Topic: Undervolting RX570 doesn't stick (Read 81 times)
That's most likely just due to voltage drop. It's normal for the effective voltage to be lower than the number you set.
Okay, this seemed to have made me go below 875mV.
I ended up flashing the bios to set the GPU core voltage to 812mV. Always stays that level at high clockspeeds now.
All seems to be well, although it seems my mem is running at a higher voltage (835mV) than the core (812mV).
I ended up flashing the bios to set the GPU core voltage to 812mV. Always stays that level at high clockspeeds now.
All seems to be well, although it seems my mem is running at a higher voltage (835mV) than the core (812mV).
Set memory voltage equal or lower than the core voltage in high P-states.
I have several RX570s in my rig I'm atttempting to undervolt.
Using AMD Wattman, I lower the voltages to 835mV.
However!
1. HWInfo tells me the GPU core voltage never drops below 875mV. Strange, but I can live with that.
What's worse is the following.
2. The GPU completely ignores the undervolting and stays at 1.000V, EXCEPT when I manually change the undervolt to 836mV and apply in Wattman.
The annoying thing is that as soon as the GPU leaves the "busy" P7 state (at which point the clock drops to 300MHz), the undervolt setting gets reset.
Next time I start Claymore's dual miner, GPU voltage is at 1.000V again.
Can anyone explain this behaviour?
The undervolting results in 20-30W less power consumption per card, so it's important I get it right.
Using AMD Wattman, I lower the voltages to 835mV.
However!
1. HWInfo tells me the GPU core voltage never drops below 875mV. Strange, but I can live with that.
What's worse is the following.
2. The GPU completely ignores the undervolting and stays at 1.000V, EXCEPT when I manually change the undervolt to 836mV and apply in Wattman.
The annoying thing is that as soon as the GPU leaves the "busy" P7 state (at which point the clock drops to 300MHz), the undervolt setting gets reset.
Next time I start Claymore's dual miner, GPU voltage is at 1.000V again.
Can anyone explain this behaviour?
The undervolting results in 20-30W less power consumption per card, so it's important I get it right.
Jump to: