• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

What exactly stops us from putting a thermal monitor tied to a relay, on the 12VHPWR connector?

S

StoleMyOwnCar

2[H]4U
Joined
Sep 30, 2013
Messages
3,481
Serious question, it's been something I've been considering for a while. We have these incredibly complex, current-based monitoring solutions for the connector, or ASUS trying to build their shunt resistor crap (which doesn't even work). But what about the simplest solution? Just attaching a bunch of thermal sensors to the connector, calibrating it for normal expected and peak temperatures while the GPU is running, and then wiring that to a tiny microcontroller that tells a relay (connected to the PSU power outlet) to just shut off instantly if things get too toasty? It may not completely solve the issue, I think some damage to the connector would/could still be sustained (and the PSU will provide some power even after shut off)... but on the other hand it should be dead simple. Thermal sensors are pretty mature at this point (afaik anyway), and not all that expensive. As are relays. With some proper calibration, I think you could also account for normal running temp, and it should be pretty obvious when the connector gets "way too toasty". I think frankly you could build this solution out with a cheap Arduino in probably <50$ with some duct tape to put the sensors on? I could be well out of my field here, but I just feel like this would at least help with barely any investment.
 
It's like the old Athlon days when the motherboards had a termistor in the middle of the socket. It was far off from an on package diode.
The only way to see if temps are too high on these connectors is using a thermal camera. It's possible to monitor in realtime and have ability to alert or shutdown but the sensitivity is such that you would know when things are going wrong long before an instant action (shutdown) is required. Comparatively speaking, it's like a low oil pressure light vs. pressure gauge.
 
A relay to handle the input power of your power supply starts to creep up on $200.

Could probably put a thermistor or sensor on the graphics card near the power connector. Some heat will transfer over and if it starts to notice too much could shut the card down and kick back to the drivers that user needs to re-seat connector. Or for most, "please call a professional".

Really should just move to better connectors and be done with it. Some nice XT90s (joking) on the board.
 
There probably is an easy solution, it's just a matter of a standard deciding if it's handled by the GPU or the PSU.
 
I mean you could use some cheap IOT thing if you want. My brain went right to UL listed relays.

not all UPS have communication ports, some do though.

Could even just get a heat sensor and a beeper. I mean you are gaming and sitting right there at peak usage anyways right?
 
Sure but do they fail gradually? Like is it something that a human would even have enough time to stop, or would it be better for it to be scripted?

I guess what you could do is something in the middle, have the Arduino/whatever hooked up to your PC via a USB port, constantly polling. If it detects a large heat increase in the connector, just instantly stop whatever game you're playing. I would think most games would be using some common backend, so maybe there's some graphics driver they could just kill...
 
MSI was apparently showing off a new 50 series model at Computex with resettable fuses on it and current monitoring.
 
Darunion said:
A relay to handle the input power of your power supply starts to creep up on $200.

Could probably put a thermistor or sensor on the graphics card near the power connector. Some heat will transfer over and if it starts to notice too much could shut the card down and kick back to the drivers that user needs to re-seat connector. Or for most, "please call a professional".

Really should just move to better connectors and be done with it. Some nice XT90s (joking) on the board.
Click to expand...

30A double pole contactors are inexpensive. 12V coils are a bit harder to find at cheaper prices. 24V is most common (HVAC).
 
A fascinating concept. My main worry is that the connector might already be broken by the time a thermal sensor notices too much heat. However, compared to other of the extremely complicated monitoring methods we've seen, it appears to be far more realistic as an inexpensive backup safety measure.
 
neman said:
A fascinating concept. My main worry is that the connector might already be broken by the time a thermal sensor notices too much heat. However, compared to other of the extremely complicated monitoring methods we've seen, it appears to be far more realistic as an inexpensive backup safety measure.
Click to expand...
We need high power cards to have a single two wire input that can handle 4AWG wire like this:
On the power supply it could be the same or use a 100A Anderson connector. No worries about melting!
12VDC-IN.png
 
cpufrost said:
30A double pole contactors are inexpensive. 12V coils are a bit harder to find at cheaper prices. 24V is most common (HVAC).
Click to expand...
Do you think that IOT relay I linked would be fine? It says it's rated for a decent amount of amperage and it's quite cheap. I think you could assemble it and and a cheap Arduino, and then some thermal sensors, for almost nothing, like maybe 50-70$ total and it would require little to no soldering knowledge or anything, just a breadboard. Alternatively, there's my software based hard cutoff idea.
 
StoleMyOwnCar said:
Do you think that IOT relay I linked would be fine? It says it's rated for a decent amount of amperage and it's quite cheap. I think you could assemble it and and a cheap Arduino, and then some thermal sensors, for almost nothing, like maybe 50-70$ total and it would require little to no soldering knowledge or anything, just a breadboard. Alternatively, there's my software based hard cutoff idea.
Click to expand...
Yes, the relay can handle it. The device itself has a 12A limit so if your load stays under 12A continuous, that is fine.
 
It comes down to... Could they, sure. If only the engineers were running the company.. Then you have to loop in C level, marketing, legal, accounting etc. Every one has their points to add, their particular needs for the product and their expectations for success. Once everyone has had their say, they look at the landed cost against what they can reasonably sell it for. After all of that, Accounting & Legal says it will cost you X in bad publicity, RMA's, repairs, replacements etc and we can make 3X by not doing all the things that others said was needed. Guess which way they will choose!?!?!?
 
StoleMyOwnCar said:
Serious question, it's been something I've been considering for a while. We have these incredibly complex, current-based monitoring solutions for the connector, or ASUS trying to build their shunt resistor crap (which doesn't even work). But what about the simplest solution? Just attaching a bunch of thermal sensors to the connector, calibrating it for normal expected and peak temperatures while the GPU is running, and then wiring that to a tiny microcontroller that tells a relay (connected to the PSU power outlet) to just shut off instantly if things get too toasty? It may not completely solve the issue, I think some damage to the connector would/could still be sustained (and the PSU will provide some power even after shut off)... but on the other hand it should be dead simple. Thermal sensors are pretty mature at this point (afaik anyway), and not all that expensive. As are relays. With some proper calibration, I think you could also account for normal running temp, and it should be pretty obvious when the connector gets "way too toasty". I think frankly you could build this solution out with a cheap Arduino in probably <50$ with some duct tape to put the sensors on? I could be well out of my field here, but I just feel like this would at least help with barely any investment.
Click to expand...
The shutoff mechanism is the easy part. It's the accurate monitoring of the pin temperature (the temperature we really care about), as opposed to the temperature of the plastic housing, that's hard. This is why all the attempts at doing this, like we've seen shown at Computex last week, or ThermalGrizzly has offered for a while, handle this by performing per-pin current measurement, which is what we really care about anyway.

Cooler Master is supposedly going to offer a power supply that has per-pin monitoring built in, with a shutoff mechanism for the 12V high failure connector built into it, triggered by any pin going over a preset current threshold. This is the proper way, since it allows the shutdown to be triggered long before the pins get hot enough to damage the connectors.
 
RazorWind said:
The shutoff mechanism is the easy part. It's the accurate monitoring of the pin temperature (the temperature we really care about), as opposed to the temperature of the plastic housing, that's hard. This is why all the attempts at doing this, like we've seen shown at Computex last week, or ThermalGrizzly has offered for a while, handle this by performing per-pin current measurement, which is what we really care about anyway.

Cooler Master is supposedly going to offer a power supply that has per-pin monitoring built in, with a shutoff mechanism for the 12V high failure connector built into it, triggered by any pin going over a preset current threshold. This is the proper way, since it allows the shutdown to be triggered long before the pins get hot enough to damage the connectors.
Click to expand...
Nah I think that's overcomplicating it. To actually melt plastic, the connector would have to get REALLY FUCKING HOT. I've burned out a component in circuits class, they get hot insanely fast and they get really hot. You could detect it probably within fractions of a second, even within a plastic housing. You probably just need enough thermal sensors to surround the thing. Then you calibrate them to the maximum running temp you see out in the wild, which should be pretty simple. Give +10-15% for ambient variance, and that should realistically be as hot as it ever gets normally. If that thing's melting, it'll easily skyrocket way past that limit, and it'll be fast. I'm thinking that this sort of shutoff might not be fast enough to completely stop it from melting, but I figure you might have some chance of preventing some damage. And it would be hella cheap, that already-assembled IOT strip is like 35 bucks. Then all you need is an Arduino, probably like 15-20. Then some thermal sensors to just plaster onto it, let's say 3 of them. Maybe 60-80$ total depending on what you get. Some wires and duct tape, with a breadboard that the thing probably comes with, and done. Could probably have AI code basically the entire microproject in one step these days, too.
 
You must log in or register to reply here.
Back
Top