AM4 4 phase boards temps issue

[vitalic66]

Hey guys,

I recently updated to ryzen. It's a ryzen 1700 with a gigabyte ax370 gaming k3. Not the highest end board neither the lowest end.

Like a lot of AM4 boards, it have 4 phases for the CPU. I saw in a spreadsheet of all the AM4 board models that basically all the non top tier boards have the exact same mosfets model in 4+2 phase configuration.

I was watercooling and overclocking it. Like 4ghz at 1.45v. CPU temps were really good but during gaming i saw the VRM spike to around 70-80 on smaller games and 90+ on big CPU games.

Then i tested prime, and, oh boy, 120c climbing untill crash !

I dropped a fan on the backplate of the GPU pointing at the VRM cooler to see, and the temps dropped dramatically to around 60 in games and 70-80 on load (maybe more on prime i can't recall exactly but lower than previously).

Since there is no mosfet WB for my card, i decided to uninstall the loop and put back the wraith cooler. At this point, at stock but still 1.38v splike when XFR kick in, mosfet temps never exceded 65c. The fan of the cooler giving a good air flow on the heat sink.


I decided to try out an high end air cooler, the phanteks dual tower one. And right now, after 30 mins of Prime small FFT, at 3.9ghz 1.38v, the CPU is stable at 82c, package at 120w, core power at 220w, CPU+SOC at 243w, SOC mosfets at 56C AND... DRUM ROLLS... VRM mosfets at 120c !! Stable and not crashing like with the watercooling but still !



I wonder how many people have bought ryzen 7 with AIO coolers to overclock and are reading the false HWmonitor temps ?

If somebody with a ryzen build and a non high end motherboard can do some tests and share, it will be much apreciated. Also, read temps with HWinfo64, since these temps are the real correct ones. Not the 60c non moving temps of HWmonitor. I also like to know how's the temps on 6+phase boards.

Thanks for the read.


EDIT :

After half an hour of battlefield 1 64 player map, CPU is stable at 64 and Mosfet at 75. The game uses all 8 real cores at around 50-60% and barely uses the hyperthreads. The fans are barely spinning since i like it to be quiet. I think it's acceptable but in the future it might become a problem.

 

[trents]

I'm not sure about all the board temps but as I recall HWMonitor gave the same readings on the package temp of the R5 1600 processor as did Ryzen Master on my ASRock AB350M Pro.

 

[wingman99]

I run those temperatures is gaming all the time so don't worry the rig will last a decade. I was wondering if you could show a screenshot of BF1 8 real cores at around 50-60% and barely uses the hyperthreads?

 

[vitalic66]

I'm not sure about all the board temps but as I recall HWMonitor gave the same readings on the package temp of the R5 1600 processor as did Ryzen Master on my ASRock AB350M Pro.

HWMonitor give the right CPU temps. But the Mosfet temps are frozen to the point when you launched the app. Ryzen master gives only the CPU temps. HWinfo64 gives all the temps and they are true.

I run those temperatures is gaming all the time so don't worry the rig will last a decade. I was wondering if you could show a screenshot of BF1 8 real cores at around 50-60% and barely uses the hyperthreads?

As you can see one out of two cores are used in the actual and minimal values. The max values are a spike who does not represent 99% of the time. In ryzen mode, windows 10 prioritize real cores over threads. Since half of the cores are used more than the other half, i bet these are the real cores.

 

[trents]

How do you know they are true?

 

[vitalic66]

Because i highly doubt a CPU would prioritize a thread before a core.

 

[trents]

I was talking about HWInfo64 mosfet temp readings.

 

[vitalic66]

Because they fluctuate with the load of the CPU, unlike HWmonitor. Also, i pointed a fan at the mosfets and the temp was decreasing. Finally i touched the heat sink when the CPU was on load and it was burning. More on par with mosfets at high temps than just 60c like HWmonitor says.

 

[Woomack]

I'm not sure if VRM reading is correct. When I had GB AB350 Gaming 3 then under full load software was showing ~105°C. However VRM heatsink wasn't really hot ( could be bad contact ). My friend bought 2 of these boards for work ( like 24/7 rendering etc ) and they're working fine for 2 months+ fully stable. I'm just saying that at 110°C+ these boards would be unstable or shutdown ( 110-120 usually is max ). Maybe reading is correct but barely any software is showing all correct voltages or temps. Hwinfo64 was working the best for me on ASUS, Gigabyte and Biostar boards but still had some issues with some voltages or temps ( depends on the board ).

 

[wingman99]

I never heard of VRM's being unstable, it's not like folks can overclock them, they should be foolproof by design for 3 years.

 

[Johan45]

I never heard of VRM's being unstable, it's not like folks can overclock them, they should be foolproof by design for 3 years.

As the temps goes up so does the resistance which lowers out put requiring a higher voltage. It's a vicious circle. They get to a ppint where they're tapped out and can't deliver a clean signal either they die or the board shuts them down via throttling. In a sense they are unstable at that point

 

[trents]

Personally, I just use the finger touch method of reading mosfet temps.

 

[Tír na nÓg]

Personally, I just use the finger touch method of reading mosfet temps.

Youn made my evening!

 

[wingman99]

As the temps goes up so does the resistance which lowers out put requiring a higher voltage. It's a vicious circle. They get to a ppint where they're tapped out and can't deliver a clean signal either they die or the board shuts them down via throttling. In a sense they are unstable at that point

When the heat increase then the resistance increases electronics 101, so to compensate the duty cycle increases on the MOSFET to maintain the voltage within specifications until throttling or shutdown with Foolproof specification to prevent damage. What does a clean signal have to do with MOSFET resistance do to heat and what is a vicious circle?

A MOSFET is a on and off switch.

 

[vitalic66]

Reading is absolutely correct since the heat sink is burning hot. Like touching a frying pan. With 1.3v the mosfets are a lot cooler. Like 60c. This mean the dissipation power of the heat sink is not good enough for higher voltages. This is sad, given all the gigabyte (even aorus have the same heat sink) and i think other brands uses the exact same heat sink on every motherboard models untill ultra high end.

 

[trents]

Reading is absolutely correct since the heat sink is burning hot. Like touching a frying pan. With 1.3v the mosfets are a lot cooler. Like 60c. This mean the dissipation power of the heat sink is not good enough for higher voltages. This is sad, given all the gigabyte (even aorus have the same heat sink) and i think other brands uses the exact same heat sink on every motherboard models untill ultra high end.

Okay, I can accept that as valid. Anything from about 90c and up is pretty dangerous to prolonged touch.

 

[Johan45]

When the heat increase then the resistance increases electronics 101, so to compensate the duty cycle increases on the MOSFET to maintain the voltage within specifications until throttling or shutdown with Foolproof specification to prevent damage. What does a clean signal have to do with MOSFET resistance do to heat and what is a vicious circle?

A MOSFET is a on and off switch.

Right it's an on and off switch in an electrical circuit which delivers a signal. As the fets are pushed out of their optimum range that signal gets noisy because of the inefficency of the switch at higher temps. C'mon I thought the vicious circle was obvious. more heat more resistance need higher input to overcome which in turn causes more heat.etc.. etc..

 

[Dolk]

Johan is correct, but its not just the FETs, you have to remember the inductor has resistance as well.

http://www.overclockers.com/forums/...rature-or-why-you-should-keep-that-board-cold

(Need to add that final section on inductance though)

@wingman99, the instability happens because you can no longer turn on or off the FETs in their specified duty cycle range. Changing that parameter only works for a very limited time and is the wrong way to fix the issue. Once the FET's cannot cleanly cycle power through their phase, the inductor can stay "hot" and cause lower inductance to flow across (thus keeping the circuit hot and reducing current input into the CPU). At this point you are a few steps away from thermal run off which the VRMs cannot handle cleanly.

@Vit (OP)
Finger test for temp is not an ideal mean of saying its hotter than 40C, 50C, or anything.... Remember 1st 2nd/3rd degree burns can occur at 50-60C depending on conditions. So if you touched something at 120C, you'd probably have a large welt or similar. Please use a temp probe (not the one on the board) to verify your worries.

AM4 boards with 4+2 power phases have the potential to handle the OC of these chips, but with a Ryzen 7 it may have difficulties. As your testing has confirmed, depending on Air cooling quality, your OC potential will vary. By this feedback, it seems like the phase was designed decently well. I'd suggest strapping a fan on top of the VRM fins and than finding an OC that works best, than take it down by 100mhz. Your board is working hard to achieve what others would call a simple OC, so it looks like you need more cooling to handle it.

 

[wingman99]

Right it's an on and off switch in an electrical circuit which delivers a signal. As the fets are pushed out of their optimum range that signal gets noisy because of the inefficency of the switch at higher temps. C'mon I thought the vicious circle was obvious. more heat more resistance need higher input to overcome which in turn causes more heat.etc.. etc..

The MOSFET does not deliver a signal, it delivers only DC voltage in a pulse form to the inductor that changes it to continuous DC with the capacitors. The operating range of the MOSFET Temperature and Frequency is set by the motherboard companies and is FOOLPROOF. When you see 1.425v with a Digital voltmeter that is what the VRM is producing for DC voltage even when running the MOSFET at the highest factory calibrated temperature before throttling and shutdown.

Dolk is correct if the factory parameters were off we would see Mosfets buning up and massive complaints in the forums and review sites demanding the manufacture to take care of the problem. We don't design motherboard VRMs here, so folks don't have to worry about running a 4-2 phase with 1800X because manufactures set the calibration so that you will make it through the warranty period.

Most people don't even join forums when building a PC manufactures have to make the motherboards foolproof. I will give one very small manufacture example

EVGA GeForce GTX 1080/1070 PWM Operating Temperature Update

Recently, it was reported from several sources, that the EVGA GeForce GTX 1080 FTW PWM and memory temperature is running warmer than expected during Furmark (an extreme stress utility). https://www.evga.com/thermalmod/

So if there is even a slight problem like EVGA had folks would find the problem in motherboards by the masses.

The best selling motherboard at Newegg is 4+2 phase GIGABYTE GA-AB350-GAMING 3.

 

[trents]

It's okay if you already have scar tissue from previous finger testing.