Frametime spikes and stuttering after switching to AMD CPU?

[Montelo]

Been there, done that; still do
Though, the case of/for timer tweaking is more "responsiveness" and less about raw FPS. The difference seems academic, but they're 2 entirely different 'interests' within the Enthusiast market/community. Stuttering and High Input latency are far bigger issues, for anyone that's not in it merely for the tweaking/benchmarking (or sensationalism).

Resplendence Software - LatencyMon: real-time audio suitability checker

LatencyMon: real-time audio suitability checker

www.resplendence.com

More/Less, you run it to monitor+log latency for system processes, etc.
There will be a latency spike associated to some process or driver when your system 'hitches' (stutters), LatencyMon can tell you what it was (Or, at least the most-upstream process of- the cause)

Should I let it run alone or in the background while playing?

 

[LabRat 891]

i dont have a modular PSU

im assuming SoC and Curve Optimizer has something to do with overclocking? if yes i didnt do that

If no then i sadly have no idea what this means

No problem. My comments about extensions and bent-over+frayed and/or worn/mis-seated power cables remain (potentially) relevant.
Though at this point, I'm not personally suspecting power as the issue. Your video of the spikes in CS:GO are nothing like my experiences w/ PCIE powerline issues.
All 2 or 3 times I've experienced PCIE powerline issues, the stuttering was almost rhythmic/regular, and very frequent.

Overclocking/tweaking/undervolting, yes. So, totally inapplicable in your case (I think?)

Having no idea means that you've likely not messed with any of those settings.

[Apologies if already gone-over] DOCP/XMP-settings are often* not 100% stable w/o SoC voltage tweaking.
Even though performance will suffer greatly, I would be troubleshooting w/ JEDEC 'auto' non-XMP/DOCP RAM settings.
*My experience across 5 AM4 boards, and 5+ AM4 CPUs, 2x APUs; 5 different sets of XMP/DOCP-supporting DDR4 3200-3600.

 

[Montelo]

Overclocking/tweaking/undervolting, yes. So, totally inapplicable in your case (I think?)

Having no idea means that you've likely not messed with any of those settings.

most likely yea .

Only thing im using is XMP i also tried disabling it because i thought maybe 4 Ram sticks with 3600Mhz are too much but it wasnt the case, i also tried a different kit with 2x8GB Kit with 3200Mhz same thing sadly

 

[LabRat 891]

most likely yea .

Only thing im using is XMP i also tried disabling it because i thought maybe 4 Ram sticks with 3600Mhz are too much but it wasnt the case, i also tried a different kit with 2x8GB Kit with 3200Mhz same thing sadly

IMHO, Continue any testing, diagnostics, etc. w/ only 2x sticks and no XMP/DOCP.

You would be correct to think 4x sticks are too much.
IIRC, 'rule of thumb' for running all 4 slots on AM4, is go a JEDEC-spec down, or few. ie, DDR4-3200 supporting CPU might only run 4x dual rank sticks at 2400 or 2133 w/o actively 'tweaking' the CPU and SoC.

(Off the top of my head on LatencyMon)
Run LatencyMon (press the lil play button in the top-left after opening) on the desktop after a normal bootup. Let it run, and see if anything spikes.
If still no spikes noted after polling the system awhile, try giving the GPU and system some load (something that did spike it) and see what spikes.

My system seems to be operating 'normally' (though not especially impressive).
Either of the "Highest Reported-" categories, is what you're looking for.

 

[Montelo]

IMHO, Continue any testing, diagnostics, etc. w/ only 2x sticks and no XMP/DOCP.

You would be correct to think 4x sticks are too much.
IIRC, 'rule of thumb' for running all 4 slots on AM4, is go a JEDEC-spec down, or few. ie, DDR4-3200 supporting CPU might only run 4x dual rank sticks at 2400 or 2133 w/o actively 'tweaking' the CPU and SoC.

Latencymon just told me this after 15 Minutes

 

[LabRat 891]

View attachment 404760

Latencymon just told me this after 15 Minutes

What was recorded as the "Highest Reported" latency process/driver?

 

[Montelo]

What was recorded as the "Highest Reported" latency process/driver?

this is what im seeing

 

[LabRat 891]

View attachment 404765

*boom* storport.sys is the Windows System Driver(?) for storage devices.
In my experience, and aside from missing/bad Chipset drivers, etc...

Spoiler: AMD B550 Chipset Drivers

https://www.amd.com/en/support/downloads/drivers.html/chipsets/am4/b550.html

Your drive's likely overheating, dying, or powersaving switching when it's not supposed to.

What NVMe drives are in the system? A malfunctioning drive need-not be the boot drive or game drive, to cause this sort of issue.

 

[Montelo]

*boom*
Aside from missing/bad Chipset drivers, etc....

You drive's overheating, dying, or powersaving switching when it's not supposed to.

these are my drives (sorry that it is in German)

2 M.2 SSDs
The one with 85% is a crucial one that
The other one is a Samsung 970 Evo Pro

What are too high temps for a Drive?


EDIT: CrystalDiskInfo doesnt say anything bad about the drives everything in is "Blue" condition

What NVMe and/or SATA drives are in the system? A malfunctioning drive need-not be the boot drive or game drive, to cause this sort of issue.

I tried to install Windows on both Drives once it didnt seem to make any difference really
First i used the Crucial one then the Samsung one, and now im back on the Crucial one same problem everytime

 

[LabRat 891]

View attachment 404766View attachment 404767
these are my drives (sorry that it is in German)

2 M.2 SSDs
The one with 85% is a crucial one that
The other one is a Samsung 970 Evo Pro

What are too high temps for a Drive?

IMO on NVMes, 70c is 'hot' over 75c is 'too hot', 80c+ is really not good.
That's CrystalDiskInfo, and it takes static readings (unless told to update every 60secs). Meaning, you're not getting an accurate picture of the temps while under gaming load.
For example: I've 'cooked' an NVMe from a combination of poor thermal contact w/ the mobo thermalshield, and getting heatdumped on by a 7900XTX.

Where are those drives physically installed into the board, and what cooling is on them?
-my roomate had similar happen because he didn't want to pull the plastic off the mobo shield's thermal pads; nearly killed 2x Optane P1600X drives...

What I'd do...
First, update/reinstall chipset drivers.
DL and install+run(as admin) https://www.hwinfo.com/download/ (sensors only)
leave it running while you start up and run a game. After a stutter or few, exit the game, and go look through the "maximum" column on the GPU, CPU, and esp. the SMART temps on the NVMe drives.

edit:

I tried to install Windows on both Drives once it didnt seem to make any difference really
First i used the Crucial one then the Samsung one, and now im back on the Crucial one same problem everytime

If one of the NVMEs are malfunctioning, it will 'lock up' Windows' storport driver, regardless of whether it's the boot drive or not.

IDK about Crucial, but the Samsung drive's firmware can be checked and updated with Samsung's management software.

https://org.downloadcenter.samsung.com/downloadfile/ContentsFile.aspx?CDSite=US&CttFileID=7242927&CDCttType=SW&ModelType=C&ModelName=MZ-V7S1T0B/AM&VPath=SW/202501/20250113152210689/Samsung_Magician_Installer_Official_8.2.0.880.exe

Semi-related:
I had 'an anomalous and intermittent performance issue' on my Samsung 990pro w/ Heatsink I put in my Gen3x4 10th gen Intel laptop, until I
-updated the drive's firmware
-verified full read/write load thermals
then
-disabled ASPM/'power savings' in Samsung Magician (it will warn you that the warranty is void if you overheat the drive w/ power savings disabled)

 

[Montelo]

IMO on NVMes, 70c is 'hot' over 75c is 'too hot', 80c+ is really not good.
That's CrystalDiskInfo, and it takes static readings (unless told to update every 60secs). Meaning, you're not getting an accurate picture of the temps while under gaming load.
For example: I've 'cooked' an NVMe from a combination of poor thermal contact w/ the mobo thermalshield, and getting heatdumped on by a 7900XTX.

Where are those drives physically installed into the board, and what cooling is on them?
-my roomate had similar happen because he didn't want to pull the plastic off the mobo shield's thermal pads; nearly killed 2x Optane P1600X drives...

What I'd do...
First, update/reinstall chipset drivers.
DL and install+run(as admin) https://www.hwinfo.com/download/ (sensors only)
leave it running while you start up and run a game. After a stutter or few, exit the game, and go look through the "maximum" column on the GPU, CPU, and esp. the SMART temps on the NVMe drives.

edit:

If one of the NVMEs are malfunctioning, it will 'lock up' Windows' storport driver, regardless of whether it's the boot drive or not.

IDK about Crucial, but the Samsung drive's firmware can be checked and updated with Samsung's management software.

https://org.downloadcenter.samsung.com/downloadfile/ContentsFile.aspx?CDSite=US&CttFileID=7242927&CDCttType=SW&ModelType=C&ModelName=MZ-V7S1T0B/AM&VPath=SW/202501/20250113152210689/Samsung_Magician_Installer_Official_8.2.0.880.exe

I had 'an anomalous and intermittent performance issue' on the 990pro w/ heatsink I put in my Gen3x4 10th gen Intel laptop, until I
-updated the drive's firmware
-verified full read/write load thermals
then
-disabled ASPM/'power savings' in Samsung Magician (it will warn you that the warranty is void if you overheat the drive w/ power savings disabled)

uhhh the Drives don’t really have any cooling

1 drive is under the CPU and the other one is beneath the GPU


I will try updating the chipset drivers and use HWmonitor tomorrow I can just send a Screenshot in here then im assuming?

And there I can see if a Drive is making any problems?

 

[LabRat 891]

uhhh the Drives don’t really have any cooling

Then... AFAIK, this is "expected performance". When the NVMe throttles to preserve its life, (while being actively accessed) it's like hitting a wall to the user.

1 drive is under the CPU and the other one is beneath the GPU

The GPU-occluded drive is probably cooking when the GPU is exhausting onto it (assuming, you don't have a blower card).
and
The CPU (boot, hopefully?) drive is probably throttling under load, esp. if it's not seeing active airflow from a downdrafting cooler.

I will try updating the chipset drivers and use HWmonitor tomorrow I can just send a Screenshot in here then im assuming?

And there I can see if a Drive is making any problems?

Do update drivers, check drive firmware, etc. but, I think you might have a thermal issue.

Run HWinfo(or AIDA64, or w/e) in the background while loading-up the system.
After some stutters, exit/alt tab out and go look through Maximums on temps, check for Thermal Throttling: "Yes" on anything, and also look to see if there's been any WHEA errors.
[Note: don't freak out if you see some really high or really really low readings from the mobo, etc. Often-times, those are errant readings, and can generally be ignored, after verifying their inaccuracy]



ESPECIALLY if you don't have a mobo-included thermal-spreader/shield/sink installed on the drives, you need (at least) something like these, on the NVMe drives.

https://www.amazon.com/dp/B0CR69GG3W

I know you're not US-based, but these have been generally 'sufficient*' for:
Optane P1600X
SK Hynix/KLEVV CRAS C910
Solidigm P41+
Intel Pro 7600p
Samsung 990Pro
and more...
I've not found a better deal, Stateside



*

Spoiler: NVME cooling minutia and experiences

The NAND and/or PCM on a given NVMe drive, likes to be warm. However, the controller is much less tolerant.
At the bare minimum, NVMe drives should have a 'thermal spreader' applied to them, to 'balance' heat from the controller into the NAND and the rest of the drive.
Most drives come with an aluminized and/or allotropic-carbonized sticker, to provide *some* heat spreading but, this is generally insufficient in my experience. -Especially, when the drive is actively being heatdumped onto by a GPU or, is otherwise starved of active airflow.

 

[Montelo]

Then... AFAIK, this is "expected performance". When the NVMe throttles to preserve its life, (while being actively accessed) it's like hitting a wall to the user.

The GPU-occluded drive is probably cooking when the GPU is exhausting onto it (assuming, you don't have a blower card).
and
The CPU (boot, hopefully?) drive is probably throttling under load, esp. if it's not seeing active airflow from a downdrafting cooler.

Do update drivers, but I think you might have a thermal issue.

Run HWinfo(or AIDA64, or w/e) in the background while loading-up the system.
After some stutters, exit/alt tab out and go look through Maximums on temps, check for Thermal Throttling: "Yes" on anything, and also look to see if there's been any WHEA errors.



ESPECIALLY if you don't have a mobo-included thermal-spreader/shield/sink installed on the drives, you need (at least) something like these, on the NVMe drives. View attachment 404770
I know you're not US-based, but these have been generally 'sufficient' for:
Optane P1600X
SK Hynix/KLEVV CRAS C910
Solidigm P41+
Intel Pro 7600p
Samsung 990Pro
and more...
I've not found a better deal, Stateside

I have this heatsink is This what you mean? I didnt had it installed though its for the M.2 under the CPU I installed it now

The other one should get some air i think since I have a a lot of Fans that should do some good airflow

 

[LabRat 891]

View attachment 404771
I have this heatsink is This what you mean? I didnt had it installed though its for the M.2 under the CPU I installed it now

The other one should get some air i think since I have a a lot of Fans that should do some good airflow

Don't forget to use the thermal pad, and remove the protective film first.

That's a step in the right direction (for diagnosing storport.sys spiking). Worst-case,
try that for a bit; stutters persist, remove the non-boot drive, and ensure the boot drive is cooled.

 

[Montelo]

Don't forget to use the thermal pad, and remove the protective film first.

That's a step in the right direction (for diagnosing storport.sys spiking) Worst-case,
try that for a bit; stutters persist, remove the non-boot drive, and ensure the boot drive is cooled.

the idle temp is a little bit higher now though with the pad in the other screenshot i sent it was 38 now its already at 44

 

[LabRat 891]

the idle temp is a little bit higher now though with the pad in the other screenshot i sent it was 38 now its already at 44

Being a 'spreader' primarily, this is expected.
It's under load that's the concern, and the heatsink/spreader will aid in evening-out temps across the mass of the NVME drive.

Without the thermal mass of the heatsink bonded to the controller, the controller will likely quickly 'heatsoak' and bounce off its thermal limits when doing Read/Write (shader caching, asset loading, texture streaming, etc.).



*something* is causing storport.sys to hiccup (really badly).
Drivers
Cooling
Firmware
are the 0/low-cost correctable likely causes.

Technically....
SoC/IoD on the CPU going slightly unstable for some reason can cause the whole PCIe bus to hiccup.
ASPM may be improperly implemented, causing the drive(s) to 'get stuck' between power states, or in a low-power state.
Even on rare occasion, a particular drive might just flat-out not like a particular motherboard or firmware release for that particular board.

I'm probably missing a few more 'out there' reasons for storport.sys lagging-out, too. You can just Search "storport.sys latency" anywhere, and see it's an issue w/ a myriad of purported causes.

 

[Montelo]

Being a 'spreader' primarily, this is expected.
It's under load that's the concern, and the heatsink/spreader will aid in evening-out temps across the mass of the NVME drive.

Without the thermal mass of the heatsink bonded to the controller, the controller will likely quickly 'heatsoak' and bounce off its thermal limits when doing Read/Write (shader caching, asset loading, texture streaming, etc.).

alright^^ thank you, ill give HWMonitor a try later/tomorrow since its already 5 in the morning here i need to catch some sleep haha, im gonna update this thread then after the results

This is what im seeing i dont know why it is in german i installed the app in English, i had some stutters ingame but the temp seems fine i guess?
Game is installed on my C Drive (that is the one beneath my GPU)

i noticed that this is red what does this mean?

 

[Apocalypsee]

alright^^ thank you, ill give HWMonitor a try later/tomorrow since its already 5 in the morning here i need to catch some sleep haha, im gonna update this thread then after the results

View attachment 404792

This is what im seeing i dont know why it is in german i installed the app in English, i had some stutters ingame but the temp seems fine i guess?
Game is installed on my C Drive (that is the one beneath my GPU)

View attachment 404795
i noticed that this is red what does this mean?

You let the program run in background while you gaming and that is the max temperature while you gaming I assume? That temperature is fine if you gaming.

That power reporting deviation is the way the motherboard telling the CPU how much power it consume, if its 100% then it reports the right power, less than that it consume less than it should which means the CPU can boost more.

We at loss here, if I were you I'll try to put the Power Management mode in nvidia control panel to 'Prefer maximum performance' and while you at it go to BIOS and disable CPU Boost, just wanted to sure that the fluctuated clockspeed isn't the one that causing the stutter.

 

[Montelo]

You let the program run in background while you gaming and that is the max temperature while you gaming I assume? That temperature is fine if you gaming.

That power reporting deviation is the way the motherboard telling the CPU how much power it consume, if its 100% then it reports the right power, less than that it consume less than it should which means the CPU can boost more.

We at loss here, if I were you I'll try to put the Power Management mode in nvidia control panel to 'Prefer maximum performance' and while you at it go to BIOS and disable CPU Boost, just wanted to sure that the fluctuated clockspeed isn't the one that causing the stutter.

yea these were the temps while i was playing cs2.
Im gonna try that with the Power Management mode.

There has to be a reason tho why this is happening it's so frustrating

Im gonna try that with the CPU boost aswell, what is the name of setting exactly and where do i find it

I tried fortnite a bit, the game is really stuttery for me

I dont need to say any timestamps should be easy to see

The first few seconds was my internet but you can see what stutters i mean in the frametime

 

[3x0]

Fortnite stutters for everyone, ignore it

 

[Montelo]

Fortnite stutters for everyone, ignore it

but that extreme?
my friends dont experience that much stutter

 

[evernessince]

Have you tried moving your OS to your other drive?

I've had similar issues in the past and it turned out that my motherboard didn't like a specific SSD I was using. Thought the machine was defective until I cloned the OS to another SSD and it worked flawlessly.

 

[Montelo]

Have you tried moving your OS to your other drive?

I've had similar issues in the past and it turned out that my motherboard didn't like a specific SSD I was using. Thought the machine was defective until I cloned the OS to another SSD and it worked flawlessly.

i did try that yea, same issues

 

[evernessince]

i did try that yea, same issues

When you tested the memory, did you only try 2 sticks at a time in the same slot? It might be worthwhile to try 1 stick at a time in different slots as it's possible an individual slot / trace is problematic.

but that extreme?
my friends dont experience that much stutter

I concur. Given the video you've provided, it seems the behavior is similar across multiple games which indicates to me that it isn't a game related issue.

 

[jaszy]

If your issue didn't happen on NVIDIA, its quite possibly a shader catching issue with the configuration of hardware being used. AMD and NVIDIA have different methods of compiling shaders.