Random frametime spikes 50 - 200 ms causing stutter

[Jerg3qhaerth]

I have searched many threads and tried many solutions, yet nothing seems to fix it for me. I wanted to make my own thread as a last hope since i feel lost on where the problem is. What i do know is that my frametime spikes mostly are between 50 - 200 ms and that is the stutter i feel. Everywhere from 5 seconds to a minute apart. I believe its mostly when i walk around and load stuff but im not sure. i will have my msi afterburner below with both moving and standing still scenarios.

It stutters these games:
Ark
jedi: fallen order
Assassins creed origins (if i remember correctly)

It does not stutter LEAGUE OF LEGENDS so it's not in all games but most games.

What iv'e tried of the top of my head:
- bunch of nvidia control panel settings
- framerate limiting
- unpluging usb 3.0
- setting to high performance in: Task manager, power options and in graphics settings.
- overclocking gpu
- stopping background apps via group editer
- vsync
- different graphics settings in game
- fullscreen, windowed, borderless windowed
- switching microsoft accounts
- reset computer
- disabling full screen optimization
- reset bios to default


I have not tried switching out any pc parts, i believe that is the "hardware" im refering to?
i don't want to waste money switching out parts without knowing what the problem is.


specs: (i don't know what more to put or how to find it, but more info will probably be on userbenchmark)
Intel(R) Core(TM) i5-8600K CPU @ 3.60GHz 3.60 GHz
32,0 GB ram
Nvidia geforce GTX 1060 6GB

USERBENCHMARK: https://www.userbenchmark.com/UserRun/59484010

There are some red color i dont like to see, but i don't know if it's the cause of stuttering or what it means really.
The pc runs pretty well other than the stutters, it can reach at least above 60 fps with high settings in most games.

Here is my msi afterburner where you can see the frametime spikes im talking about. the last bit that is more flat but still with some spikes, is where i STOOD STILL. Game: ARK / Settings: MEDIUM

Let me know if you need more information about anything, and probably how to get it if it's system or specs or somthing like that.

Thanks to anyone who helps

 

[ehdbom]

I am new to this forum and while I do know my way around computers in general this one may be out of my hand.
And yes hardware is refering to the physicial components of your pc such as CPU, GPU, RAM and so on.

The graphs from MSI Afterburner seems mostly fine to me except the framtimes and fps of course. I am not 100% sure about the power and voltage limit graphs and if that is normal, but I think it is that is my only concern as far as these graphs. But as I said that those graphs are out of my ball park. As far as User benchmark it is pretty wildely regarded by most enthusiasts to not be the best or that usefull. From what I gathered from it, it seems OK but it doesn't really provide much expect the components of your PC and that they are operating not great but fine. I cannot really make any assumptions based on the benchmark as Userbenchmark in general isn't a good indicator.

But I have some suggestions that you may not have tried yet:
1. Reseating the CPU, to make sure the pressure onto the socket is applied correctly.
2. Updating the BIOS of the motherboard and perhaps CPU chipset driver as well
3. Checking the power connections of your PC components making sure that they are fully seated.
4. Doing a clean install of Windows of a USB drive, if you haven't done this I highly recommend trying this!!
5. Check your disk usage during load, if your disk where Windows and/or the game is installed at is at 100% or close to 100% (in task manager) something is most likely either wrong with your Windows install or the disk itself. If you have a very cheap or budget ssd try using a regular hard drive instead if you have one available, some cheap ssd are prone to fail or are in some cases much slower than regular hard drives and can cause stuttering.
6. Try manually defragmenting your hard drive(s), do not do this on an SSD!
7. Run tests on your hard drive/ssd. I am not an expert in diagnosing bad drives but I think I can recommend running CrystalDiskInfo to see what is says about the drive health and running a benchmark on CrystalDiskMark and comparing it to the same drive.
8. While I doubt this won't do much but you could also try a memory test with e.g. memtest86 to see if your RAM is working as it should.

 

[Mussels]

Does this happen with your network cards completely disabled, in offline testing?

network traffic, mechanical hard drives, wallpaper changes, specific network cards (wifi cards scanning for new channels) and so on can all cause this sort of problem, you can even get spikes from hitting Vsync, or VRM's without a temperature sensor overheating

Userbench is terrible and should not be used, CPU-Z and GPU-Z will give you most hardware information you need and HWinfo will give you the realtime readings (and things like minimum/maximum/average) while figuring things out

 

[Deleted member 185158]

Looks like the game is using mad virtual memory. Even a bit more than the physical memory being used and plenty left.

Loading textures off the hard drive maybe?

 

[Mussels]

Edit: I'll put your next steps here at the top since theres a lot of text below

You should be able to:
1. run HWinfo in sensors only mode and see VRM sensor temperatures
2. run cinebench R23 in the 30 minute test
3. Add some GPU load on top (Unigine Heaven is a simple test here, just make sure it's got no FPS limits/Vsync slowing it down)
4. Check those VRM temperatures, CPU clocks and effective CPU clocks in HWinfo, and post screenshots of the results. One youtube commenter said he had a sensor that read 0c or 127c, basically a yes/no for throttling and not of any actual value. You'd need a temperature probe or stick your finger on the VRM heatsink/PCB areas if thats true.
(The arrows in the bottom left of the sensors window allow you to add multiple columns, a fullscreen window lets you show *ALL* of these in one image)

5. If you see the expected VRM overheating, you can fix it with a 40mm fan or two and new thermal pads.
6. You can also adjust/reduce the overclocking you're using to lower the wattages, but if the thermal pads have dried out you'll need to lower settings a LOT to compensate, so replace those first.
My ROG Strix x570-E chipset pad didnt even last a year before it was dust and overheating.

Tweaktown had no issues with their CPU overclocking, but they also ran some extremely high perfomance, high airflow fans that sound like jet engines directly on them

During ALL TESTS, fans above the VRM that cool the CPU cooler's (Corsair H110i) radiator are turned on to high (12v)

The multi-core performance was enabled in the BIOS it was shipped with (0402), but after we flashed to the new BIOS (0605), MCE wasn't on by default.

The boards launched with a feature that made them perform better but added more heat to look good in reviews, they disabled it in the future (to keep the boards alive and reduce warranty claims, i guess)

*Original post below*
What's standing out from these graphs is the CPU usage spikes
The GPU usage is dropping at the same time the CPU usage goes up - it's not at EVERY spike, but there are a few of them

If this was stutter from say, loading from a mechanical drive - the CPU usage should lower, too.

CPU usage is dependent on CPU performance so something like VRM throttling on the motherboard can cause the CPU's performance to drop, making the usage appear higher (If it throttled to 1/2 speed, usage would double)

We see the clock speed say it's at a flat 4100Mhz which is below the maximum turbo for this CPU, makes me think it's a conservative all core overclock since it never varies at all. Afterburner says the CPU is only using 31W - 67W which are definitely low wattage values, but we arent seeing VRM temperatures so it's possible thats the issue

I had to click that userbench link to get the motherboard model, seriously - post the full specs next time and dont give UB the web traffic. They're outright lying scum.

Tweaktowns review of the board shows it's got a misleading 8+2 VRM config, which means it's really only a 4+2 - 4 phases could be overheating, since al all core overclock is going to prevent the CPU from lowering its clocks to cool down.

ASUS ROG STRIX Z370-G GAMING (WI-FI AC) (Intel Z370) Review (tweaktown.com)

The ASUS Digi+ PWM controller labeled ASP1400 is used as the main PWM controller in a 4+2 phase PWM output mode. It's most likely from Richtek or UPI Semiconductor as the drivers seem to be Richtek drivers

Mobo PCB Breakdown: ASUS Strix Z370-G Gaming - YouTube
"A watered down board that includes far too much water"
"Theres a reason they don't put ROG on the board, it'd ruin the name" (But then marketing put it on the box and website)


The video describes how they've made it LOOK like an 8 phase, but as they run in pairs they truly aren't seperate phases - it's a way to use two shite components and have them work better than four shite components, but still nowhere near as good as 8 actual phases on quality boards

at the four minute mark he then states "It gets worse" and gives the sad laugh of the terminally depressed at what he's seeing

Mobo PCB Breakdown: ASUS Strix Z370-G Gaming

at 10 minutes he states "If you get the motherboard and dont put direct airflow on the VRM's, just dont put an 8700K in it" - that refers to stock settings, but with an all core OC or even just MCE being enabled on your 8600K, that same rule would apply.

I don't understand everything he says, but makes some simple claims like the low-end mosfets in those pairs are 4.3 million Ohms (mOhm) and "quality ones start at 3 and get lower" with 'proper ROG boards' being around 2
(I'm not familiar with what he's talking about, but it's RDS - the resistance of the components. The higher the resistance, the more heat they output at a given wattage)

He gives the specific example of an 8700k at max load and math behind it all, but his summary there is simple - they're crap and prone to overheating.
He gave an example of a 1.42v 140A CPU setup (~199W) putting as much as 27W per phase due to the ineffeciency of them - and thats a lot of heat in such a small space (the small components to the right of the ones labelled with the numbers, these)

I'm sure the math doesnt scale linearly, but your 65W spikes could be pushing 7W+ out of each of them, or even more or less if they're poor with scaling - with no airflow from a stock cooler and aged thermal pads, I can see that overheating. It only takes *one* of them overheating to throttle them all down.

In the video he comments on how running the board near its limits for a year or two is going to run into VRM temps above 100C, degrade the thermal pads and long term possibly start damaging nearby components - the nearby capacitors labelling indicates they're only rated for 5000 hours at 100c - meaning long term exposure to the expected temps of those VRM's is going to result in a fast board death.


In so many of these threads about micro stutter, we're finding more and more shitty boards misleadingly named and designed to appear like high end products supporting high end CPU's and overclocking - but in reality they just cant, especially when used with AIO cooling that doesnt blow air over the VRM's and a high heat GPU pre-heating them