CSGO stuttering every 5 seconds mystery (ex. 5800x3d stutter topic)

[Mussels]

Sorry forgot to mention: The heat from the GPU is going to kick the boards VRMs into overheating, especially if the thermal pads have aged

A faster CPU can push the GPU harder, resulting in it throttling even faster again - definitely look into a PPT limit in the BIOS, this will harm MT performance but not gaming performance

I'd start at 100W PPT and work your way backwards 10W at a time if stuttering continues

 

[3x0]

Yeah, I'm willing to bet money it's not the motherboard. But hey if you're willing to try, make sure to limit the PPT like @Mussels said... don't need to rebuild the entire PC just to test this.

 

[tabascosauz]

Also, this is a great suggestion.

The stutters in wz2 I described above were caused by Fullscreen borderless that cod switched to automatically after a small update. I have enabled the Fullscreen Exclusive and the stutters are 99% gone.

Now i see fps drops in WZ2, but that is a whole other topic, since its game related. I have seen same setups with the identical fps drops. What a game to be playing

I noticed MW2 seems to be reasonably good at staying in Fullscreen. Only after the Season 2 update did it slip back into Borderless, which did cause some additional (noticeable) stuttering but easy fix.

MW19 was horrendous, and I got into the regular habit of checking window mode at every start.

Yeah, I'm willing to bet money it's not the motherboard. But hey if you're willing to try, make sure to limit the PPT like @Mussels said... don't need to rebuild the entire PC just to test this.

MW2 runs all-core for short periods when reinstalling shaders, but that's still only 100-120W tops, more like 90W for me. In actual gameplay it draws like 50 watts, for Warzone about 80-90W steady.

I agree VRM capability is not the problem here, recent COD is notorious for various forms of regular stuttering and frame drops. 90W is right in line with 3700X all core, so the 53C results for Gaming X in HWUB testing are appropriate.

 

[3x0]

@tonypivo had issues in CSGO as well, and the CPU doesn't use more than ~50W

 

[LabRat 891]

Trash motherboard


It's overheating the VRM's on the x3D, it's one of those terrible ones that cant even handle a 65W CPU
B550 Motherboards VRM Cheatsheet - Comparison by Hardware Unboxed : buildapc (reddit.com)

The "10+3" VRM layout is... not.
It's all phase doublers, so it's more like 5 low quality VRM's
View attachment 287084


To be clear, it's the worst in the entire roundup
These are the VRM temperatures on the back of the PCB
View attachment 287085

In certain setups it bounced around between worst and second worst place in heat, but that's also because it's one of the boards that lies oops i mean sorry "It's inaccurate" with power and temperature reporting, so it would report incorrect temperatures in software for the VRM's (hence the thermal probe for their testing) and would throttle the CPU in some tests, keeping itself cooler at the cost of performance



The *only* way to use that board and NOT have it stutter is to enable PBO, and then create your own eco mode with lower settings, specifically keeping the PPT values right down




With a 3900x and PBO limits enabled, heres how that board fit in the standard testing lineup - but that's not taking any performance issues into account
The 3900x had a 105W PPT, while if PBO is enabled - like trying to use the curve optimiser for the x3D - that's going to shoot up as high as 142w
View attachment 287086

oO
Highly insightful.

Guess my obsession for putting 'peel-n-stick heatsinks' even on boards that'll never see an OC, and my personal SOP to make sure there's always active airflow by/on VRMs isn't a bad thing.
Over the generations I noticed nearly all stock-coolers sacrifice ultimate BOM-cost/performance for ensuring airflow around the socket; clearly that's been for a reason.

 

[Mussels]

Yeah, I'm willing to bet money it's not the motherboard.

I take paypal.
That board is literally the worst of all B550 boards tested on every VRM testing i could find.

oO
Highly insightful.

Guess my obsession for putting 'peel-n-stick heatsinks' even on boards that'll never see an OC, and my personal SOP to make sure there's always active airflow by/on VRMs isn't a bad thing.
Over the generations I noticed nearly all stock-coolers sacrifice ultimate BOM-cost/performance for ensuring airflow around the socket; clearly that's been for a reason.

Exactly - the moment you change the cooling around the socket you disrupt the VRM airflow. Budget boards are designed with stock coolers in mind for low wattage CPU's - when you add more wattage AND remove the cooling, its a recipe for disaster.

There are some VERY very trash B450, B550 and x570 boards out there that cant even sustain 55W loads. Intel has similar problems, but this needs to stay on topic with relevant hardware to the OP - a lot of budget boards are made with the lowest end 4 core and APU setups in mind, and 'support' the higher end CPU's by throttling their speeds.

The moment PBO gets enabled, the stock wattages (~105W for most Zen3) jumps to 142W, and curve undervolting gives you higher clocks at that same 142W. Many guides and youtubers focusing on this miss that entirely since they're using top tier $500 USD motherboards, and not these budget ones. It's neccesary to manually enter settings, even if they're just the stock values.

 

[3x0]

I take paypal.

So do I. You're on 10 bucks?

 

[tabascosauz]

The moment PBO gets enabled, the stock wattages (~105W for most Zen3) jumps to 142W, and curve undervolting gives you higher clocks at that same 142W. Many guides and youtubers focusing on this miss that entirely since they're using top tier $500 USD motherboards, and not these budget ones. It's neccesary to manually enter settings, even if they're just the stock values.

Where are you getting these wildly off numbers? Stock 105W TDP parts since 2019 have always been 142W PPT. When you Auto PBO, only then do 2CCD parts usually land around 160-180W depending on board and CPU.

After an hour at sustained 142W the Gaming X lands at 73C which is just fine. The 5800X3D basically never exceeds 130W worst case in any normal usage, and basically never exceeding 100W in games even without Curve Optimizer. The overheating results were from the 1 hour 200W static OC test. Even the VRM on my Unify-X gets a bit lukewarm after that much current for an hour.

I understand that you don't like the low effort doubled lo-side 5 phase that the Gaming X has (I don't like it either), but you're really blowing this out of proportion. There has never been a B550/X570 board that struggles at 55W PPT - even unheatsinked 3-phase boards don't struggle with 55W loads unless you put them in an oven.

Have you ever played MW19 or MW2? Sometimes the stuttering isn't even client-side.

 

[Deleted member 185158]

Where are you getting these wildly off numbers? Stock 105W TDP parts since 2019 have always been 142W PPT. When you Auto PBO, only then do 2CCD parts usually land around 160-180W depending on board and CPU.

After an hour at sustained 142W the Gaming X lands at 73C which is just fine. The 5800X3D basically never exceeds 130W worst case in any normal usage, and basically never exceeding 100W in games even without Curve Optimizer. The overheating results were from the 1 hour 200W static OC test. Even the VRM on my Unify-X gets a bit lukewarm after that much current for an hour.

I understand that you don't like the low effort doubled lo-side 5 phase that the Gaming X has (I don't like it either), but you're really blowing this out of proportion. There has never been a B550/X570 board that struggles at 55W PPT - even unheatsinked 3-phase boards don't struggle with 55W loads unless you put them in an oven.

Have you ever played MW19 or MW2? Sometimes the stuttering isn't even client-side.

Before 2019... try 2017.

105w is at base clocks period.
Boost is enabled by default, 142w.

Mosfets are measured generally up to temps greater than 150c.

The information was pulled out of a 3 letter body part.

It's not the board, I have an additional 10$ for a bet.

 

[3x0]

@tonypivo Just to rule out other factors other than VRM, set PPT to a lower value in BIOS or PBO2Tuner. Don't be crazy to replace the board and rebuild your PC just because of some opinion on the net.

 

[Mussels]

Where are you getting these wildly off numbers? Stock 105W TDP parts since 2019 have always been 142W PPT. When you Auto PBO, only then do 2CCD parts usually land around 160-180W depending on board and CPU.

After an hour at sustained 142W the Gaming X lands at 73C which is just fine. The 5800X3D basically never exceeds 130W worst case in any normal usage, and basically never exceeding 100W in games even without Curve Optimizer. The overheating results were from the 1 hour 200W static OC test. Even the VRM on my Unify-X gets a bit lukewarm after that much current for an hour.

I understand that you don't like the low effort doubled lo-side 5 phase that the Gaming X has (I don't like it either), but you're really blowing this out of proportion. There has never been a B550/X570 board that struggles at 55W PPT - even unheatsinked 3-phase boards don't struggle with 55W loads unless you put them in an oven.

Have you ever played MW19 or MW2? Sometimes the stuttering isn't even client-side.

the 142W is if PBO is enabled - I've seen this on many, MANY boards now where the BIOS defaults lock the wattage down, but the second you enable PBO to access those menus the defaults are raised to "motherboard limits" which means "hahahah no limits at all"
Users trying to get the curve undervolt end up adding wattage instead of reducing it, making things worse

You may be using higher end boards where thats the default option, but i'm dealing with people bringing me these garbage tier systems from local stores and this is how they behave

This isn't me making something up, i posted links and images to the reviews where they literally discuss the performance issues plaguing these boards

 

[3x0]

@Mussels What 142W? CSGO that stutters for him barely uses 60W.

 

[Deleted member 185158]

@Mussels What 142W? CSGO that stutters for him barely uses 60W.

Wanna see a low end board kick ass??

2700X and B450m-a 4300mhz 16 threads, no vrm throttling. Prism cooler.

Screen shot taken in 2019, just sold the cpu (lid-less) here on the forums.

 

[Mussels]

@Mussels What 142W? CSGO that stutters for him barely uses 60W.

Mixing different conversations here: I literally stated these boards overheat at 55W, the BIOS tries to force more than the VRM's can handle - and i also clearly stated high heat GPU's push that to happen even faster

The rapid throttling on/off causes the stutter, the CPU slows down causing the GPU to slow down, things cool down - they speed up, they throttle again, repeat

 

[3x0]

I literally stated these boards overheat at 55W

They don't. Fact.

 

[Mussels]

Wanna see a low end board kick ass??

2700X and B450m-a 4300mhz 16 threads, no vrm throttling. Prism cooler.

Screen shot taken in 2019, just sold the cpu (lid-less) here on the forums.

View attachment 287198

That board has a considerably higher reputation and better VRMs
AM4 B450/X470 Motherboard Vcore VRM Tier List v1.3 - Google Sheets

They don't. Fact.

Prove it. Do what i did and find reviews of people testing the board(s) involved and do what the reviewers did by verifying performance and temperatures matched what you're expecting. I have, so if you want to disagree with me try backing that up with some form of evidence.

I'll find the specific review and model that had the 55W issue, this has come up in several threads (especially the zen garden thread) and theres literally thousands of posts and reviews in my browser history to find that exact one again - but there was a common garbage design used in B450/B550/x570 boards that genuinely was and is trash, and causes these problems.

One thread here had a TPU regulars boards VRMs explode on him, but i can never find the thread since the title was about his assumed problem, not the VRM's he discovered later

The only solution is a better board in every case, but power limits (and avoiding boards using the max boost) is the only option without spending extra money.


One x570 example:
When a 3900x and a 5800x/3D use the same max wattage with PBO, you get the same results
Some of the boards that ran cooler only did so because they were throttling and had the temp sensors placed further from the actual heat source and dedicated thermal probes were needed, as i linked previously in this thread about the exact board used by the OP

The same crap happens on the intel side too, and it happens with various brands before anyone tries accusing me of being a shill or whatever bizarre argument was coming next.
PSA: Don't Buy This Asrock Motherboard | TechSpot


LTT has a thread where they verify the true VRM's on each board and rate them accordingly
Motherboard VRM Tier List v2 (currently AMD only) - CPUs, Motherboards, and Memory - Linus Tech Tips

Don't buy any of these unless it's for an APU office PC.

Another great resource with the details of the VRMs
B550 VRM DB sheet - Google Sheets

People get massively misled by similar product names, thinking that they read a fantastic review on a particular product and buying something with a tiny difference in the name and assuming it's just the wifi or something that changed

The reality is marketing bullshit where a 6+2 board has 90A MOSFETS while the "12+2" has a doubler with trash, when you google those OnSemi mosfets they're just 46A




And as for the obvious fact people arent clicking the links and not understanding i'm providing just the teaser image to get you to actually click the links, heres the HUB video with subtitles

They also did follow up comparisons with comparing the boards against more models, where while it was the worst in the original video others managed to become even worse
They specifically cover in this video (and this screenshot) that boards with the exact same VRMs can perform wildly different, because of the heatsink design - and all of these tests are pure CPU load only, NO added heat from a GPU

And the final screenshot shows my entire point:
At the end of their roundup video
Of all these boards tested, only one throttled.
Only one.
Just the one.
Guess which one?

Yes, its an overclocked setup. No, it's wattage isn't that extreme - but again you add some heat from a GPU or higher than the 21c ambients they had, and you'll reach these temperatures easily.

 

[3x0]

Prove it.

You're the one making the stupid claim, you prove it

I have built two PCs with that exact same boards, one running a 3900x and the other 5600. Both run fine.

 

[Deleted member 185158]

That board has a considerably higher reputation and better VRMs
AM4 B450/X470 Motherboard Vcore VRM Tier List v1.3 - Google Sheets
View attachment 287201


Prove it. Do what i did and find reviews of people testing the board(s) involved and do what the reviewers did by verifying performance and temperatures matched what you're expecting. I have, so if you want to disagree with me try backing that up with some form of evidence.

I'll find the specific review and model that had the 55W issue, this has come up in several threads (especially the zen garden thread) and theres literally thousands of posts and reviews in my browser history to find that exact one again - but there was a common garbage design used in B450/B550/x570 boards that genuinely was and is trash, and causes these problems.

One thread here had a TPU regulars boards VRMs explode on him, but i can never find the thread since the title was about his assumed problem, not the VRM's he discovered later

The only solution is a better board in every case, but power limits (and avoiding boards using the max boost) is the only option without spending extra money.


One x570 example:
When a 3900x and a 5800x/3D use the same max wattage with PBO, you get the same results
Some of the boards that ran cooler only did so because they were throttling and had the temp sensors placed further from the actual heat source and dedicated thermal probes were needed, as i linked previously in this thread about the exact board used by the OP
View attachment 287200

The same crap happens on the intel side too, and it happens with various brands before anyone tries accusing me of being a shill or whatever bizarre argument was coming next.
PSA: Don't Buy This Asrock Motherboard | TechSpot


LTT has a thread where they verify the true VRM's on each board and rate them accordingly
Motherboard VRM Tier List v2 (currently AMD only) - CPUs, Motherboards, and Memory - Linus Tech Tips

Don't buy any of these unless it's for an APU office PC.

View attachment 287202

M-A. The very bottom Asus B450 board. Rev 1 actually. Still in use actually.

Bought it on a newegg combo! 550$ cpu, board and lpx garbage.

Had fun with that board, learned a lot about transistor leakage with it.

-30c, Tec cooled sustain 4ghz at 1.190 vcore.

Cpu high temp alert is 70c.
With a default system, cpu fan commanded 100%. User only able to raise this to 75c.

 

[3x0]

Yes, its an overclocked setup. No, it's wattage isn't that extreme - but again you add some heat from a GPU or higher than the 21c ambients they had, and you'll reach these temperatures easily.

Your point is, a user running an OCed 3900X with auto voltage applied @200W on a test bench will have the same experience of VRM throttling as a user running a 60W workload in a case with some airflow with a GPU adding a bit of heat... that's quite the stretch there.
I'm out of this discussion, since you're determined to run your asinine point into the ground. Waiting for OP to respond

 

[Huubsterr]

Also, I can't stand VSYNC; I set FPS limits if I need to.

Best tip ever!

I turned off vsync again and set (in-game) the max FPS on 5fps below my screens 165hz, and DMZ (MW2) never felt so smooth, no stutter anymore!

Thanks a lot!

 

[LabRat 891]

Best tip ever!

I turned off vsync again and set (in-game) the max FPS on 5fps below my screens 165hz, and DMZ (MW2) never felt so smooth, no stutter anymore!

Thanks a lot!

Glad to have helped.
I usually set it in-driver (AMD Adrenaline Settings, IIRC). I can't advise it, but on my Vega(s) I just set it globally to 140-144 fps. (yes, slightly under is 'better' for adaptive sync displays. ie. G-sync/FreeSync/VRR)

 

[Mussels]

You're the one making the stupid claim, you prove it

I have built two PCs with that exact same boards, one running a 3900x and the other 5600. Both run fine.

I did.

Baseless claims with no evidence are enough for me to not waste time on you any longer, but i'll still gladly deny any and all false claims people make on these forums that mislead end-users into bad decisions on shitty products. Any crap motherboard can be made to behave if you keep wattages low enough, but that doesnt make it a good board, or make it good with ENTIRELY DIFFERENT CPUS TO THE ONE THE THREAD IS ABOUT THAT USES MORE WATTAGE.

Best tip ever!

I turned off vsync again and set (in-game) the max FPS on 5fps below my screens 165hz, and DMZ (MW2) never felt so smooth, no stutter anymore!

Thanks a lot!

Always recommended to do this, many games have broken Vsync implementations, frame limits keeping your GPU under 99% usage will prevent frames being rendered ahead as well (the same as Nvidias ultra low latency mode), halving or thirding your input latency

 

[ev1lk1ll]

Good day to everyone.
I read the material of this forum thread and information on the Internet, made various settings, but it is not possible to solve the problem with the frame time in full-screen warzone 2 mode (it is constantly growing up and when it reaches 30 ms and above, while the FPS is 180-200, slowdowns begin, and the picture does not become smooth), I will describe the situation in more detail:
Recently updated the components:
Palit RTX 3080 updated to -> ASUS RTX 4080
Ryzen 7 5800X updated to -> Ryzen 7 5800X3D

In the bundle (Palit RTX 3080 and Ryzen 7 5800X3D), there was no problem with frame time in full-screen warzone 2 mode, it appeared in the bundle (ASUS RTX 4080 and Ryzen 7 5800X3D).
Unfortunately, there is no longer an opportunity to return to the bundle with Palit RTX 3080 for verification.

I tried to make settings, they did not lead to success:
1. Various versions of Nvidia drivers.
Install version 531.29;
2. Various BIOS versions of the motherboard (1.40, 1.60, L2.62).
Installation version L2.62;
3. I changed the settings in the game itself (- 5 FPS from the refresh rate of the screen, etc.) and Nvidia panels, as well as Windows settings (switches "Game mode" and "GPU scheduling with hardware acceleration");
4. Enabled and disabled the XMP RAM profile, the rest of the Bios settings are set by default;
5. Reinstalled the game;
6. I also noticed that there is no Nvidia DLSS and Nvidia Reflex item in the warzone 2 settings, I don't understand why they are not there, has anyone encountered this?

As a temporary solution, I set the screen mode to "Full-screen (without frames)", but with this mode the FPS is less by about 20-30 and the frame time is less stable and the picture is not as smooth.

Configuration of components at the moment:
ASRock B550 PG Riptide
АМD Ryzеn 7 5800Х3D
ID-Cooling FROSTFLOW X 360
A-Data XPG GAMMIX D20, 32 GB, 3600, [18-22-22]
ASUS TUF Gaming GeForce RTX 4080 OC Edition
Phanteks AMP 1000W [PH-P1000G]
SSD М.2 WD 1 ТB
SSD M.2 Kingston 500 GB
24.5" Dell S2522HG (240 Hz, G-SYNC)

If someone has ideas and recommendations that you can try again, I will be glad.
Thank you in advance.

 

[3x0]

In the bundle (Palit RTX 3080 and Ryzen 7 5800X3D), there was no problem with frame time in full-screen warzone 2 mode, it appeared in the bundle (ASUS RTX 4080 and Ryzen 7 5800X3D).
Unfortunately, there is no longer an opportunity to return to the bundle with Palit RTX 3080 for verification.

Do you remember which driver version you used with the RTX 3080 when there was no stutter?

Warzone is a very problematic game by itself, plenty of people experience stutters because of their game optimization.

 

[ev1lk1ll]

Yes, I remember, this is version 528.24.
With it, the performance in this wonderful game is no better, as with version 531.29.

Frame time graph from MSI AB:

Spoiler: image

The game is very poorly optimized, and it's strange to me why it worked pretty well in a bundle of RTX 3080)

For the sake of experiment, I'll download Warzone 1 the other day, see how things are there.