How would I properly set specific core Voltage on a Ryzen CPU?

[chopinanopolis]

(Ryzen 5 5600, ASRock B450M Pro4 R2.0) I'll post my full specs at the bottom of the post, I don't know if the mobo and CPU would be enough, but it can't hurt. Imma try to keep this short, basically my PC has been randomly rebooting since I built it in Feburary, mostly while gaming. I've basically tried everyhting from A-Z trying to fix it, but nothing has worked, including sending almost every single component in for RMA or flat out replacing it. When the PC reboots it shows me Machine Check Exception; Cache Hierarchy Errors with different APIC ID's in the Event Viewer, which point to the CPU. The one thing I havent tried is setting Core voltage, since that error apparently can be cause by cores not getting enough Voltage. The downside is I'm very tech illiterate when it comes to PC stuff, so I dont want to follow a random guide for a different Ryzen CPU. Thank you in advance for any help

GPU: AMD Radeon 6650XT
CPU: AMD Ryzen 5 5600
Mobo: ASRock B450M Pro4 R2.0
Ram: 16GB Kingston Fury Beast DDR4-3200 (Now Aegis G.Skill DDR4-3200)
SSD: 1TB Lexar NM620 M.2
PSU: 700 Watt be quiet! System Power 9
OS: Windows 10 Pro

 

[A Computer Guy]

Which components have you not replaced?

If your motherboard supports LLC you can try raising your LLC values in the UEFI/BIOS for vcore which should help with vdroop problems (based things I have read in the past). Another round about way to give your cores more voltage is to enable PBO and use curve optimizer with positive values instead of negative ones. The last method of course is to set a direct CPU voltage but I would only do that as a last resort.

Can you make your PC crash if run OCCT CPU or CPU+RAM stress tests? (https://www.ocbase.com/occt/personal)
If you can then you can try applying higher vcore LCC (if available) or vcore offset (it is likely a pick list of +/- fixed values) or use positive curve optimizer values incrementally to see if that helps reducing crashes.

Being a lower tier motherboard I think it's a bit of a gamble getting one that is problem free. I have the original ASRock B450M Pro4 and luckily not really any problems with a 3950x but I've read others with this motherboard have issues which is probably why there is a R2.0 in the first place. Assuming your CPU is not bad a better motherboard would probably work better for you all around. It sucks when it doesn't work out of the box correctly.

 

[chopinanopolis]

Which components have you not replaced?

If your motherboard supports LLC you can try raising your LLC values in the UEFI/BIOS for vcore which should help with vdroop problems (based things I have read in the past). Another round about way to give your cores more voltage is to enable PBO and use curve optimizer with positive values instead of negative ones. The last method of course is to set a direct CPU voltage but I would only do that as a last resort.

Can you make your PC crash if run OCCT CPU or CPU+RAM stress tests? (https://www.ocbase.com/occt/personal)
If you can then you can try applying higher vcore LCC (if available) or vcore offset (it is likely a pick list of +/- fixed values) or use positive curve optimizer values incrementally to see if that helps reducing crashes.

Being a lower tier motherboard I think it's a bit of a gamble getting one that is problem free. I have the original ASRock B450M Pro4 and luckily not really any problems with a 3950x but I've read others with this motherboard have issues which is probably why there is a R2.0 in the first place. Assuming your CPU is not bad a better motherboard would probably work better for you all around. It sucks when it doesn't work out of the box correctly.

The only component I haven't either sent in for just checks or outright replacement is my SSD. My CPU and Mobo I got replaced, my GPU was sent back to me cause they couldnt find any issues with it, same with my ram, but I got a new kit yesterday (the Aegis Gskill kit, thats why its in parenthesis in the spec list) and my PSU I replaced a few months ago cause I was told that it was the cause of the issue.
The PC doesnt crash at all while running stress tests, I've tried different programs for different components, and nothing has crashed. I'll have to check for LLC in a sec in bios.
It would be strange if its actually the motherboard, since I've gotten it replaced with a brand new one, and the exact same issue with the same error still happens, unless of course I'm just stupidly unlucky
Edit: Seems like the motherboard doesn't have LLC

 

[tabascosauz]

(Ryzen 5 5600, ASRock B450M Pro4 R2.0) I'll post my full specs at the bottom of the post, I don't know if the mobo and CPU would be enough, but it can't hurt. Imma try to keep this short, basically my PC has been randomly rebooting since I built it in Feburary, mostly while gaming. I've basically tried everyhting from A-Z trying to fix it, but nothing has worked, including sending almost every single component in for RMA or flat out replacing it. When the PC reboots it shows me Machine Check Exception; Cache Hierarchy Errors with different APIC ID's in the Event Viewer, which point to the CPU. The one thing I havent tried is setting Core voltage, since that error apparently can be cause by cores not getting enough Voltage. The downside is I'm very tech illiterate when it comes to PC stuff, so I dont want to follow a random guide for a different Ryzen CPU. Thank you in advance for any help

GPU: AMD Radeon 6650XT
CPU: AMD Ryzen 5 5600
Mobo: ASRock B450M Pro4 R2.0
Ram: 16GB Kingston Fury Beast DDR4-3200 (Now Aegis G.Skill DDR4-3200)
SSD: 1TB Lexar NM620 M.2
PSU: 700 Watt be quiet! System Power 9
OS: Windows 10 Pro

To boost Vcore in hopes of solving cache Hierarchy, the idea is not to set a static Vcore. Static OC hasn't really been particularly practical or meaningful for years now. The idea is to set positive Curve Optimizer values to essentially overvolt through changing the V-F curve. Default is 0, going higher values (up to max 30) is the idea.

If the vendor doesn't place a special CO menu in the main Tweaker page of the BIOS, then usually in Advanced > AMD OC > Precision Boost Overdrive > PBO to Advanced. For ASRock

If your CPU has been replaced, then it might be your board. When Cache Hierarchy is caused by CPU, a simple RMA will fix it, but when it's caused by a board, it's best to just go to a different board model rather than sticking with the same. Problematic WHEA 18 boards tend to just stay that way, even after BIOS updates.

Who did you send your parts in to? Unless it's an official RMA with AMD, chances are non-zero that they just sent your CPU back to you, unless you have verified through IHS markings that it is a new one.

 

[chopinanopolis]

To boost Vcore in hopes of solving cache Hierarchy, the idea is not to set a static Vcore. Static OC hasn't really been particularly practical or meaningful for years now. The idea is to set positive Curve Optimizer values to essentially overvolt through changing the V-F curve. Default is 0, going higher values (up to max 30) is the idea.

If the vendor doesn't place a special CO menu in the main Tweaker page of the BIOS, then usually in Advanced > AMD OC > Precision Boost Overdrive > PBO to Advanced. For ASRock

If your CPU has been replaced, then it might be your board. When Cache Hierarchy is caused by CPU, a simple RMA will fix it, but when it's caused by a board, it's best to just go to a different board model rather than sticking with the same. Problematic WHEA 18 boards tend to just stay that way, even after BIOS updates.

Who did you send your parts in to? Unless it's an official RMA with AMD, chances are non-zero that they just sent your CPU back to you, unless you have verified through IHS markings that it is a new one.

I'll try to find some PBO settings then, thank you. Should I try to even find specific settings, or is it more a matter of messing with values until it's stable? Or do you have any recommendations of where to start in terms of values?

My CPU and Mobo were both replaced. The CPU I sent directly to AMD and then Mobo I know is new, because my old one was damaged in shipping back from a RMA that I sent it in for. The Mobo I sent to the retailer I got it from (Mindfactory). Then only thing that hasn't been fully replaced or checked by the retailer is my SSD.

 

[tabascosauz]

I'll try to find some PBO settings then, thank you. Should I try to even find specific settings, or is it more a matter of messing with values until it's stable? Or do you have any recommendations of where to start in terms of values?

My CPU and Mobo were both replaced. The CPU I sent directly to AMD and then Mobo I know is new, because my old one was damaged in shipping back from a RMA that I sent it in for. The Mobo I sent to the retailer I got it from (Mindfactory). Then only thing that hasn't been fully replaced or checked by the retailer is my SSD.

Do the APIC values consistently point to specific cores? Or do they seem random? If they always point to specific cores then you can start adding a bit of voltage for those, maybe in increments of 5 to save time.

If the WHEAs seem to hit all different cores, then it might be wise to try a different model of board instead of replacing for the same one again. Preferably something newer B550 or X570, and something that isn't so bottom of the barrel as Pro4.

 

[chopinanopolis]

Do the APIC values consistently point to specific cores? Or do they seem random? If they always point to specific cores then you can start adding a bit of voltage for those, maybe in increments of 5 to save time.

If the WHEAs seem to hit all different cores, then it might be wise to try a different model of board instead of replacing for the same one again. Preferably something newer B550 or X570, and something that isn't so bottom of the barrel as Pro4.

They seem relatively random, some show up more often than others though, 0,9,8,1 for example.

I don't really want to be spending another 150 on a mobo right now honestly. The whole PC only cost around 650 when I built it, so it's not a stupidly expensive one. I think I'll just try the voltage stuff for now, and hope that it fixes it. If not, I guess I'll have no choice but to buy a new Mobo. I don't have access to any other Mobo right now either, so I can't test any other model sadly.

 

[tabascosauz]

They seem relatively random, some show up more often than others though, 0,9,8,1 for example.

I don't really want to be spending another 150 on a mobo right now honestly. The whole PC only cost around 650 when I built it, so it's not a stupidly expensive one. I think I'll just try the voltage stuff for now, and hope that it fixes it. If not, I guess I'll have no choice but to buy a new Mobo. I don't have access to any other Mobo right now either, so I can't test any other model sadly.

In the CO menu, leave it on All Cores then and it will apply your offset universally.

There should be very little risk with trying positive CO (although the vast majority use it the opposite way, as an undervolting method). The only consequence should be that your Vcore stays relatively the same while clocks will progressively decrease as with positive CO values the V-F curve calls for less clock at each voltage point. Ideally, positive CO should not be necessary at all, as you are literally reducing performance, but alas.

I didn't mean immediately commit to a new mobo at random, but buy it from somewhere with a liberal return policy so that you can ditch it with no financial loss if it doesn't work out for whatever reason.

 

[chopinanopolis]

In the CO menu, leave it on All Cores then and it will apply your offset universally.

There should be very little risk with trying positive CO (although the vast majority use it the opposite way, as an undervolting method). The only consequence should be that your Vcore stays relatively the same while clocks will progressively decrease as with positive CO values the V-F curve calls for less clock at each voltage point. Ideally, positive CO should not be necessary at all, as you are literally reducing performance, but alas.

I didn't mean immediately commit to a new mobo at random, but buy it from somewhere with a liberal return policy so that you can ditch it with no financial loss if it doesn't work out for whatever reason.

Alright, thank you. I turned it up to 5 and it still rebooted, so now I'm running it at 10 and gonna see if that works. The crashes aren't consistent, so sometimes it runs perfectly fine for a day or two, and sometimes it crashes immediately, so it's gonna take a minute to see if it actually makes a difference sadly.

I'll see what I can find. Would I need to worry about the ram qvl list of other Mobos? Someone told me those are basically irrelevant now as long as it's DDr4 on a DDr4 board it should be fine?

 

[tabascosauz]

Alright, thank you. I turned it up to 5 and it still rebooted, so now I'm running it at 10 and gonna see if that works. The crashes aren't consistent, so sometimes it runs perfectly fine for a day or two, and sometimes it crashes immediately, so it's gonna take a minute to see if it actually makes a difference sadly.

I'll see what I can find. Would I need to worry about the ram qvl list of other Mobos? Someone told me those are basically irrelevant now as long as it's DDr4 on a DDr4 board it should be fine?

The QVL is not really that useful anymore, but the best thing you can do for yourself is just to get a more recent X570 board (a few late release boards from 2020+), X570S board (2021+ I think?), or a B550 board (which all are from 2020+). All B550 boards saw a general improvement in memory topology and that carried over to those new X570/X570S as well.

There are a handful of B450s that are essentially B550 PCB designs with a B450 PCH (e.g. B450M TUF Pro S), but they are the exception not the rule. Boards like that really aren't much more affordable than their B550 parent design anyway.

 

[chopinanopolis]

The QVL is not really that useful anymore, but the best thing you can do for yourself is just to get a more recent X570 board (a few late release boards from 2020+), X570S board (2021+ I think?), or a B550 board (which all are from 2020+). All B550 boards saw a general improvement in memory topology and that carried over to those new X570/X570S as well.

There are a handful of B450s that are essentially B550 PCB designs with a B450 PCH (e.g. B450M TUF Pro S), but they are the exception not the rule. Boards like that really aren't much more affordable than their B550 parent design anyway.

Nice alright. I'll probably go with a 550 board then, some are under 100 bucks, so that's definitely a plus, as well as being mATX.

The machine-default permission settings do not grant Local Activation permission for the COM Server application with CLSID
{C2F03A33-21F5-47FA-B4BB-156362A2F239}
and APPID
{316CDED5-E4AE-4B15-9113-7055D84DCC97} to the user DESKTOP-5H1C96B\My Name SID (S-1-5-21-935824414-3933982711-1161374692-1001) from address LocalHost (Using LRPC) running in the application container Unavailable SID (Unavailable). This security permission can be modified using the Component Services administrative tool.

Stuff like this can't cause random reboots, right? Cause I've been getting these since I built the PC aswell. I got them down to only one error but they've mulitplied again, now theres like up to 6 sometimes