Cant seem to figure out why my game is stuttering

[Mussels]

Thank you heaps, seriously that's giving us what we need to see.

Your CPU peaked at 131W (looks like PBO is enabled in the BIOS), maxing at 68C - those are typical, good numbers for a 2700x.

Your CPU was drawing 140A (amps, not watts) during that test - reddit was saying this boards VRM's can overheat past 100A.

System (aka VRM's) peaked at a flat 90C, which makes me think it's throttling - do you have any leftover computer fans you can stick right onto that?
If a fan blowing onto it doesnt lower the temps, that means the thermal pad is bad and needs replacing (literally may just need to be removed and placed back on, not bought new)


One method to confirm this is one of the few sensors you minimized, showing the average effective clock and core effective clock - if you're VRM throttling and the CPU is bouncing up and down, you'd see this lower than the max value

Reset the stats in HWinfo (the little clock icon, bottom right) after starting the cinebench multithreaded test - if these values dont match up then you know its throttling
You can see mine are all roughly the same at 4.4GHz, so no throttling (By tweaking PBO settings, i can get that around 4.6GHz, with 5.05GHz single threaded in gaming)

(You're missing a screw in that bottom photo, into the mobo standoff. Please try and fix that)

 

[BaiterSkater]

So this is what I've found under the VRMs. Idk if any of this is normal. I found moisture under the VRMs which is odd because i never spilt anything on this motherboard. Its quite a sticky substance. No idea what it is.

 

[Mussels]

Damn you're taking this serious, thank you
The moisture is a type of oil leaking from the thermal pads, its absolutely normal but also a sign of them being exposed to high heat - after a very long time (years) at high heat all the oil is gone, they dry out and go crumbly and need replacing

Yes thats normal for them to squish, you can replace them with higher quality ones if you want... for now flip it over, and try and get full contact on everything
Literally one piece not having full contact there, can cause the VRMs to throttle

 

[BaiterSkater]

Damn you're taking this serious, thank you

Yes thats normal for them to squish, you can replace them with higher quality ones if you want... flip it over, and try and get full contact on everything?
Literally one piece not having full contact there, can cause the VRMs to throttle

Wait you want me to turn it around or keep it the same way it was on the board? I just wanna make sure i hear you correctly.

 

[Mussels]

Wait you want me to turn it around or keep it the same way it was on the board? I just wanna make sure i hear you correctly.

Carefully and slowly peel off the thermal pad (try not to stretch it too much), then turn it over so the indented side is on the heatsink, and the good side is on the motherboard
The contact on the mobo components is whats critical, and the fresh flat side will do that for us


If you measure it while you're there, you can find replacements online to upgrade it later (0.5mm, 1mm, 1.5mm are common)

Edit
'Expensive', top tier example: 15W/k (bigger number transfers heat faster, stock ones usually around 5-7)
GELID SOLUTIONS GP-Ultimate 120×20x 1.0mm Thermal PAD Single Pack

 

[BaiterSkater]

So i just booted after flipping the pad over on my heatsink and noticed my core speed is jumping around alot on cpuz. Is this something weird ?

Edit:
The core clock speed on hwinfo is jumping from 3.1 ghz to 4.1 ghz. Is that something to be concerned about as well?

 

[tabascosauz]

The core clock speed on hwinfo is jumping from 3.1 ghz to 4.1 ghz. Is that something to be concerned about as well?

No, that's just it boosting like every other modern CPU and Ryzen. HWInfo's "Effective Clock" metrics give you a better idea of what the CPU is actually doing. Tip - HWInfo often spazzes out, if it does that then maximize the window then resize it and hopefully the data goes in the right columns. You can also enable Minimize Sensors Instead of Closing in settings (right click tray icon > settings), then it'll keep logging even if you close the window.

I wouldn't go so far as to say the VRM on the B350M Mortar is trash, as it's more middle of the road than anything. It's just that

• The heatsink is so tiny that I don't know if you can even cool it down with active airflow
• Stock 2700X is allowed to draw 142W like a 12-core 3900X/5900X, you are asking wayyyyy too much of this board, a board that should strictly be limited to 65W CPU like the 2700 or 3700X
• Is that picture correct and you have 1 case fan and 1 CPU fan for airflow? If so you are really asking wayyyyy too much of this board

I'd just turn down the power limit to 90W by enabling Eco Mode in the BIOS. But you seem like you might be slightly BIOS-averse, not sure how willing you are. B350M Mortar should still have a visually up-to-date Click BIOS design, so there should be a search function where you can simply type in Eco Mode and enable it.

About throttling potentially causing performance drops, kinda skeptical. The RT8894A on this board is so dumb I don't recall reading any sort of overtemperature protection on its spec sheet, wouldn't expect the PK632BA to have any either. "Overheating" is generally around 100-115C for most discrete mosfets so while 95C is undesirable idk if I would pin the stuttering on it, throttling needs hardware that's at least smart enough to protect itself and I'm not sure the RT8894A is. High temps can negatively affect mosfet performance though I think though

 

[Mussels]

Sorry, was off in the real world teaching a new ryzen owner some fancy tricks.

CPU-Z only shows one core at a time, so when they go to sleep (as often as 1000 times a second!) the reading bounces to another core. Thats why HWinfo's min/max/average readings are better.
The board you have was designed BEFORE the 2700x was made. It's VRM cooling just cant keep up, and it was designed for the stock AMD cooler to be blowing air at it.

As long as the thermal pad is making good contact, airflow on the VRM's will actually work.
Step one is to get it as cool as possible, with the adjusted pad next up is just attaching a fan to it.

With the fan attached, we then see what BIOS settings we have available and tweak them to get the max power you can get, without throttling.

Hardware info showed you at 140A and 130W

We can simply limit those to 110A and 100W if your board has PBO options or ECO mode (it may not, sadly) and the problem could be solved. Otherwise, we may need a modest all core overclock like 4.2GHz at a low voltage, to keep those VRM's chill.

 

[cst1992]

Wow guys, I've been browsing forums like this and Tom's for over 10 years and this is stuff even I have never ever come across.

Do you think the VRM overheating could be caused due to the CPU cooler? That is a tower cooler which blows air front to back, not straight down(like Intel's and AMD's stock coolers). That could potentially cause low airflow around the CPU(like in this case) and cause the surrounding components to overheat. Just a speculation.

it was designed for the stock AMD cooler to be blowing air at it.

Yeah, this.

 

[Mussels]

I'm reading your BIOS manual now, and i'll try and post suggested settings as i find them


I'm unsure if you'll be using the 'normal' or 'advanced' views, so the locations of items may change.
The manual will also be outdated, things may have been added or moved.

Since the manual only lists Ryzen 1000 series features, PBO and ECO mode are not listed. If your BIOS has them, take a photo please

What i can see:

Game mode off. Thats a power booster, opposite of what we need.

OC page:
A-XMP on
CPU features sub menu:
AMD cool and quiet: on
Core C6 state: enabled

Wow guys, I've been browsing forums like this and Tom's for over 10 years and this is stuff even I have never ever come across.

Do you think the VRM overheating could be caused due to the CPU cooler? That is a tower cooler which blows air front to back, not straight down(like Intel's and AMD's stock coolers). That could potentially cause low airflow around the CPU(like in this case) and cause the surrounding components to overheat. Just a speculation.


Yeah, this.

VRM issues were very very common on low end AMD boards in the FX era, and it continued into first gen ryzen as board manufactures saw AMD as the budget option, and made the boards as cheap as possible, aimed at the lower wattage CPU's assuming they'd be used for light office work and such.

This B350M was designed for a 4 core CPU, but AM4 being universally compatible meant that suddenly the 140W capable 2700x could fit in there. MSI was slow to learn their lesson, they even have some of the worst x570 boards on the market

MSI are not universally bad, but they definitely released a lot of AM4 boards with garbage VRM cooling. Hardware unboxed has many videos on this.
Under the same conditions (All core OC'd 3900x) look at this madness:
MXI MAG x570 Tomahawk? 58c
MSI MPG x570 Gaming edge wifi? 105C
The MSI x570-A pro? 115C




So: whatcha got for cooling those VRMs, and what BIOS settings do you have available for Overclocking and power control? If the latest BIOS added the PBO menus, we're in luck
Edit: a 2 year old reddit thread says that yes, you should have PBO settings in there.
It'll be like Advanced, AMD overclocking, PBO - we want manual control
Maybe 100W PPT, 85A TDC, 110A EDC ?

Eco mode settings be more like
85W PPT, 75A TDC, 90A EDC

Wow guys, I've been browsing forums like this and Tom's for over 10 years and this is stuff even I have never ever come across.

Do you think the VRM overheating could be caused due to the CPU cooler? That is a tower cooler which blows air front to back, not straight down(like Intel's and AMD's stock coolers). That could potentially cause low airflow around the CPU(like in this case) and cause the surrounding components to overheat. Just a speculation.


Yeah, this.

People love finding an answer they approve of ("MSI is trash!" "AMD is bad!" "Red goes faster!") and never actually find the root cause of the problem.
Reddit, facebook etc love finding short easy answers that usually are simply incorrect, or only correct under specific circumstances.
This is 100% an overheating VRM situation, the goal now is to reduce temperatures by increasing cooling, AND reducing the load on the VRMs.

 

[BaiterSkater]

Okay so i disabled PBO, not sure where c6 state is and ECO IS.

 

[cst1992]

Check CPU power states.

I've not used an AMD BIOS before, so I can't help you specifically, but you should be able to find something in power configuration.

 

[Mussels]

View attachment 222135

Okay so i disabled PBO, not sure where c6 state is and ECO IS.

Global C-states is under a "CPU features" for me (and should be on - however some boards have C6 as a seperate option instead of all in one)
Eco is only available on some boards, with some CPU's, when PBO is on

What you have done there by disabling PBO, is drop that 140W to 105W (Run HWinfo and Cinebench again) - let's see the current/amperage/temps again

If PBO disabled doesnt cool things enough, we turn it back on and find the custom/manual options within it, and enter values ourselves

 

[BaiterSkater]

okay so this is hwinfo after I fixed all the settings i could in the BIOS. I only found a PBO option in my BIOS, sadly.

 

[Mussels]

Damn, system (VRM) is still at 90C - and still seeing 131W CPU usage
If you reset HWinfo after starting cinebench, then the effective clocks tell us a story: the minimum and average are not 3.9GHz, they're dropping off to impossible numbers as the cores shut down to save power (50Mhz and 1.9Mhz are not physically possible, so they're readings from a throttled core)


So PBO only has on/off, no advanced or custom menus?

If theres not, you're stick with either adding cooling to the VRM's, or an all core overclock designed to get those wattage and amperage figures down
We can see your board is running 1.352v for 3.9Ghz, which is more voltage than it needs there for sure - the final, imperfect option would be to try for a 4GHz all core OC with as low voltage as possible, and get the amps and watts under 100 and see how the VRMs go

 

[tabascosauz]

(50Mhz and 1.9Mhz are not physically possible, so they're readings from a throttled core)

Uh no, it's certainly possible, in fact it's normal behaviour for Effective Clock at idle for a properly idling CPU.........that's kind of the point of Effective Clock, and has nothing to do with entire cores being throttled. Total CPU usage is recorded at 100% so the CPU is being fully loaded as it should during Cinebench.

The problem here is that it's hard to tell whether 131W is:

• Normal behaviour for a 2700X in CB R23 (on Ryzen 5000 it's common to see a hair under 142W during R23 as it doesn't always max out stock power)
• "Throttling" from Precision Boost as the CPU gets hot but doesn't hit Tjunction (I can't remember if the boost algorithm was that smart back then for Pinnacle CPUs)
• VRM throttling (again, impossible to tell as iirc the controller is dumb, there is no PROCHOT EXT flag being tripped, and it's nowhere near 100C, etc.)

 

[Mussels]

Uh no, it's certainly possible, in fact it's normal behaviour for Effective Clock at idle for a properly idling CPU.........that's kind of the point of Effective Clock, and has nothing to do with entire cores being throttled. Total CPU usage is recorded at 100% so the CPU is being fully loaded as it should during Cinebench.

View attachment 222164

The problem here is that it's hard to tell whether 131W is:

• Normal behaviour for a 2700X in CB R23 (on Ryzen 5000 it's common to see a hair under 142W during R23 as it doesn't always max out stock power)
• "Throttling" from Precision Boost as the CPU gets hot but doesn't hit Tjunction (I can't remember if the boost algorithm was that smart back then for Pinnacle CPUs)
• VRM throttling (again, impossible to tell as iirc the controller is dumb, there is no PROCHOT EXT flag being tripped, and it's nowhere near 100C, etc.)

You missed an earlier post where i wanted him to reset the values while the test was running, so there'd be no idle results from before the testing
It's one method to see clock stretching

2700x is meant to be 105W with PBO disabled, that's an issue with running a zen+ on a first gen board - it doesnt support PBO properly.
I owned one of these MSI B350's as my very first ryzen board, sold it upgraded to an MSI B450 to get NVME support only to find they still did VRM's really badly. I still have that B450 as my test bench/spare parts. (This is still 4+2... but without even a heatsink)

Spoiler

You can find list upon list of tested VRM's and ranking the boards, and unfortunately many MSI boards like these rank at the bottom. (i lost the complete 300 series list, but this 450 has the same setup)
Motherboard VRM Tier List AMD | Tech Other | Carbonite

 

[tabascosauz]

You missed an earlier post where i wanted him to reset the values while the test was running, so there'd be no idle results from before the testing
It's one method to see clock stretching

I'm fully aware of that. OP clearly didn't listen to you, because the Average column for Effective Clock shows 2.6GHz. The CPU is running the test at ~3.9GHz without thermal throttling, so obviously there's some appreciable amount of idle downtime included in HWInfo's data to bring down the average clock over the logging period to 2.6GHz.

 

[BaiterSkater]

Im so confused..

 

[Mussels]

Im so confused..

That's okay, discussions happen that aren't always relevant and may go over your head... and i make mistakes too (which is why i edit my posts a lot)

With HWinfo, If you start it after cinebench testing begins and take a screenshot during the testing, the current (now) and average (while tested) results are helpful
If you start HWinfo before, or take the screenshot after those values are now including idle and may not be useful
(Things like max temperature are useful, always)

What you need to focus on:
1. Can you put a fan onto that VRM heatsink or not
2. Can you enable manual PBO settings, or is it just on/off
3. If you cant tweak PBO, lets do an all core overclock of 4GHz at say 1.1V, and run the HWinfo/cinebench testing again.

 

[BaiterSkater]

I put one of my case fans above the VRM to cool it off. I think PBO is just enable, disable, or auto. so now I should do a OC to 4 GHz at 1.1V in the BIOS and then run cinebench and show you guys the stats again?

 

[RJARRRPCGP]

You missed an earlier post where i wanted him to reset the values while the test was running, so there'd be no idle results from before the testing
It's one method to see clock stretching

2700x is meant to be 105W with PBO disabled, that's an issue with running a zen+ on a first gen board - it doesnt support PBO properly.
I owned one of these MSI B350's as my very first ryzen board, sold it upgraded to an MSI B450 to get NVME support only to find they still did VRM's really badly. I still have that B450 as my test bench/spare parts. (This is still 4+2... but without even a heatsink)

Spoiler

View attachment 222166

You can find list upon list of tested VRM's and ranking the boards, and unfortunately many MSI boards like these rank at the bottom. (i lost the complete 300 series list, but this 450 has the same setup)
Motherboard VRM Tier List AMD | Tech Other | Carbonite
View attachment 222170

That B450 motherboard is poop! (in the photo) Reminds me of the AM3 days and also looks like something I would expect on an A320-class motherboard. Looks at least almost as bad as an A68 motherboard!

 

[BaiterSkater]

My mistake, PBO does have a manual option in my BIOS. What should I set ?

 

[eidairaman1]

Wow guys, I've been browsing forums like this and Tom's for over 10 years and this is stuff even I have never ever come across.

Do you think the VRM overheating could be caused due to the CPU cooler? That is a tower cooler which blows air front to back, not straight down(like Intel's and AMD's stock coolers). That could potentially cause low airflow around the CPU(like in this case) and cause the surrounding components to overheat. Just a speculation.


Yeah, this.

No because my vrm on my board would be overheating then...

That's okay, discussions happen that aren't always relevant and may go over your head... and i make mistakes too (which is why i edit my posts a lot)

With HWinfo, If you start it after cinebench testing begins and take a screenshot during the testing, the current (now) and average (while tested) results are helpful
If you start HWinfo before, or take the screenshot after those values are now including idle and may not be useful
(Things like max temperature are useful, always)

What you need to focus on:
1. Can you put a fan onto that VRM heatsink or not
2. Can you enable manual PBO settings, or is it just on/off
3. If you cant tweak PBO, lets do an all core overclock of 4GHz at say 1.1V, and run the HWinfo/cinebench testing again.

Wow for this person to be able to pull a heatsink off the vrm tells me the name itself holds true (baiter)

 

[Mussels]

Yes! you have manual settings!



These are the ECO mode settings the newer boards have, try these