Power Limit Mod had no effect!
Power Limit Mod had no effect!
I own a GTX 1070 ROG Strix.
It runs at around 40% power consumption, which is already quite high. While testing with software like Kombustor, it operates near the maximum power draw. It’s set to 120% but throttles around 103%.
Using some components from my 8700K, I decided to experiment with adjusting the power limit.
I connected the two shunt resistors (2mOhm and 5mOhm), restarted the system. No changes noticed.
Then I tried soldering the resistors together to short them out completely. I thought this would help if a fault occurred, allowing me to trace back from there. The same numbers appeared: 40% idle, throttling easily.
Finally, I removed both shunt resistors and replaced them with 1mOhm resistors.
THE RESULTS ARE IDENTICAL. This can’t be right unless all the power is coming from the PCI-E slot and not reaching the molex connector. (That’s the only resistor I didn’t swap)
Anyone have a suggestion? I’m really confused.
If you wish to exceed your stock power cap, a tailored BIOS would be ideal. I’ve crafted my own a couple of times using my 980tis, allowing me to adjust the power limit freely—usually around 140% since the 980ti. For a custom BIOS on the 1070, you should expect greater overclocking potential thanks to higher core voltage and increased power limits.
I notice you still have a Maxwell card. For Pascal cards, the BIOS is encrypted. The card has been out for two years and no one has managed to modify the BIOS like I have with every card since my 9800GTX+. That's why I've opted for hardware changes, which, as shown above, haven't worked.
Posted a gpuz image.
Never reached 90% with 1050ti using 6pin connector, though the maximum was 125% on power limit. Wish I could implement the shunt mod.
🙁
Sorry, I shared this on another platform too and fixed the idle usage issue. For those who noticed, I needed to use Nvidia Inspector to turn on "multi-display power-save." My monitors were pushing my clocks too high. Once the adjustment was made, idle usage dropped to around 2%. However, the main takeaway is that swapping a 5mOhm and 2mOhm shunt resistor for 1mOhm ones would have caused such a significant apparent power reduction that the card might have triggered a power fault. Instead, I observed essentially no change.
The power control mechanism measures current by observing the voltage drop across a shunt. It reads the input voltage (12v) and uses that to calculate the current via Ohm’s law (V=IR), then determines power consumption. With a 1/5th resistance value, the system expected only a 20% reduction in power usage. But since the readings stayed the same, it suggests something is amiss. I’m trying to identify what might be missing.