask yourself were you only good at slide canceling? A skilled player would be unaffected by it.

But my tool isn't about gaining an advantage. You won't get that advantage if you're suffering from input lag due to IO overhead while using my tool. It might help a little but it's more to help people identify if they're suffering from input lag due to kernel IO overhead. The graphs help a lot to make it obvious and help identify if a tweak worked or not. But it can also give you an a slight edge as well once you've solved all the latency issues. My game feels a lot smoother while using the tool vs when not using it.

i think you missed the point i was trying to make. im not talking about adding stupid mechanics everyone can use. im talking about fixing scenarios. adding more or less interpolation to certain players. you brought up cod which is a good example for their patented manipulation of hitreg, ping, gunfights, packet burst, etc etc.
im saying that a lot of mp games or rather, competitive mp games, addicting games are all using some form of this matchmaking manipulation, way more in tact shooters because it keeps people hooked. the only game i ever played that didnt feel counter intuitive to skill is csgo. valorant was and is a game built around trying to be the most competitive esport shooter, with the most random bullet accuracy, crappy animations where it looks like ppl still run n gun on perfect setups. so bad their poster boy tenz repeated these horrible aspects of the game probably 20 different times and didnt even wanna play. and lately Riot is trying to make gunplay more 'accurate' and nerfing every character in a update a few months ago. these are things you can visibly see and notice if you arent under the spell of a gaming studio. what about the stuff that goes on in the backends of these games like call of duty?

in terms of my initial claim, i still think its pointless to tweak a system beyond debloating and common optimization because of everything i mentioned. lets say youre better than me and have less input lag or better specs, better fps, faster mouse, better everything, but i still perform better than you on average, i place higher than you, or that my games stats imply im better. what then? thats when you get ego, you try to buy an even better mouse or tweak your system even more. some people try to defend the gaming studio if i try to criticize the game. why is this the state of mp/comp mp gaming? because it keeps both of these type of people playing.
have you ever noticed that sbmm feeling cranked in games is like a normal thing now? and most of the time feels more like ranked than the actual ranked mode? you get better teammates+more fair games than the ranked mode. but everyone in ranked just wasting their time fueling their egos because gaming is a money rake more than ever. ranked isnt like bo2 ranked anymore. now everyone wants to be pro, ranked is a form of keeping people hooked now. you do that by finding ways to fuel the ego thru a fake ranking system. how can some games not fix smurfing? its literally the most simple thing ever yet you play against duos or stacks against people actually trying, but most of the time your team has a boosted player, no one calls out with you, teammates troll while youre being chill trying to win a game. thats not true ranked or testament of skill, thats just matchmaking retention. and that duo or stack just rank up thinking theyre the goats, creating more boosteds and smurfs. then those people tell others to just get better or that its a skill issue, the same people tweaking tf out of their system, trying to get the next best peripheral etc.

dude i understand this was a dump but the comp mp/tact shooter gaming community is cringe now and they are the louder voice now which ultimately makes all these studios just keep making complete garbage games while boasting about how they care for competitive integrity or this or that, and these people gobble it up. all the fortnite kids who used to crank 90s trying to go pro thinking theyre the shit with their egos are the majority, and the ones who are brainwashed. which again, tells these studios to keep making these shitty games with rigging, manipulation - all for extra money because any valid criticism just gets shut down from all the other players who just mindlessly play the game to fuel their ego. tweaking a system wont do anything and it wont help you get closer to pro. having relatively new and decent hardware will do enough. debloating the system is good enough for any system. if your system is old, you can tweak your system to the moon, but it still isnt gonna make you better. having a good system/setup is a good baseline but that is all.

[soul4kills]

Your point wasn't lost on me. You really weren't specific to anything particular so I used a general example.

But in either case. I play to have fun, not to be better than anyone else. I don't have an ego like most gamers. But the reason I optimize is this. I hate having that feeling where when I shoot someone, i'm pretty sure i had that headshot, but somehow it wiffed. After my optimizing journey. I finally figured out what works for me and what the problem is. I honestly have zero latency issues now. And I never get that feeling anymore when I play a game. Now when I miss, i know for sure it was because of my bad aim.

That's just how good my latency is now and it feels nice. I honestly don't care if I get shit on or play bad. My reasons for gaming/optimizing is not the same as everyone elses. Here are the things I've done to optimize on a stock windows 11 system. 1 Registry tweak, just a handful of bios settings changes, disabled devices not being used in device manager, and using my new app. That's it. All the other stuff people do is overkill and probably hurting them rather then helping. I have a 2060, not the top of the line either. My system is pretty low aside from my 5900x. I optimize to have a better gaming experience, not to gain an edge or be better than someone.

And back to your point. Skill is always a factor no matter the level of RNG developers implement. Althought higher levels of RNG makes it less fun. In that situation just don't play the game. Find one you enjoy. Stop forcing yourself to play a game you can't enjoy.

[zeasty09]

Your point wasn't lost on me. You really weren't specific to anything particular so I used a general example.

But in either case. I play to have fun, not to be better than anyone else. I don't have an ego like most gamers. But the reason I optimize is this. I hate having that feeling where when I shoot someone, i'm pretty sure i had that headshot, but somehow it wiffed. After my optimizing journey. I finally figured out what works for me and what the problem is. I honestly have zero latency issues now. And I never get that feeling anymore when I play a game. Now when I miss, i know for sure it was because of my bad aim.

That's just how good my latency is now and it feels nice. I honestly don't care if I get shit on or play bad. My reasons for gaming/optimizing is not the same as everyone elses. Here are the things I've done to optimize on a stock windows 11 system. 1 Registry tweak, just a handful of bios settings changes, disabled devices not being used in device manager, and using my new app. That's it. All the other stuff people do is overkill and probably hurting them rather then helping. I have a 2060, not the top of the line either. My system is pretty low aside from my 5900x.

And back to your point. Skill is always a factor no matter the level of RNG developers implement. Althought higher levels of RNG makes it less fun. In that situation just don't play the game. Find one you enjoy. Stop forcing yourself to play a game you can't enjoy.

Can you tell us the exact "1 Registry tweak, just a handful of bios settings changes" in details? what you changed in reg and what you changed in bios?
And about your app? what should be done exactly? just run and play? or you adjust something? or it's only benchmarking?

[soul4kills]

Can you tell us the exact "1 Registry tweak, just a handful of bios settings changes" in details? what you changed in reg and what you changed in bios?
And about your app? what should be done exactly? just run and play? or you adjust something? or it's only benchmarking?

Actually 2 registry tweaks, the "GlobalTimerResolutionRequest" to on. Which can be enabled with my app. Then setting "Win32PrioritySeperation" to 41 in decimal.

Bios, the only ones related to latency are disabling "SVM" & "IMMOU" which are related to virtual machine stuff. SVM is needed by some anticheats as it's used by core isolation, faceit for example. Disabled onboard ethernet and audio. I use wifi. Disabled spread spectrum. I also underclocked by cpu, but that's just preference. But with my tool, I found out my underclock was unstable at -25, and put it at -20. Now it's more stable. Unstable underclocks/overlocks will affect your latency. There are others, but it's unrelated to latency.


Device manager stuff. This is all the stuff I disable. DO NOT RUN THIS SCRIPT ON YOUR SYSTEM. Your device paths will not be the same as mine. You gotta pick n choose which one applies to you, depending on how you use your system, you may not want to disable some things, so do your research. So this is a part of a larger script I run after any windows updates. Windows likes to reset things after an update.

Code: Select all

ECHO Sony Hands-Free (System Devices)
pnputil /disable-device "BTHENUM\{0000111E-0000-1000-8000-00805F9B34FB}_VID&00010094_PID&0004\A&594E843&0&042144F78C3F_C00000000"
ECHO Airpods HandsFree (System Devices)
pnputil /disable-device "BTHENUM\{0000111E-0000-1000-8000-00805F9B34FB}_VID&0001004C_PID&2014\A&594E843&0&141A97475D7B_C00000000"
ECHO A-Volute Bullshit - Nahimic
pnputil /disable-device "SWD\DRIVERENUM\{CE86418F-53D0-B4E5-36B3-3D0907CFA3B4}#AVOLUTE_NH3APO&6&501DF3C&0"
ECHO Bluetooth Device (Personal Area Network)
pnputil /disable-device "BTH\MS_BTHPAN\9&2F5861F2&0&2"
ECHO Nahimic Audio Bullshit
pnputil /disable-device "SWD\DRIVERENUM\{CE86418F-53D0-B4E5-36B3-3D0907CFA3B4}#AVOLUTEMIRRORINGVAD&6&96DD16D&0"
ECHO NVIDIA USB Type-C Port Policy Controller (System Devices) - NonExistent USB Port 
pnputil /disable-device "PCI\VEN_10DE&DEV_1ADB&SUBSYS_20673842&REV_A1\4&2283F625&0&0319"
ECHO Graphics Card NonExistent USB Port (System Devices)
pnputil /disable-device "PCI\VEN_10DE&DEV_1ADA&SUBSYS_20673842&REV_A1\4&2283F625&0&0219"
ECHO NVIDIA High Definition Audio (System Devices)
pnputil /disable-device "PCI\VEN_10DE&DEV_10F9&SUBSYS_20673842&REV_A1\4&2283F625&0&0119"
ECHO Microsoft Wi-Fi Direct Virtual Adapter (System Devices) - For being a hotspot or direct wifi connections
pnputil /disable-device "{5D624F94-8850-40C3-A3FA-A4FD2080BAF3}\VWIFIMP_WFD\7&34160812&0&11"
ECHO NDIS Virtual Network Adapter Enumerator (System Devices) (For Virtual Machine Stuff)
pnputil /disable-device "ROOT\NDISVIRTUALBUS\0000"
ECHO Composite Bus Enumerator (System Devices) For USB Devices that have multiple devices attached, Causes Frame Drops
pnputil /disable-device "ROOT\COMPOSITEBUS\0000"
ECHO Microsoft Kernel Debug Network Adapter (For systems admin debugging remotely)
pnputil /disable-device "ROOT\KDNIC\0000"
ECHO Microsoft Virtual Drive Enumerator (For handling virtual drives like mounting ISO's or Virtual Machine Drives)
pnputil /disable-device "ROOT\VDRVROOT\0000"
ECHO System speaker (Makes those Beeps and Boops)
pnputil /disable-device "ACPI\PNP0800\4&1096A61D&0"
ECHO Microsoft Hyper-V Virtualization Infrastructure Driver (Virtual Machine Stuff)
pnputil /disable-device "ROOT\VID\0000"
ECHO Remote Desktop Device Redirector Bus (For Remote Desktop Connections)
pnputil /disable-device "ROOT\RDPBUS\0000"
ECHO Microsoft Wi-Fi Direct Virtual Adapter (For Wifi Direct Connections/ Hotspot, Screen Mirroring)
pnputil /disable-device "{5D624F94-8850-40C3-A3FA-A4FD2080BAF3}\VWIFIMP_WFD\7&3BE7C3D&2&15"
ECHO TAP-NordVPN Windows Adapter V9
pnputil /disable-device "ROOT\NET\0000"
ECHO OpenVPN Data Channel Offload
pnputil /disable-device "ROOT\NET\0001"

And that's it.

As for my app, yes you just turn it on and play your game. Preferably you can use it on pause, but if you want to see how good your kernel is responding, set sample rate to 1000ms - 250ms. And that's it. You can benchmark by pressing b. Do it when you have a game on. And it will find the most stable timer resolution for you. Then you just play with better response times. But the main purpose of the tool is to provide a visual for you too see how your kernel is responding. To tweaks, or to having certain apps open. For example having one drive on will give me really high deviations. And if you don't understand graphs, i'm not going to explain it, go back to school.

Edit: I forgot, I use NVCleaninstall. First selection, legacy control panel, hd audio, physx. Second selection, Disable Installer Telemetry & advertising, Unattended Express Installation, Perform a Clean Installation, Disable Ansel, Show Expert Tweaks, Disable Nvidia Container, Disable Nvidia HD Audio Device sleep timer, Enable Message Signaled Interrupts (Policy: Default, Priority: High), Disable HDCP.

Edit 2: I recently discovered that cppc enabled & cppc preferred disabled is the ideal bios setting for my AMD dual ccd 5900x. It's specifically for dual ccd's. Can also be advantageous for other cpu types. Main gain here is cppc preferred causes windows scheduler to favor and assign processes to the 2 best cores. This causes those 2 cores to get overloaded and suffer constant context switching. This can also cause processes to jump cores a lot because they are constantly competing to use the 2 best cores. This causes processes to migrate to the other ccd requiring recaching memory to the new ccd. CPPC prefered disabled prevents this by utilizing all cores not favoring any core. Resulting in less context switching, overloading, ccd migration and competition for cpu time. It also reduces hotspots since the workload is spread across all cores rather than the 2 best cores. Which in turn minimizes constant fan ramping. As when the load is only focused on 2 best cores, the heat spread is lessened resulting in constant fan ramping to dissipate that hotspot. It's easier to dissipate heat when it's spread out rather than being localized to 2 distinct areas.

It also solved my periodic frame jitter problem. Using the blur busters mouse test https://testufo.com/mouserate . I used to get consistent large gaps in my mouse reports. This fixed that jitter for me. I'm really happy I finally figured that out. It was the last and only performance problem I had with my machine that I gave up on for the longest time.

[soul4kills]

Final Update:

-Added Power Plan switching with [k]
-Reduced the overhead even more by removing things from the main loop
-Added an about/help page
-Added a section for custom startup/benchmark settings in the batch file
-Fixed the buckets, still had bad mappings in the higher numbers

That's it. I think it's pretty polished now.

Reminder for people who don't understand the purpose of the tool. It doesn't fix your latency issues. It's a tool to help you visualize the problem if you do have latency issues. You can use it to test if the solutions you tried work or not. After you've solved your latency issues, you can improve your latency even further by setting your global system timer to 0.500ms with the app. Allowing for a better response time from your cpu/kernel.