Hello /sci/,
Mechanical Engineering undergrad here.
In cryptocurrency mining using graphics cards or ASIC's, or even just in server rooms, waste heat seems to be a large untapped resource that could be used to reduce the power consumption of the components.
Using a small thermoelectric device, the temperature differential could be used to drive a voltage which could be used to produce a small amount of energy which the component could use, to even say reduce the power consumption by 1-5%, which for a 150W graphics card could turn into a substantial saving. Especially when the manufacture of the small thermoelectric device is fairly cheap (£5ish), whereas a new graphics card (say rx 580) is £200-300.
Do you guys think this idea could form a decent dissertation paper?
Cheers
>>9167352
>Do you guys think this idea could form a decent dissertation paper?
I think it's a neat idea. I...is that good enough?
>>9167357
Thanks senpai.
I was more curious to see if there was something obvious I was missing, or if this was already done by Nvidia/AMD/IBM or something
>>9167361
I think that you'd get a better idea if you just ran it by your supervisor.
>>9167352
That's a good idea. also look into stirling engines. they produce a tiny ammount of power, but they are used in sumarines to power some shit idk.
AAAAAAAAAAAAAAAAAAAAA ITS MENTIONED EVERYWHERE I GO
Ok, but how are you going to cool the cards?
>>9167456
Remember, buy high, sell low
>>9167352
>Do you guys think this idea could form a decent dissertation paper?
Sure.
You might want to optimise the heat path, with a very cold cold side drain, and gain heat from RAM, then disks and in the last stages CPU/GPU before again sending heat to the thermoelectric device.
>>9167352
The heatsinks need to be at an elevated temperature to dissipate heat. The thermoelectric generators need a temperature differential between the sides to produce power. The only way to achieve both would be to increase the temperature of the processor.
>Do you guys think this idea could form a decent dissertation paper?
You're a fucking idiot.
>>9167618
>>9167456
AAAAAA
>>9167662
that sounds like a good method of extracting as much heat as possible
>>9167671
>computer is off
>heat sink and cpu/gpu = ambient room temperature
>computer is turned on
>cpu/gpu heat up
>heat is transferred from the cpu/gpu into the heatsink
instead, why not add a TEG to extract a current from the heat differential
>>9167691
the TEG presents itself as an additional thermal resistance between the processor and heatsink
how fucking hard is it to understand this?
>>9167699
yeah, and what's the problem with that?
Component efficiency has almost no bearing on temperature between like -20'C and +90'C, unless the thermal resistance is enough that it would cause the chip to overheat (maybe a better heatsink or liquid cooling?)