Ok by editing the Bios on RX570 i understood something about undervolting that i should have expected (damn studies were long ago) … I will use figures from my tests on RX570 to illustrate.
For those who dont know about RX570, you can edit the bios with tools like Polaris and can hardly code Voltage for CPU and RAM, but also TDP of the card. And it is 100% efficient, if you say TDP is 100W, card will never go above 100W (with timesamples of 1 second). The current metters have been checked with a killowat plug and HW64.
If you do not modify the max TDP in the bios, wich is at 120W by default, but undervolt the GPU and the memory for example at 900mV. The result is a massive increase in performance (in case of high consuming algorithms like dual mining). And the reason is logical, i should have expected
. If i understand correctly, the GPU only limit TDP but doesnt limit current. So if you undervolt your GPU, it can result in a proprotional increase in current delivered to your GPU (the old P= UI law). By undervolting my GPU to 900mV i allowed my GPU to receive 25% more current if it required it. The global current at 220V remaining the same of an out of a box card at its max TDP.
A one day test with an algo at full load (dual mining) confirms that this setting is error free Under HW64.
From a heat point of view it is as safe as the out of the box card at its max TDP.
But my question is from a component point of view : i have quite no doubt that undervolting should be safe for the electronic components as long as it remains reasonnable. Though i am wondering if the possible result of lets say a 25% incresase in current compare to the out of the box specification is safe in the long run (6 months? a year? 2 years).
If anyone has done such an undervolting without limiting the TDP so that the max current in the card is not limited at the same value as unmoded card, please feel free to give information.
What would be great is feedback answering these questions :
- how much you undervolted GPU (and/or RAM) in % = A
- how much you limited TDP compared to nominal in % =B
- optional : max current increase in GPU you allowed (= B/A if you simplify by not taking RAM and other components into account)
- for how long and how many cards (by type of cards ie GTX 1060, RX480, etc) you did it
- feedback on faulty cards and your opinion if its related to increase in max current
And if you re an expert on electronic components your general feedback is also welcome
