Question 5900x Drops clock speed under sustained load

[_Peejay]

Hey all

I'm helping a friend troubleshoot a problem he has with his gaming PC. These are his specs:
CPU: 5900x
Mobo: Gigabyte A520M H
RAM: Corsair 2x8GB 3200MHz CL16 RAM
GPU: Nvidia RTX3070
Be Quiet!: 700W

Two days ago while playing Warzone 2 he suddenly started to encounter frame drops. He sent me a video of his task manager where the CPU was dropping from a normal clockspeed to about 500MHz and then jumping up again. This was happening about every 10 seconds. This only started to happen after about 5-10 minutes in game. With a reboot it was gone again for the first 5-10 minutes.

We’ve been troubleshooting and found that the anomaly is repeatable in Cinebench R23 and Prime95. For the first run of Cinebench, the CPU stays nicely at 100% usage. After about 2 minutes, the usage drops to about 10%¸ the clocks drop to 549MHz in HWmonitor and task manager. After that it shoots back up to 100%, back down, back up, ... The Cinebench score is way lower than it should be.

In Ryzen Master the CPU Power drops from 104W to 61W or even 15W. PPT drops from 92% to 20-60ish%.

I’ve searched all over but can’t really find an explanation why this is happening.
He removed dust from his PC before this started happening, so I asked him to remount his cooler, with no improvement.

Does anyone have a clue what is causing this issue?

 

[Misgar]

Is the CPU being overclocked (PBO etc).
What are the maximum temperatures of the individual CPU cores?
What CPU cooler are you using?
Is new TIM used after removing cooler?
What are the fan profiles?

If core temperatures are too high thermal throttling might be kicking in. The resolution of the screenshot is too low for me to make out the details.

 

[_Peejay]

Sorry for the resolution of the screenshots.

To aswer your questions:

Is the CPU being overclocked (PBO etc).
No, I tried turning it on but it makes no difference exept clocking a bit higher with higher temperatures.

What are the maximum temperatures of the individual CPU cores?
While stress-testing the CPU package hits about 74° max, idle it is running 60-65°.
Indivual core temps:

What CPU cooler are you using?
Auraflow x240,

Is new TIM used after removing cooler?
Yes, brand new paste has been applied!

What are the fan profiles?
The profiles look okay, fans are ramping up to 80-100% when the CPU is getting hotter

If core temperatures are too high thermal throttling might be kicking in. The resolution of the screenshot is too low for me to make out the details.

 

[Misgar]

Well, that all seems fine to me. It doesn't look as if thermal throttling is the culprit unless the BIOS is detecting very short duration high temperature peaks which we cannot see.

Is there anything odd in Windows Event Viewer?

Anyone else care to comment?

 

[_Peejay]

Yea it seemed fine to me as well..
Event Viewer looks fine, no major errors or warnings.

The strange thing is that it started so suddenly. He didn't update any drivers or BIOS. The only thing I can see that there were 2 Windows updates recently: KB5027215, KB5027538.

I'm helping him troubleshoot through AnyDesk and notice the PC is feeling normal, opening programs and general navigation feels fast as it should be. Nothing seems out of the ordinary. It only happens when you stress the CPU for a longer period (Cinebench or in game).

 

[Roland Of Gilead]

Hey there,

Which exact model PSU? Which bios are you running?

Is chipset driver and all system drivers up to date?

The A520 mobo, whilst it can run these chips is not a good solution. The VRM's aren't robust enough which could be causing throttling.

I'd ensure all drivers and bios are up to date, then try again. Be sure to clear CMOS after the bios update to ensure it takes properly.

 

[_Peejay]

Hey thanks for your response!

He is running the be quiet! System Power9 700W.
I can check if the drivers are up to date. Can they cause this sudden kind of behavior? Last week this was not happening.

Thank you for the clarification on the motherboard, he hasn't had issues up to now. Can it start suddenly due to outside temperatures for example? -> Heat wave in Europe.

We'll try your suggestions now.

 

[Misgar]

Does the motherboard have large heatsinks over the VRMs or can you see naked MosFETs next to chokes (inductors) and a bunch of capacitors?

If you don't have enough phases in the VRM area and the MosFETs go above 110 to 120C, you may encounter thermal throttling in the VRMs when the ambient air temperature is higher than normal.

Remove the side panel from the computer, point a large desk fan into the computer and set it to maximum air flow. Then repeat the stress test with forced air cooling.

 

[_Peejay]

No it does not have heatsinks, you can see it clearly.

He also has a smaller case, so everything runs a bit hotter.
The theory about the VRM's overheating and causing a throttle seems to be a logical explenation.

Went into the bios and disabled everything to do with PBO. In the last stress test I noticed this:

 

[Misgar]

Having just checked the Gigabyte web site I agree wholeheartedly with Roland of Gilead. The A520M-H is not a good choice for anything as powerful as the 5900X, regardless of what Gigabyte might say on their web site regarding CPU compatibility. The VRMs are indeed naked and will overheat with a 5900X in Europe during a heatwave.

As it happens I've just upgraded a similar Gigabyte A320M-H motherboard and installed a 95W 2600X CPU in place of a 65W A6-9500 APU. Since I'm an electronics design engineer, I decided to augment the VRM cooling with a "bodge" which seems to have worked.

Prior to installing the 2600X, I affixed small metal heatsinks on all the exposed MosFETs to reduce their surface temperatures slightly. I was forced to cut small sections out of some heatsinks to make them fit between the electrolytics in the VRM area, but the result works without fireworks or explosions.

There is no way this "bodge" is anywhere near as good as a high quality motherboard with double the number of VRM stages and proper heatsinks. I enjoy messing around with old systems and this is probably computer number 15 down from my most important build (7950X) so it doesn't matter if it all goes pear shaped.

A link to the heatsinks with double sided 3M adhesive tape is shown below:

https://www.ebay.co.uk/itm/256087593545

If you do decide to perform this mod, don't blame me if it all goes wrong and the computer dies. During the heatwave you might consider undervolting or some other method to reduce power draw. Alternatively, remove the side panel and aim a large desk fan at the VRMs. They'll be roasting underneath the NH-D15.

If the current CPU cooling does not direct a strong "breeze" across the VRMs, consider adding a high speed 80mm fan, pointing down over the MosFETs, but this will be very diificult with the NH-D15. Plenty of cooling is required to keep these VRMs below 100C.

By far the best solution is to buy a proper motherboard with decent VRMs.

Take care and good luck.

 

[Roland Of Gilead]

Hey thanks for your response!

He is running the be quiet! System Power9 700W.
I can check if the drivers are up to date. Can they cause this sudden kind of behavior? Last week this was not happening.

Thank you for the clarification on the motherboard, he hasn't had issues up to now. Can it start suddenly due to outside temperatures for example? -> Heat wave in Europe.

We'll try your suggestions now.

You're welcome.

How old is the PSU? About 5 years, right?

Yes, drivers can cause all sorts of issues. THe CHipset contain the Ryzen balanced power plan. If you don't have that option, switch it to Ryzen Balanced in the Power settings.

 

[_Peejay]

Having just checked the Gigabyte web site I agree wholeheartedly with Roland of Gilead. The A520M-H is not a good choice for anything as powerful as the 5900X, regardless of what Gigabyte might say on their web site regarding CPU compatibility. The VRMs are indeed naked and will overheat with a 5900X in Europe during a heatwave.

As it happens I've just upgraded a similar Gigabyte A320M-H motherboard and installed a 95W 2600X CPU in place of a 65W A6-9500 APU. Since I'm an electronics design engineer, I decided to augment the VRM cooling with a "bodge" which seems to have worked.

Prior to installing the 2600X, I affixed small metal heatsinks on all the exposed MosFETs to reduce their surface temperatures slightly. I was forced to cut small sections out of some heatsinks to make them fit between the electrolytics in the VRM area, but the result works without fireworks or explosions.

There is no way this "bodge" is anywhere near as good as a high quality motherboard with double the number of VRM stages and proper heatsinks. I enjoy messing around with old systems and this is probably computer number 15 down from my most important build (7950X) so it doesn't matter if it all goes pear shaped.

A link to the heatsinks with double sided 3M adhesive tape is shown below:

https://www.ebay.co.uk/itm/256087593545

If you do decide to perform this mod, don't blame me if it all goes wrong and the computer dies. During the heatwave you might consider undervolting or some other method to reduce power draw. Alternatively, remove the side panel and aim a large desk fan at the VRMs. They'll be roasting underneath the NH-D15.

If the current CPU cooling does not direct a strong "breeze" across the VRMs, consider adding a high speed 80mm fan, pointing down over the MosFETs, but this will be very diificult with the NH-D15. Plenty of cooling is required to keep these VRMs below 100C.

By far the best solution is to buy a proper motherboard with decent VRMs.

Take care and good luck.

Thank you for this information.

Sorry if it is confusing, I am troubleshooting for a friend, it is not for my PC. He is not using the NH-15, but a Auraflow x240. I have seen a picture of his setup and it seems that no air is directly going over te VRM's. Due to the recent weather, I am pretty sure we have found the culprate.

I'll try to get a slight undervolt running to stop the VRM's from overheating. I'll advise him to look into a beefy'er motherboard. In the meantime i'll ask him to point a fan at the VRM's.

 

[_Peejay]

You're welcome.

How old is the PSU? About 5 years, right?

Yes, drivers can cause all sorts of issues. THe CHipset contain the Ryzen balanced power plan. If you don't have that option, switch it to Ryzen Balanced in the Power settings.

He states that it is only been in use for about 2 years now. I'll try the Ryzen balanced power plan!

 

[Misgar]

My mistake, I spotted NH-D15 and forgot the earlier post said Auraflow x240. It makes more sense now.

 

[_Peejay]

I did a quick curve optimizer with -30 offset on it. We have a big fan blowing cool air directly on the VRM's.

Temperatures of the VRM's dropped by about 15°. Package temperatures dropped by about 8°.
No more weird drops in clock speed at the moment. Everything is rock solid and stable.

The motherboard will be replaced for a more sustainable solution.

Thank you to all contributors! 😉