Gigabyte GA-AB350M gaming 3 VRM cooling

[SMOKEU]

I'm a little concerned about the VRM temperatures on my motherboard, as when the CPU is under heavy load, the VRM temperatures go just over 100°C, and that's in winter with a 21°C room temperature. I'd hate to think how hot they'd get when summer arrives.

The case has fairly good airflow with 3X Noctua 140mm fans that spin around 1000RPM under load. I was thinking of getting a VRM heatsink kit and using thermal tape to stick the heatsinks onto the VRM components. I've worked pretty hard to build a powerful, quiet gaming rig and I don't want to ruin that with a tiny, high RPM fan. If the temperatures are still excessive after I've installed the heatsinks, then I'll put a 40mm or 60mm PWM fan to blow air directly onto the VRMs.

Being completely new to this, I'd also like to confirm exactly which components I'm trying to cool. I know the main VRM is already cooled by the black/red/white heatsink on the top left of the pic. I was planning on installing the heatsinks on the components that are in the blue boxes. Are these all of the VRM components covered?

 

[Johan45]

What you have outlined are the chokes and shouldn't need any cooling. Where the heat comes from are these small chips I have circled in red. Those are the hi and low mosfets that deliver the power.

 

[SMOKEU]

What you have outlined are the chokes and shouldn't need any cooling. Where the heat comes from are these small chips I have circled in red. Those are the hi and low mosfets that deliver the power.

Thanks for that! Since it's a midrange motherboard with no MOSFET cooler mounting holes, should I use a thermal glue? I know this is permanent and will void the warranty, but it's a cheap board and I couldn't find any decent mATX overclocking boards locally so I'm stuck with this one.

Or would thermal tape be better? I'm guessing the glue would be more effective. I've got lots of room to mount a fairly big heatsink since there's a lot of empty space on top of the motherboard.

 

[trents]

You can also buy heat sink kits for mosfets that come with thermal adhesive pads: https://www.amazon.com/Enzotech-MOS-C1-MOSFET-Heatsinks-Pack/dp/B004CLDIHK

Just placing a spot fan to blow on the bare mosfets can also help.

 

[wingman99]

The MOSFETS without a cooler is not for the cores it's for the SOC. The watts used for the SOC (memory controller) is minimal. I would not worry about a heatsink, unless you want to do it for looks.

 

[trents]

Ah, good to know. I always assumed they were also for the cores. What does SOC stand for?

 

[wingman99]

Ah, good to know. I always assumed they were also for the cores. What does SOC stand for?

SOC (System on Chip) for AMD it is the memory controller and integrated south bridge controller.

 

[trents]

But wouldn't those mosfets get pretty hot when overclocking the RAM or when running a large amount of RAM?

 

[wingman99]

Overclocking the ram controller is about 3-4c maximum increase in temperature 24/7. Very minor amount of increase of watts.

 

[PetrolHead]

What CPU are you using with this motherboard and how do you know the VRM temperature goes over 100 C?

 

[demien88]

I don't believe that said 100C is correct sensors might be showing wrong values.

 

[PetrolHead]

The sensor is probably fine, but the software is likely not interpreting the data correctly. There have been such issues with ASRock boards, with "Motherboard" temperatures being over 100 C while idling on certain software unless the user used some other software first (in which case the reading dropped to near ambient temperatures). However, I doubt this board even has a VRM temperature sensor in the first place. My guess is the reading is again for the motherboard (or something similar), which in turn is most likely the chipset temperature. And the temperature is probably fine in reality, despite what the reading is.

 

[Buddywh]

I don't believe that said 100C is correct sensors might be showing wrong values.

Hi...I've got the same motherboard and it has a thermal sensor located in the midst of the FET's; one for each the CPU and the SOC VRM. HWInfo64 reads those sensors.

I can also confirm it's very easy for the CPU VRM temperature to exceed 100C when overclocking. At 105-110C the controller will initiate CPU throttling in the latest BIOS's but the BIOS my board came with...F2 i think it was...didn't enable throttling and it would get to around 115-120C before instability would crash the system.

The FET's used...OnSemi 4C10N / 4C06N low-side...are rated to something north of 125C so I think they're safe for it. But even so I'm sure operating above 95C isn't good so I put a fan on the VRM area that helps keep temp in the mid 80's during hand brake encoding. I feel a fan blowing on the VRM board area works well since these FETs' die heatsink just as effectively through their leads to the power and ground plane in the motherboard.

The SOC VRM will also power the IGPU's when Zen APU's come available. I expect that's when we'll learn if they're adequate for the job intended.

 

[PetrolHead]

I checked the specification of the board previously but there was no mention of the temperature sensors. I just checked the key features page and the board does indeed have five temperature sensors and two VRM sensors just like Buddywh says. A bit surprising for a board like this, but no doubt a good thing. Forget my previous post, I was wrong.

 

[Buddywh]

I checked the specification of the board previously but there was no mention of the temperature sensors. I just checked the key features page and the board does indeed have five temperature sensors and two VRM sensors just like Buddywh says. A bit surprising for a board like this, but no doubt a good thing. Forget my previous post, I was wrong.

These temp sensors (in particular the CPU VRM sensor) and the three SmartFan 5 fan headers that control multiple fans, PWM or volts, all the way down to 0 RPM are about the only good thing about this board.

The CPU core voltage reading is ludicrously bad, but at least in a way that if you trusted it will just leave you with a low (but safe) overclock and convinced you've the worst silicon ever.

 

[wingman99]

Hi...I've got the same motherboard and it has a thermal sensor located in the midst of the FET's; one for each the CPU and the SOC VRM. HWInfo64 reads those sensors.

I can also confirm it's very easy for the CPU VRM temperature to exceed 100C when overclocking. At 105-110C the controller will initiate CPU throttling in the latest BIOS's but the BIOS my board came with...F2 i think it was...didn't enable throttling and it would get to around 115-120C before instability would crash the system.

The FET's used...OnSemi 4C10N / 4C06N low-side...are rated to something north of 125C so I think they're safe for it. But even so I'm sure operating above 95C isn't good so I put a fan on the VRM area that helps keep temp in the mid 80's during hand brake encoding. I feel a fan blowing on the VRM board area works well since these FETs' die heatsink just as effectively through their leads to the power and ground plane in the motherboard.

The SOC VRM will also power the IGPU's when Zen APU's come available. I expect that's when we'll learn if they're adequate for the job intended.

4C10N operational Tjmax is 150c.

When I asked Infineon Technologies what is a safe temperature for 24/7 they said.

Normally the reliability tests are done at Tjmax, that means device can work at Tjmax for long period. But to use part at worst case is not reasonable, usually about 20% margin must be considered.

 

[Buddywh]

4C10N operational Tjmax is 150c.

When I asked Infineon Technologies what is a safe temperature for 24/7 they said.

I guess that's 'north of 125C' Still, I think my concern is how it affects the surrounding PWA material: I've seen boards where FET's have been driven really hot for long duration and it's guite apparent the board material is degrading.

 

[wingman99]

That must be hot, because video card MOSFETs run 80-90c+ for 24/7 projects.

 

[Buddywh]

That must be hot, because video card MOSFETs run 80-90c+ for 24/7 projects.

I can imagine that to be typical of those "barely used" crypto-mining cards that pop up on Craigs list every now and then.

Even so, I just play it safe using that fan to keep my VRM temps in the mid 80's during video encodes. I do prefer that to it bumping off the CPU throttle anyway.