Can AMD 1700 OC and drop voltage @ idle?

lightsout

[H]ard|Gawd
Joined
Mar 15, 2014
Messages
1,211
I haven't got this working with my 2600x.

I plan to build a server around a Ryzen 1700 (already have the board and ram).

I would like to OC but don't want the thing running full speed/voltage 24/7.

It seems like whenever I set a static clock on my 2600x everything stays maxed out.
 
Ryzen power profile is needed for windows 7 or 10 if I remember correctly in order to allow it to idle in the lower state possible when not pushed.
 
Ryzen power profile is needed for windows 7 or 10 if I remember correctly in order to allow it to idle in the lower state possible when not pushed.
I do use Ryzen profile. I believe I read that you need to set the minimum clock state to a low percentage like 5%, but thought even when I did that it only dropped the clocks not the voltage.
 
I do use Ryzen profile. I believe I read that you need to set the minimum clock state to a low percentage like 5%, but thought even when I did that it only dropped the clocks not the voltage.
probably have to set things to auto in the bios for it to drop the voltage as needed
 
  • Like
Reactions: GHRTW
like this
Their is a " cheat " you can do........, not sure which dude; on which forum! i read this ........:

After finding max OC and stable; on my MSI X370 Gaming Pro Carbon, i have to save this into bios slot, then apply, then F10 to reboot.

Boot into desktop to confirm OC settings and volts or frequency not moving ( as per AMD spec! ); THEN......................................., reboot back into bios and set frequency to auto and CnQ also back to auto ( crucial point dudes! ) F10 and reboot back into desktop.

With Frequency + CnQ set to auto they should fall; unless applied fixed voltage?. Reboot back into bios and reset frequency and CnQ to auto, F10 and boot back into desktop.

Works.........; sometimes :) for me. Worth trying if you know you are not going to change anything in bios; frequency change, changes CnQ to disabled or enabled etc.

Good Luck:)

Edit : Just tested .....; so in HWiNFO64 can confirm both CpuVid + Frequency fall and rise, yet Cpu core voltage SVI2 TFN remains @ bios idle load max, which for me with 3.8Ghz UV = @ ~1.144v with CPB disabled! ( 1.138v desktop ) for 1.100v Full Load Floor ( P95 ~ 20mins Short-form test :) ).
 
Last edited:
  • Like
Reactions: GHRTW
like this
I did a mild oc on a 1600/msi b350 and left all the speed settings in bios to auto and it clocks and volts up and down just fine.
 
Edit : Just tested .....; so in HWiNFO64 can confirm both CpuVid + Frequency fall and rise, yet Cpu core voltage SVI2 TFN remains @ bios idle load max, which for me with 3.8Ghz UV = @ ~1.144v with CPB disabled! ( 1.138v desktop ) for 1.100v Full Load Floor ( P95 ~ 20mins Short-form test :) ).
Just to weigh in here based on my experience...
I think that the SVI2 TFN voltage (aka VDDCR_CPU as far as I know) is the max voltage value the CPU can pull from the MOSFET, and the CPU-VID is the voltage the CPU is requesting to be run at currently.

Reason I think this because in my HWiNFO Sensor Status --mind you, while overclocked so CnQ is not enabled-- my SVI2 TFN reads 1.337V while my CPU MOSFET VOut is reading 1.338V (same for the SoC's SVI2 TFN and the CPU-NB MOSFET VOut, their readings both match). So that's what my CPU is being fed actually.
Whereas my BIOS is set with a VCore of 1.35V, which is what my Core VID reads since that's what it's expecting to be delivered.

That being said... The curious thing about all these readings, at least on my MSI X370 Titanium, is that under a load the value of the SVI2 TFN (both for CPU and SoC) do not fluctuate at ALL. They remain steady at those reported voltages. However the CPU VID drops, depending on the load, down to 1.3V for the AIDA64 FPU Stress and 1.275V for the AIDA64 Cache Stress. BUT, the VCore in the Motherboard category of HWiNFO increases from 1.35V to 1.392V for both FPU or Cache stress tests. What I can indeed say is that the Motherboard VCore reading changes based on what the CPU LLC option in the BIOS is configured to. If I changed it to a lower setting I know for sure that one will change, but I don't recall if the Core VID or the MOSFET do as well.


As far as lightsout is concerned... You'll want to dabble with the PState options on your board, provided it has them (my Titanium never was given them, sadly), as that just overrides the default multipliers for the Power States (ie, Boost and normal clocks).
In order to keep it fed the voltage you want instead of what AMD has it programmed with, you'll want to use the Voltage Offset mode for adjusting voltage. Using "Offsets" will change the voltage at the MOSFET stage, and so the CPU doesn't know that the voltage is any different; thus, not entering "0C" mode and not disabling CnQ or Boost.

Of course, since it's still employing the Boost states, you'd basically want to set PState 2, 1 and 0 all to the same value, because 2 is full-boost which will only apply to one core, 1 is partial boost so all the cores, 0 is default clocks. If you only set 0 and left 1 and 2 alone, it'd actually cause it to reduce the clock speed under load when it switched to a "Boost" state :p
 
Just to weigh in here based on my experience...

As far as lightsout is concerned... You'll want to dabble with the PState options on your board, provided it has them (my Titanium never was given them, sadly), as that just overrides the default multipliers for the Power States (ie, Boost and normal clocks).
In order to keep it fed the voltage you want instead of what AMD has it programmed with, you'll want to use the Voltage Offset mode for adjusting voltage. Using "Offsets" will change the voltage at the MOSFET stage, and so the CPU doesn't know that the voltage is any different; thus, not entering "0C" mode and not disabling CnQ or Boost.

Of course, since it's still employing the Boost states, you'd basically want to set PState 2, 1 and 0 all to the same value, because 2 is full-boost which will only apply to one core, 1 is partial boost so all the cores, 0 is default clocks. If you only set 0 and left 1 and 2 alone, it'd actually cause it to reduce the clock speed under load when it switched to a "Boost" state :p

Thats what i was missing...............; wanted to use P-States like this :
P-State 2 = Max Efficient OC/UV @ 3.8Ghz ~ 1.1000v F.L.F
P-State 1 = Max OC Temp
P-State 0 = Max OC Frequency...............!??

Now i understand that P2= Max Single/Duel core? - P1=All core p0= Default ...GRrrrrrrrrr; are these functions fixed in that way? I want AMD Overdrive to take my P2/1/0 values and either Gaming boost or Desktop efficiency ! Sounds reasonable to me :)
 
I looked at Pstates with my 2600x but it was more than I wanted to fool with.

I may just spend the money on a 1700x and be happy with the core clocks (the [H] in me has a hard time saying that). We'll see.
 
I looked at Pstates with my 2600x but it was more than I wanted to fool with.

I may just spend the money on a 1700x and be happy with the core clocks (the [H] in me has a hard time saying that). We'll see.


agree; but having fun with the non-x 2600. Almost certain after my adventures that p-states are doing nothing but introducing more instability!!! Do they only work for the x models?
 
agree; but having fun with the non-x 2600. Almost certain after my adventures that p-states are doing nothing but introducing more instability!!! Do they only work for the x models?
No I don't think so, but the non-x models have weaker clocks. I started with a non-x 2600, but XFR was pretty weak so opted for a 2600x. I am content just to let it do its auto magic.
 
I can do a 4.2 all core OC on my 2600 by setting just P state 0 (highest clock) to 4.2 and using a voltage offset + to add the voltage needed to keep it stable under load. Everything else CPU wise is stock. If your board does not have voltage offset then find your highest OC on stock volts and set P state 0 to that freqency and see how it goes.

Use Windows balanced power plan with Ryzen 2xxx cpus. Set lowest to 5% and max to 100%.

CPU down clocks and drops voltage at idle just like stock settings.


Asus CH7 Hero wifi. currently . Also worked fine on a 1600 with Asrock AB350 itx at 3.8ghz P state 0 -- no voltage changes made.

Caveat for me is my Ram can do 3200 c14 4x8 for 32gb on stock CPU but Ram OC falls apart under cpu OC to 4.2 ... lol Have to drop ram to 2933 to OC the CPU that far. Currently I have the ram set to 3200 c14 and CPU stock since it hits 3900 all core with no adjustments. The extra 300mhz are cute but really don't notice in day to day use.
 
I can do a 4.2 all core OC on my 2600 by setting just P state 0 (highest clock) to 4.2 and using a voltage offset + to add the voltage needed to keep it stable under load. Everything else CPU wise is stock...........".

On my MSI X370 GPC , we now have both Offset + Override on Cpu + NB/Soc ...:) Yay!!!; since bios 1.K0 i think ( currently on 1.L0 )?
What we do not have is individual setting of p=state like you have. Have messed around with P-state OC'ing but not feeling anything on my R5 2600 really; besides more instability, which leads me to think MSI are rubbish in this department; or...., just for X-models :-(

Can confirm same CnQ behaviour, i mostly use High performance mode ( with minimum sat @ 35% ), since noticed small amounts of lag being introduced ( edge-cases dude; not in all instances! ) when using Balanced.

Ram !!!!!! Right, i have CMK8GX4M2B3200CL16....... Corsair vengeance LPX ( Samsung E-die ). These, would not pass 3200 XMP with my R5 1400 @ 1.45v+. Needed atleast one nudge on both NB/SocV + CpuV to stabilise. Currently @ 3333 auto everything ( 1.41v bios ); will do 3400 and seems stable enough; just not quite finished OC'ing the snot out of my 2600 yet :). Will eventually get aroung to OC'ing DDR; been as low as 40Ohm and as high as 3566 ( would not finish startup loading before logon! ).

Edit 1 : Currenty have Ryzen Master applying my best OC - 4.2Ghz @ Full_Load_Floor of 1.375v :). So getting 1.425v Desktop_Idle-Load_Max ( ouch :-(, wish the Vdroop was less!!! ), so frequency falls as well as Cpu vid ( but only under sustained idle load; any dektop activity pushes vid back to D.I.L.M again! ).

Still, better that nothing :)

Edit : spelling + clarity.
 
Last edited:
On my MSI X370 GPC , we now have both Offset + Override on Cpu + NB/Soc ...:) Yay!!!; since bios 1.K0 i think ( currently on 1.L0 )?
What we do not have is individual setting of p=state like you have. Have messed around with P-state OC'ing but not feeling anything on my R5 2600 really; besides more instability, which leads me to think MSI are rubbish in this department; or...., just for X-models :-(

Can confirm same CnQ behaviour, i mostly use High performance mode ( with minimum sat @ 35% ), since noticed small amounts of lag being introduced ( edge-cases dude; not in all instances! ) when using Balanced.

Ram !!!!!! Right, i have CMK8GX4M2B3200CL16....... Corsair vengeance LPX ( Samsung E-die ). These, would not pass 3200 XMP with my R5 1400 @ 1.45v+. Needed atleast one nudge on both NB/SocV + CpuV to stabilise. Currently @ 3333 auto everything ( 1.41v bios ); will do 3400 and seems stable enough; just not quite finished OC'ing the snot out of my 2600 yet :). Will eventually get aroung to OC'ing DDR; been as low as 40Ohm and as high as 3566 ( would not finish startup loading before logon! ).

Edit 1 : Currenty have Ryzen Master applying my best OC - 4.2Ghz @ Full_Load_Floor of 1.375v :). So getting 1.425v Desktop_Idle-Load_Max ( ouch :-(, wish the Vdroop was less!!! ), so frequency falls as well as Cpu vid ( but only under sustained idle load; any dektop activity pushes vid back to D.I.L.M again! ).

Still, better that nothing :)

Edit : spelling + clarity.
I am just using PB2 or XFR which ever it is at level 2. And I watched the voltage dance around at 1.45v yesterday. I have it on auto. I need to do some messing with that to get it down.
 
Can you apply a negative value in offset voltage on that board and still enable Precision boost?

That might do it.
 
Back
Top