[SOLVED] R5 5600x pushing up to 1.5v idle because PBO tells it to

[erik62905]

Hey guys,

So I enabled PBO around a month ago and set the max boost clock override to 200 MHz. It's definitely working and boosting very high, hwinfo says it's boosting as high as 4.85 GHz. Expectedly, this is only happening in short bursts when I'm running something like chrome. The problem is PBO is pushing very inconsistent voltages (it stays at a given voltage and clock speed for only a second or two). These points are what it seems to like the most:

• 0.98v @ 3.6 GHz
• 1.43v-1.46v @ anywhere from 3.6-4.75 GHz depending on what it feels like doing
• 1.48v-1.5v for a short spike @ 4.75-4.85 GHz

My temps are around 42C at idle and I think that's why PBO decides to act like this. In games this is not a problem at all and it stays at around 1.2v while clocking high. I haven't had any crashes for as long as I've had PBO enabled, so that's good. Also, hwinfo tells me the number given is the requested amount of voltage by the core and not necessarily the actual voltage supplied by the voltage regulator. So is the real voltage being applied to the CPU much lower than this and I'm worried over nothing? Or is the number given at least somewhat close? Should I be worried, and is there any way I can lower my idle voltages without disabling PBO? Thanks

 

[Lutfij]

What is the make and model of your motherboard and the BIOS version for said motherboard? I'd advise on turning PBO off. Keeping voltages below 1.35v is a good idea to have the processor last a longer time frame.

 

[erik62905]

What is the make and model of your motherboard and the BIOS version for said motherboard? I'd advise on turning PBO off. Keeping voltages below 1.35v is a good idea to have the processor last a longer time frame.

Asus Strix B550-F GAMING running BIOS version 1805.

Product page: https://rog.asus.com/us/motherboards/rog-strix/rog-strix-b550-f-gaming-model/

 

[hotaru.hino]

According to the author of HWiNFO, you should be using "CPU Core Voltage (SVI2 TFN) " to get the actual voltage going into the processor.

Also according to someone from AMD's technical marketing department, up to 1.5V is fine.

 

[erik62905]

According to the author of HWiNFO, you should be using "CPU Core Voltage (SVI2 TFN) " to get the actual voltage going into the processor.

Also according to someone from AMD's technical marketing department, up to 1.5V is fine.

That sensor is showing me a very consistent 1.43v with a max of 1.48v. It's not going down.

 

[hotaru.hino]

That sensor is showing me a very consistent 1.43v with a max of 1.48v. It's not going down.

It may be reading the highest voltage that's going into the processor.

 

[erik62905]

It may be reading the highest voltage that's going into the processor.

That's probably right, but it's not showing anything lower than 1.43v even when VID shows 0.96v. I'll disable PBO later and see what hwinfo says.

 

[PrinceTexasLoaf]

Ryzen cpus at idle run at higher clock speeds and so they use higher voltage so there's no instability, then when under load they lower their voltage since they're running at lower clock speeds. This should be fine.

 

[drea.drechsler]

.... The problem is PBO is pushing very inconsistent voltages (it stays at a given voltage and clock speed for only a second or two).
....

Ryzen 2nd and 3rd gen CPU's boost aggressively to finish off transient processes quickly. It's by design and they way they work. AMD has said it's normal and expected that when it boosts clocks it will also raise voltage to as high as 1.5V to keep it stable. If the transient load doesn't finish off quickly it lowers clocks and voltage as necessary to keep the processor safe. It's doing this all very rapidly, up to 100 times a second, voltage and clocks jumping up and down for various cores.

So it sounds to me like your CPU is working precisely as it's designed to work. If you want to try undervolting do it with negative offset only, a little bit at a time. Some motherboards seem to allow more than others. To know it's helping test performance with Cinebench, both multithreaded and single threaded, and as well a stress test for stability. While the processor is designed to raise voltage as high as 1.5V in light threaded workloads the payoff is with heavier loads as the processor can stay a bit cooler and hold higher clock speeds...with greater performance resulting.

 

[erik62905]

Thank you all for your help! I think I'll try a little bit of undervolting to help with the voltages at idle, but after what was said about how it's meant to go up to 1.5v, I'm assuming it's running normally and higher readings are because hwinfo can't update sensors fast enough.

 

[drea.drechsler]

Thank you all for your help! I think I'll try a little bit of undervolting to help with the voltages at idle, but after what was said about how it's meant to go up to 1.5v, I'm assuming it's running normally and higher readings are because hwinfo can't update sensors fast enough.

You can help with HWInfo...configure it as shown in this thread for more frequent polling. But the CPU makes decisions up to 100 times a second...that's 1mS intervals which is way faster than anything before.

The boost algorithm will pull back voltage as the cores get hotter and at the same time lower boost clock speeds. You can watch it (sort of) in HWInfo by setting up graphs for core clocks, core temperature and SVI2 TFN core voltage. Watch the clocks and voltage as you click around windows, or maybe run a Defender quick scan. Then start a stress test...something like Prime95...and watch what happens as it gets hotter. You should see it working in the graphs; as the core heats up voltage will drop really low, maybe around 1.25V, in P95 but the clocks also are going down. It will still be jumping around though, but much less.

If your undervolting is working the clocks won't drop quite so far, maybe 25 or 50Mhz less. But you may not be able to see it in the clocks, only in benchmarks.