[SOLVED] Im starting to pull my hair with this cpu (Ryzen 5 3600)

[JerT]

After numerous posts ive been getting mixed messages about my processors voltages on avrg idle (1,375 - 1,400 and sometimes up to 1,500) some say its ok some say not so ok and some say its dangerous and im just confused and no idea how to fix this so now ive recorded footage to show what i mean precisely https://streamable.com/l3fr94

EDIT: Thanks for drea for his/her continued patience with a Ryzen noob like myself, if anyone is having similar issues please go through our convo below about this.


Things i did that helped:

1. Updated BIOS to latest.
2. Disabled PBO
3. Enabled AMD Cool N Quiet
4. Update Chipset drivers and use the AMD Balanced power plan and make sure min cpu state is 90% and max 100%
5. Set a rather agressive chasis and cpu fan curves.

Remember that when making changes to the bios i recommend doing them one at the time to make sure they apply so for example: Disable PBO -> exit save (system reboots) -> re-enter bios enable AMD cool n quiet -> Exit save (system reboots)

 

[drea.drechsler]

View: https://www.reddit.com/r/overclocking/comments/f18cbc/reviewing_voltage_recommendations_for_zen_2/

 

[JerT]

So in short with voltages that ive got my pc is doomed if i understood correctly to put it blunt how do i <Mod Edit> this situation of mine?

 

[mdd1963]

what mainboard is being used, and, is it with latest BIOS?

Any BIOS- level overclocks selected, many of which tend to simultaneously bump up core voltage perhaps too much?

Were BIOS defaults selected, with CPU voltage, etc., all at defaults (auto, default, etc)?

What method is being used to give you the core voltage present? And what temps are being seen at idle and under an all-core intensive load, such as CPU-z/bench/stress CPU?

 

[dorsai]

Get rid of HWinfo...it supposedly triggers the core to take the V reading and then measures the millisecond peak voltage from the core loading up giving you a false idea of what the cpu is doing...under a true sustained all core load the V will typically be closer to 1.35v. Just use Ryzen Master for voltage readings.

If you're overclocking and worried about the life span of the cpu keep it under or @ 1.35v...if you don't care about 7 years of cpu life you should still try and stay under 1.4v on a 24/7 overclock.

Using PBO will also cause voltage spikes as the motherboard will push higher limits on voltage and current into the chip...if you have good cooling and temps stay below 75c under load you should be fine...the colder the better.

If you're on the stock cooler just turn off PBO and leave everything else on auto...then theres nothing to worry about and you'll likely have just as good performance 95% of the time anyway.

 

[JerT]

what mainboard is being used, and, is it with latest BIOS?

Any BIOS- level overclocks selected, many of which tend to simultaneously bump up core voltage perhaps too much?

Were BIOS defaults selected, with CPU voltage, etc., all at defaults (auto, default, etc)?

What method is being used to give you the core voltage present? And what temps are being seen at idle and under an all-core intensive load, such as CPU-z/bench/stress CPU?

MSI MPG X570 GAMING PLUS AM4 everything is auto pretty much (as they were when i got my pc and first booted into bios (prebuilt pc) and dude ive got no idea what to do in the bios etc only thing i know the basics on is ryzen master. No idea about bios version nor the way to check it. You can see my idle and load voltages and temps on the video.

 

[JerT]

Get rid of HWinfo...it supposedly triggers the core to take the V reading and then measures the millisecond peak voltage from the core loading up giving you a false idea of what the cpu is doing. Just use Ryzen Master for voltage readings.

If you're overclocking and worried about the life span of the cpu keep it under @ or under 1.35v...if you don't care about 7 years of cpu life then just stay under 1.4v

Using PBO will also cause voltage spikes as the motherboard will push higher limits on voltage and current into the chip...if you have good cooling and temps stay below 75c under load you should be fine...the colder the better.

If you're on the stock cooler just turn off PBO and leave everything else on auto...then theres nothing to worry about and you'll likely have just as good performance 95% of the time anyway.

Mate ive got CPUZ,HWMONITOR,HWINFO64,MSI AFTERBURNER,RYZEN MASTER all of them show same voltage readings but not temps really ryzen master and hwinfo64 is the most accurate on temps i think that is.

 

[JerT]

Get rid of HWinfo...it supposedly triggers the core to take the V reading and then measures the millisecond peak voltage from the core loading up giving you a false idea of what the cpu is doing...under a true sustained all core load the V will typically be closer to 1.35v. Just use Ryzen Master for voltage readings.

If you're overclocking and worried about the life span of the cpu keep it under @ or under 1.35v...if you don't care about 7 years of cpu life you should still try and stay under 1.4v on a 24/7 overclock.

Using PBO will also cause voltage spikes as the motherboard will push higher limits on voltage and current into the chip...if you have good cooling and temps stay below 75c under load you should be fine...the colder the better.

If you're on the stock cooler just turn off PBO and leave everything else on auto...then theres nothing to worry about and you'll likely have just as good performance 95% of the time anyway.

Oh yeah and reddit is full of these too thats why im confused people are reporting mass degrading over months View: https://www.reddit.com/r/overclocking/comments/eu3fbl/r5_3600_degradation_testing/

 

[drea.drechsler]

Oh yeah and reddit is full of these too thats why im confused people are reporting mass degrading over months...

Until you understand what the cause of degradation is you won't understand what's happening. Most people don't, that's why Reddit is full of mis-information. The article I posted (also on Reddit, but at least a reputable source) tries to explain it.

Voltage isn't the cause, it's current and heat that leads to degradation. High voltage at low current density, and therefore low temperature, is really quite safe. So high voltage of itself is safe but only if it doesn't exceed dielectric stress levels where it can break down oxides. That's two entirely different phenomenon: degradation from electromigration and dielectric or oxide breakdown. Nobody knows what the dielectric stress level is for the TSMC 7nm process used in Zen2 CPU's, but it's almost positively north of 1.5V since AMD has staked their corporation on the millions of CPU's they're selling that regularly boost that high in normal operation for light loads by design.

If your system is <Mod Edit>, then AMD is worsely <Mod Edit>. And I think AMD's a hell of a lot smarter than you, me and 99% of Reddit. So no, they ain't <Mod Edit>.

Your computer is a dynamical system, so you have to look at it in several modes. Look at it not only in light loads, where you see voltage regularly bouncing around in the 1.375 even up to 1.5V range as it boosts and idles cores but AVERAGE core temperatures are still low (60-65C maybe or less) and so NOT DEGRADING at a high rate. But also look at it in a heavy all-core processing load, where AVERAGE core temperature is getting pretty hot...70, 75, 80, 85C. Here, (SVI2 TFN) voltage should be dropping to below 1.3V, maybe 1.275, along with clocks dropping lower in order to keep temperature from rising up to the Tjmax of 95C. If it isn't, NOW you're <Mod Edit> if you don't fix it.

 

[dorsai]

Oh yeah and reddit is full of these too thats why im confused people are reporting mass degrading over months View: https://www.reddit.com/r/overclocking/comments/eu3fbl/r5_3600_degradation_testing/

I've been running my 3600 using a 240 AIO cooler...no issues at all so far...but i decided early on after some testing that there's no point to a 24/7 overclock on these chips. Since AMD is already pushing performance to the max on stock settings these really are set it and forget it chips...which is a big change from the intel chips most of us ran for the last 10 or more years.

 

[JerT]

I've been running my 3600 using a 240 AIO cooler...no issues at all so far...but i decided early on after some testing that there's no point to a 24/7 overclock on these chips. Since AMD is already pushing performance to the max on stock settings these really are set it and forget it chips...which is a big change from the intel chips most of us ran for the last 10 or more years.

So what are your idle voltages and temps and then on load? similar to mine? because almost everyone ive talked to have values miles below mine...

https://imgur.com/a/wglGWbL

View: https://imgur.com/a/wglGWbL
thats my avrg idling voltage(s)

 

[drea.drechsler]

...
thats my avrg idling voltage(s)

There's nothing unsafe about that. My 3700x idles in the 1.4-1.42 range if something light is running like a Firefox.

What are your average idling temperature? not the instantaneous temp, the one that spikes and then ramps down almost constantly. AVERAGE temp; RyzenMaster reports it. HWInfo reports it as Tdie(average).

EDIT add: are you manual, all-core, overclocking? That Reddit FUD you posted was a manual overclocker, his observation doesn't apply if you're keeping things in AUTO.

 

[JerT]

There's nothing unsafe about that. My 3700x idles in the 1.4-1.42 range if something light is running like a Firefox.

What are your average idling temperature? not the instantaneous temp, the one that spikes and then ramps down almost constantly. AVERAGE temp; RyzenMaster reports it. HWInfo reports it as Tdie(average).

EDIT add: are you manual, all-core, overclocking? That Reddit FUD you posted was a manual overclocker, his observation doesn't apply if you're keeping things in AUTO.

Only things i touched in the bios were the chasis fan curves and disabled pbo with no effect and enabled amd cool and quiet with minimal effect. avrg idle temprature on desktop is 42c my cooler is Be Quiet! Pure Rock CT. What im looking to do is to get the voltages to safe levels because they are jumping crazy high on minimal load /idle up to 1,450v+, also like ive said before in my posts and other people ive discussed this with they dont have such high idle voltages their dip as down as 0,200v-1,000v. If i may ask how long have you had your cpu? but to CLARIFY the volts are extremely high with NOTHING else than hwinfo64 or ryzen master on WITHOUT moving mouse etc. I do have a 2 year warranty on this if the cpu gives out due to this will my other parts take a hit aswell?

 

[drea.drechsler]

...
thats my avrg idling voltage

Only things i touched in the bios were the chasis fan curves and disabled pbo with no effect and enabled amd cool and quiet with minimal effect. avrg idle temprature on desktop is 42c my cooler is Be Quiet! Pure Rock CT. What im looking to do is to get the voltages to safe levels because they are jumping crazy high on minimal load /idle up to 1,450v+

Where you're at is perfectly safe.

What you really need to do is stop reading the FUD that abounds on Reddit. Or at least keep in mind that OVERCLOCKERS are the ones experiencing the problems with degradation as they're disabling AUTO for the CPU voltage and multiplier. That means as the CPU comes under load and heats up the algorithm is locked out, so it can't lower voltage AND clock frequency to keep temperature in safe levels as establish by the processor's FIT tables.

In other words: as LONG as you leave those in AUTO, the processor is observing it's FIT requirements and keeps itself always in a safe operating envelope of temperature, clock speed and voltage.

ADD: I've had my 3700X since last July...about two weeks after release. But I really didn't my settings ironed out until a month or so ago when I got PBO working really sweet.

 

[JerT]

Where you're at is perfectly safe.

What you really need to do is stop reading the FUD that abounds on Reddit. Or at least keep in mind that OVERCLOCKERS are the ones experiencing the problems with degradation as they're disabling AUTO for the CPU voltage and multiplier. That means as the CPU comes under load and heats up the algorithm is locked out, so it can't lower voltage AND clock frequency to keep temperature in safe levels as establish by the processor's FIT tables.

In other words: as LONG as you leave those in AUTO, the processor is observing it's FIT requirements and keeps itself always in a safe operating area of temperature, clock speed and voltage.

Even with full load on cinebench and other testing software the voltages barely dip 1,300v. Max temps ive reached is 88c on cinebench and ive tried rust once and cpu was at about 75-80c, honestly at this point i probably should take a deep dive into the bios and look for anything that looks out of place but i dont really know about that stuff so yeah.

 

[drea.drechsler]

Even with full load on cinebench and other testing software the voltages barely dip 1,300v. Max temps ive reached is 88c on cinebench and ive tried rust once and cpu was at about 75-80c, honestly at this point i probably should take a deep dive into the bios and look for anything that looks out of place but i dont really know about that stuff so yeah.

Really, really...that's perfectly normal. 1.3V in running cinebench? beautiful! it's dropping as it should.

Try Prime95. I'd like to see if it drops into the 1.275-1.29V range under that massive processing load. But it might throttle clocks pretty severely if you're on the stock Wraithe cooler, possibly all the way to the rated base clock speed for the CPU.

If you went to a really good cooler, say a 240mm AIO, the processor won't drop it's clock so low and the voltage could stay a bit higher too. But temperature would be lower and that's what allows it.

I can't stress this enough: it really sounds like your processor is operating perfectly normal. You have to get the stuff you've read on Reddit out of your mind; they are overclockers trying to get a fixed clock speed and voltage. They're finding that when you don't let the processor drop to lower clock, and voltage, to lower temperature it degrades. So they're lowering voltage even lower and accepting the loss of stability in heavy processing or also lowering clocks till it's not even as fast at light-to-moderate threaded work as the stock processor.

It's entirely different if you're letting the algorithm manage processor health as it's supposed to.

Right now, I'm running a Folding@Home CPU work unit that has my 3700X at 4.25Ghz, 1.3 volts, about 70C. I've been doing this almost constantly since the pandemic started, but did it intermittently all winter before this.

 

[JerT]

Really, really...that's perfectly normal. 1.3V in running cinebench? beautiful! it's dropping as it should.

Try Prime95. I'd like to see if it drops into the 1.275-1.29V range under that massive processing load. But it might throttle clocks pretty severely if you're on the stock Wraithe cooler, possibly all the way to the rated base clock speed for the CPU.

If you went to a really good cooler, say a 240mm AIO, the processor won't drop it's clock so low and the voltage could stay a bit higher too. But temperature would be lower and that's what allows it.

I can't stress this enough: it really sounds like your processor is operating perfectly normal. You have to get the stuff you've read on Reddit out of your mind; they are overclockers trying to get a fixed clock speed and voltage. They're finding that when you don't let the processor drop to lower clock, and voltage, to lower temperature it degrades. So they're lowering voltage even lower and accepting the loss of stability in heavy processing or also lowering clocks till it's not even as fast at light-to-moderate threaded work as the stock processor.

It's entirely different if you're letting the algorithm manage processor health as it's supposed to.

Right now, I'm running a Folding@Home CPU work unit that has my 3700X at 4.25Ghz, 1.3 volts, about 70C. I've been doing this almost constantly since the pandemic started, but did it intermittently all winter before this.

I have Be Quiet! Pure Rock CT. Is this the official prime95 site? mersenne.org. just making sure dont want to end up with some kind of ware in the midst of this also what power plan do you recommend using? all of the 4 i see have min cpu state of 85% and max 100% . on light load my clocks vary from 3,6ghz to 4,1ghz.

 

[Flayed]

The idle core voltage is 1.406v on my r5 3600 ATM. I was initially concerned until I dug around and did some research on the topic. The core voltage drops as it should as the load on the CPU increases.

 

[drea.drechsler]

I have Be Quiet! Pure Rock CT. Is this the official prime95 site? mersenne.org. just making sure dont want to end up with some kind of ware in the midst of this also what power plan do you recommend using? all of the 4 i see have min cpu state of 85% and max 100% . on light load my clocks vary from 3,6ghz to 4,1ghz.

For simplicity use the Ryzen Balanced power plan and leave the Min CPU state at 99%, which it comes with. Max at 100%. You only get that if you install the AMD chipset drivers; download them from AMD's web site. I use 1Usmus's custom Balanced power plan. If you're interested I'll send you the link, but first you have to get straight on what's normal for Zen2.

Mersenne.org is the home site for Prime95. If you're asking then it's obvious you've never used this: you'll want to run the Torture Test option. It's easily the most tortuous workload possible for an AVX-capable processor to run and most manually overclocked systems have to be careful they not run some of the routines. If you've got everything set to AUTO, voltage and multiplier, then your CPU can protect itself; it won't let temperature get too hot without dropping clocks and volts to stay safe.

One of the 'big deals' of Zen2 is it's superior power efficiency. It evaluates it's work load and idles cores all the way into C6 deep sleep states as often as it can. It's making the decision up to 100 times a second, way way faster than Windows does with it's processor power management. That's why you leave the power plan at 99% minimum, it basicaly tells Windows to leave the processor to manage it's power useage.

But all that activity is also why you see the processor clock going up and down so much. Even at idle it's doing work all the time in short bursts that need high clocks, then it drops back to go to sleep. But in C6, deep sleep, it can't report it's asleep or it wakes up so instead it just reports the last state before dropping there and that's usually about 3.6Ghz.

 

[dorsai]

Where you're at is perfectly safe.

What you really need to do is stop reading the FUD that abounds on Reddit. Or at least keep in mind that OVERCLOCKERS are the ones experiencing the problems with degradation as they're disabling AUTO for the CPU voltage and multiplier. That means as the CPU comes under load and heats up the algorithm is locked out, so it can't lower voltage AND clock frequency to keep temperature in safe levels as establish by the processor's FIT tables.

In other words: as LONG as you leave those in AUTO, the processor is observing it's FIT requirements and keeps itself always in a safe operating envelope of temperature, clock speed and voltage.

ADD: I've had my 3700X since last July...about two weeks after release. But I really didn't my settings ironed out until a month or so ago when I got PBO working really sweet.

This. I stated the same thing earlier...set to auto and forget it...and got down voted lol.

 

[JerT]

This. I stated the same thing earlier...set to auto and forget it...and got down voted lol.

Dont know why someone would downvote solid advice? anyways i upvoted thanks!

 

[JerT]

For simplicity use the Ryzen Balanced power plan and leave the Min CPU state at 99%, which it comes with. Max at 100%. You only get that if you install the AMD chipset drivers; download them from AMD's web site. I use 1Usmus's custom Balanced power plan. If you're interested I'll send you the link, but first you have to get straight on what's normal for Zen2.

Mersenne.org is the home site for Prime95. If you're asking then it's obvious you've never used this: you'll want to run the Torture Test option. It's easily the most tortuous workload possible for an AVX-capable processor to run and most manually overclocked systems have to be careful they not run some of the routines. If you've got everything set to AUTO, voltage and multiplier, then your CPU can protect itself; it won't let temperature get too hot without dropping clocks and volts to stay safe.

One of the 'big deals' of Zen2 is it's superior power efficiency. It evaluates it's work load and idles cores all the way into C6 deep sleep states as often as it can. It's making the decision up to 100 times a second, way way faster than Windows does with it's processor power management. That's why you leave the power plan at 99% minimum, it basicaly tells Windows to leave the processor to manage it's power useage.

But all that activity is also why you see the processor clock going up and down so much. Even at idle it's doing work all the time in short bursts that need high clocks, then it drops back to go to sleep. But in C6, deep sleep, it can't report it's asleep or it wakes up so instead it just reports the last state before dropping there and that's usually about 3.6Ghz.

How is it looking? https://streamable.com/qc4pe3 should i just run this for a few hours?

 

[drea.drechsler]

How is it looking? https://streamable.com/qc4pe3 should i just run this for a few hours?

That was looking OK, but I can't say it's 'right'. I'm more interested in (SVI2 TFN) core voltage reported in HWInfo.

Also, use a custom FFT size, set it in the dialogue used to launch the torture test. Set the FFT size to 1280kb both max and min, run FFT's in-place. That's a good balance between insane, tiny FFT's that execute fully in the L1 cache and larger FFT's that still seem to make the CPU draw highest power and get good and warm.

I'd only run this a few minutes, 10 should be enough. Really, just enough to heat up the processor and heatsink really good and make the algorithm pull clocks about as far as they need to go.

I'd only do a long multi-hour test with a stock hardware setup that you want to burn-in to shake out infant failures. You might want to do this early in ownership, when you're still in the window to take it back to the retailer for exchange if something does fail. Otherwise, you're still in warranty and running it hard with no good purpose just wears things out early and wastes electricity.

 

[JerT]

That was looking OK, but I can't say it's 'right'. I'm more interested in (SVI2 TFN) core voltage reported in HWInfo.

Also, use a custom FFT size, set it in the dialogue used to launch the torture test. Set the FFT size to 1280KB both max and min, run FFT's in-place.

Here is an updated one with hwinfo64 update interval is 500ms idk if its enough tho let me know: https://streamable.com/7in90p btw couldnt do 10 mins because streamable doesnt like big video sizes but ill test it out without recording.

 

[drea.drechsler]

Here is an updated one with hwinfo64 update interval is 500ms idk if its enough tho let me know: https://streamable.com/7in90p btw couldnt do 10 mins because streamable doesnt like big video sizes but ill test it out without recording.

500ms is good enough.

Your temp's looking pretty good even at 4 min's on p95, only 76C. Your FIT seems to be happy leaving voltage at 1.325 for 3.9G, which is well above base clock speed. That's pretty good performance for a stock cooler...are you on the stock Wraithe?

Now, if you really want to experiment you could lower voltage a notch or two. BUT ONLY DO IT WITH OFFSET. Dial in a negative offset one or two or two notches, then check this again and compare temps and end voltage with another run like this one. You use offsets because it still leaves the algorithm in control of lowering voltage; setting a manual voltage will over-ride everything and try to keep it at a fixed value.

If you're happy with results then make a Cinebench 20 run both multi-thread and single-thread and compare. You have to check performance as it's easy to lower voltage too much and the processor does something called 'clock compression'. Even though it's reporting similar clock speed it's actually performing worse in benchmark like Cinebench.