Question The lower the CPU volts the good?

Jan 4, 2022
5
0
10
0
Hi,

Should I lower the CPU wattage as low as possible? or should keep in mid?

Thanks!
 
Jan 4, 2022
5
0
10
0
Well, it depends. If you are a user that put a lot of load on the processor then leave it in auto.

if you manually under-volt the CPU the performance could be affected.

i didnt really understand what you said, but as far as i understood you are going to undervolt it to keep the temps down?

(like @Lafong said, give all the info the next time you are posting a message :) )
I have overclocked to 4.8 Ghz i5-8600K, Z390-F, 1.25V. Should I up the volts a bit or put it in auto? I just overclocked the cpu and yes I do put tonnes of load on CPU.
 

natcha12

Honorable
Sep 1, 2015
368
10
10,865
32
I have overclocked to 4.8 Ghz i5-8600K, Z390-F, 1.25V. Should I up the volts a bit or put it in auto? I just overclocked the cpu and yes I do put tonnes of load on CPU.
It is nigh on impossible to harm a CPU by UNDERVOLTING it, but it is easy to kill it or shorten life by going over 1.4/1.5 for long periods (AMD can go slightly higher than intel, but eventually they die too).

The issue with undervolting is system stability under load. If your PC handles 95% usage scenarios perfectly at for example 1.2 volts, but you use it 97% load, it will crash.

If you had to up the voltage to get to 4.8ghz, the chances are you would not be able to reduce it without instability. If it automatically OC to 4.8, you have a shot.

If the later, go slow and do not change voltage more than 0.01/0.02 at a time
 
Jan 4, 2022
5
0
10
0
a
It is nigh on impossible to harm a CPU by UNDERVOLTING it, but it is easy to kill it or shorten life by going over 1.4/1.5 for long periods (AMD can go slightly higher than intel, but eventually they die too).

The issue with undervolting is system stability under load. If your PC handles 95% usage scenarios perfectly at for example 1.2 volts, but you use it 97% load, it will crash.

If you had to up the voltage to get to 4.8ghz, the chances are you would not be able to reduce it without instability. If it automatically OC to 4.8, you have a shot.

If the later, go slow and do not change voltage more than 0.01/0.02 at a time
at 1.25V my CPU at 100% 4.79 Ghz works perfectly. Do I need to change anything now or everything is fine?
 

natcha12

Honorable
Sep 1, 2015
368
10
10,865
32
a

at 1.25V my CPU at 100% 4.79 Ghz works perfectly. Do I need to change anything now or everything is fine?
It doesn't matter if you put your pc at 100% for a minute, or even an hour. It can crash after 2 hours, 5 hours even. That is why pre-built pc's are stress tested for a full 24 hours before shipping out.

You can either;
- Reduce the voltage, 0.01/0.02 at a time like i said, until it instantly crashes after turning on, this tells you that you are at the limit. Then you would take voltage up until it doesn't crash. Trial and error

- Leave voltage or use it to get more overclock manually. I'm purposefully leaving this vague as you need to do a bit of reading and research before you start cooking your CPU alive.

I am assuming you used some automatic OC to get to 4.8, judging by your initial question. That is why i suggest only doing option 1.

Reducing voltage will lower temps, that's never a bad thing unless you use your pc as a radiator
 
Something to throw out regarding voltage and whether or not it damages the components or whatever.
  • The primary cause of damage is excessive current. Excessive current leads to heat build up. This can exacerbated by a property of semiconductors called thermal runaway. Once the part reaches a certain temperature, its resistance decreases, leading to even more current being drawn, which leads to more heat.
  • Too low of a voltage does not damage the part. The only issues too low of a voltage causes is instability, which would lead to the system crashing. A system crashing is a software problem, the hardware doesn't care.
  • Too high of a voltage is a problem, but any damage is usually immediate. This applies to the power delivery or signal inputs. Static discharge events are another story. In any case, when the voltage is too high, it causes insulators to break down and become conductors. In an integrated circuit, this causes shorts, which leads to excess current. Another thing of note is that the only voltage that's too high is one specified by the manufacturer.
  • Voltage does directly affect heat output. The formula is P = C * f * V^2, where P is power dissipation (i.e., heat output), C is capacitance (something you can't change), f is the operating frequency, and V is the voltage. Since voltage is squared, it has a higher impact than changing frequency, so obviously lowering voltages can dramatically decrease heat output.
 
Something to throw out regarding voltage and whether or not it damages the components or whatever.
  • The primary cause of damage is excessive current...
I just wanted to comment on that a little bit, hopefully for the better. It's arguable which is the "primary" cause. I'd say heat is, but current is a very close secondary. It's significant to keep in mind that if it's kept cool a high current is much less harmful. At the extreme end of things extreme overclockers are running with phenomenally high currents during benchmark runs but the core temperatures are kept in check by using LN2 so it's safe. They don't like losing their competition 'golden silicon' CPU's anymore than we do.

As you point out, current in addition to voltage is definitely what generates the heat at a given frequency, and for a given cooling solution voltage is the only knob we have to turn down the temperature. But Ryzen can seem to operate at high voltages (1.5V) at 5Ghz; it does so only briefly as the boost algorithm pulls clocks back quickly but by then the process is usually finished. At heavy loads, when all cores are hot, it's constantly dithering both clocks and voltage, which is quite low now, to keep the cores safe.

The voltage limit is the dielectric withstanding limit that's inherent with the process. With Ryzen, it's got to be north of 1.5V since Robert Halleck has frequently commented that it will hit 1.5V when it boosts. He's even commented that it will hit as high as 1.55V if it's 'chill', referring to a cold startup. It is unfortunate his comments on Reddit and in the AMD Community forums are the only thing we have to go on since they don't release their data sheets. I don't think he'd lie about that though, I think he's giving it plenty of margin in fact.

I remember watching a video by Buildzoid (an overclocker). He took an Intel CPU and ran up the voltage to show what happens when it's overvolted. It's getting hot at the same time but it didn't buzz out until north of 1.6V. So even 1.52V, indicated in their data sheets, doubtless has plenty of margin too.
 
Last edited:

natcha12

Honorable
Sep 1, 2015
368
10
10,865
32
Tell me this is a joke.

Auto voltage is by far the main source of instability when it comes to overclocking. An 8600K at 4.8GHz running with just 1.25V would literally be a golden sample, most require 1.3 or more to hit that clock.
There's alot on this forum about people getting them to 4.8 on 1.2/1.25, not as rare as you think.

Auto voltage is THE most stable basis for an overclock. Because it automatically goes to a voltage that is high enough for every single OOB cpu to run at, it means almost every single one can overclock on it without needing to change it higher, no?

Whether OP decides to go up or down is up to them, but if he's truly already at stable 4.8 with 1.25, yes he can probably lower the voltage.
 
I just wanted to comment on that a little bit, hopefully for the better. It's arguable which is the "primary" cause. I'd say heat is, but current is a very close secondary. It's significant to keep in mind that if it's kept cool a high current is much less harmful. At the extreme end of things extreme overclockers are running with phenomenally high currents during benchmark runs but the core temperatures are kept in check by using LN2 so it's safe. They don't like losing their competition 'golden silicon' CPU's anymore than we do.

As you point out, current in addition to voltage is definitely what generates the heat at a given frequency, and for a given cooling solution voltage is the only knob we have to turn down the temperature. But Ryzen can seem to operate at high voltages (1.5V) at 5Ghz; it does so only briefly as the boost algorithm pulls clocks back quickly but by then the process is usually finished. At heavy loads, when all cores are hot, it's constantly dithering both clocks and voltage, which is quite low now, to keep the cores safe.

The voltage limit is the dielectric withstanding limit that's inherent with the process. With Ryzen, it's got to be north of 1.5V since Robert Halleck has frequently commented that it will hit 1.5V when it boosts. He's even commented that it will hit as high as 1.55V if it's 'chill', referring to a cold startup. It is unfortunate his comments on Reddit and in the AMD Community forums are the only thing we have to go on since they don't release their data sheets. I don't think he'd lie about that though, I think he's giving it plenty of margin in fact.

I remember watching a video by Buildzoid (an overclocker). He took an Intel CPU and ran up the voltage to show what happens when it's overvolted. It's getting hot at the same time but it didn't buzz out until north of 1.6V. So even 1.52V, indicated in their data sheets, doubtless has plenty of margin too.
I would argue that current is still the primary cause of damage. A short circuit event that causes catastrophic failure can happen extremely fast, before any OCP can trigger. Yes excessive current does lead to heat which is the mechanism that causes damage, but to say temperature alone is the culprit is like saying you can dunk a chip in LN2 and claim it can take lightning strikes all day.

And with regards to voltage, the maximum limit is the safe limit. It's not the absolute limit. Sure you can push the part beyond the safe limit, but you start risking damage to the part beyond that. I've heard a rule of thumb is no more than 10% is fine, so at 1.52V, 1.67V is the "rule of thumb absolute limit", but I'd rather not tempt fate because the part may blow up sooner.
 
...
but to say temperature alone is the culprit is like saying you can dunk a chip in LN2 and claim it can take lightning strikes all day.
...
Ahh...but that conflates two entirely different failure mechanisms: one being electromigration, the other dielectric failure.

I don't think the 'safe' limit with Ryzen is easy to define. Lots of rules of thumb, basically no one has anything definitive. Some have hinted it's determined for each processor at test and fuzed in at manufacturing.

One thing I can say with confidence: a lower voltage won't harm reliability or longevity. We used to de-rate processors (and other components) to operate at a lower voltage specifically to create ultra hi-reliability/low power consumption components for aerospace. That entailed very accurately mapping each individual device's operating envelope to assure stability during low voltage, low stress operating modes.
 
Last edited:

Karadjgne

Titan
Ambassador
Auto is not used because by its very nature, is highly unstable.

Cpu demands change constantly, and almost instantly and auto simply is not fast enough to make that transition when the cpu makes a demand, on the bottom curve of the ripple. That's the purpose of LLC, to apply an additional voltage so that vdroop doesn't lead to not enough power states. But at the same time, LLC is also applying the same voltage to the top of the ripple, so Auto will decrease VID, to keep voltages in safe ranges.

And then the cpu makes a demand, VID isn't high enough, fast enough and the pc crashes.

Having a set voltage and vid and LLC bypasses all that and can remain stable under any load. The only real time there's any stability issues is when voltages and LLC are set incorrectly.
 

ASK THE COMMUNITY

TRENDING THREADS