VRM temps safe?

[VagabondWraith]

I have a Kraken G10 w/ heatsinks on the VRM of my 980 Ti. Since the GPU has no built in VRM monitoring (thanks nVidia) I decided to buy an IR Sensor temperature gun. I wasn't able to get right on the VRM except the outermost one, and I pointed the sensor to the back of the PCB directly above the power delivery area and the highest point was 45.5c. Is that a safe temp?

 

BTW, the temps are cool downstairs at 17.5c.

[don_svetlio]

45*C is too low to be an accurate reading. Most are usually 70*C-90*C under load

[Lord_Karango17]

That seems a bit low. The actual vrm temp will be way higher, but with nvidia cards there is no way of checking vrm temps in software, so the best you can do is point your thermometer at the heatsink itself, or just use your finger and see if the heatsink is too warm to touch.

 

One thing you have to watch out for is shiny surfaces, as they mess with ir sensors...

[Roawoao]

IR thermometers have a spot size that increases with distance and shiny or metallic parts can confuse the sensor. A thermal imaging camera can give you a better idea especially in detecting reflective surfaces that give false readings.

 

Sticking a piece of masking tape will give you a good emissivity to test off of just don'y leave the tape for long as it will get all gooy from the temperatures.

 

http://www.infrared-thermography.com/material-1.htm

 

Most IR thermometers are calibrated to around 0.95 while metals can give off wildly different readings due to surface treatments, paints, polish, ...

 

A thermocouple (watch out for shorts) is the best way to measure.

[VagabondWraith]

2 minutes ago, don_svetlio said:

45*C is too low to be an accurate reading. Most are usually 70*C-90*C under load

It was running for about 20 mins prior to taking the temps. Those IR guns are very accurate though. The highest I got after 45 mins is 48.5.

[VagabondWraith]

2 minutes ago, Roawoao said:

IR thermometers have a spot size that increases with distance and shiny or metallic parts can confuse the sensor. A thermal imaging camera can give you a better idea especially in detecting reflective surfaces that give false readings.

 

Sticking a piece of masking tape will give you a good emissivity to test off of just don'y leave the tape for long as it will get all gooy from the temperatures.

Oh okay. So how far off do you think it might be?

[Roawoao]

15 minutes ago, VagabondWraith said:

Oh okay. So how far off do you think it might be?

It depends strongly on the specific material and surface of the VRM and how small you can get the spot size typically IR thermometers are not designed to read tiny spots and it should say on it somewhere how this spot size increases with distance. The quickest way to try to compensate is to stick some masking tape on the VRM to measure it then if it reads much hotter then your getting a more expected reading. (Don't forget the masking tape actually reduces the actual surface temperature by insulating the VRM a bit but the 0.92 emissivity will be much better than an unknown metallic surface)

 

As others have said just touch the actual VRM bits with your finger if it burns it is hotter than ~45degC and if it does not burn then it is colder. It if really burns then it is likely >60degC. (Beware of ESD obviously) Also the coils (gray/metallic, ferrite) will likely be cooler than the actual switching ICs which are smaller black encapsulated chips very close to the coils (typically labelled with a Q## on the PCB silkscreen).

 

Most VRMs probably run quite hot and unlike a processor they can usually survive temperatures even at 105+degC max. (That would burn really bad if you touched something that hot). Don't forget with human temperature metallic parts (coils, leads, tabs) will feel hotter than plastic parts (chips) even at the same temperature as you only sense the heat flow not actual temperature. The 45degC comes from the fact that above this temperature human proteins will get cooked and this will cause pain.

 

Also looking at the PCB back is going to give very hard to interpret readings as the thermal resistances will make it difficult to determine the actual VRM temperature. This is on top of the emissivity problems a metallized PCB will have.

 

With a fan blowing directly onto the VRM area they are likely getting enough cooling even without a heatsink directly on them. The PCB itself is a sort of metal heatsink and the fans will cool it down accordingly.

 

[VagabondWraith]

@Roawoao

 

Thank you very much. That explained a lot!