Undervolting and Curve Optimizer: Ryzen 5900x lifespan question.

[b1k3rdude]

So like lot of people after reading a lot about this and watching a number of tech videos, I have enabled PBO2 & undervolted my 5900x via Curve Optimizer on my Msi MEG X570 ACE. I am seeing great improvements for both single core and all core workloads for example - 

• Prime95 - Single-core was 4825MHz@54c, now 4950Mhz@44c. All-core was 4275Mhz@63c, now 4400Mhz@53c
• CinebenchR20 - Single-core was 4875MHz, now 4950Mhz. All-core was 4175Mhz, now 4375-4400Mhz.

I then told a friend about PBO2/CO as has has the same motherboard, but he said if you lower the voltage then Ohms law means I will then be using more current.

• I measured the power usage in Argus monitor and it showed during an All-core cinebenchR20 an average of 140-145W with a single single spike to 160W. I don't know where Argus is measuring this from so I don't know how accurate it is. I never checked stock power usage before making the changes, but I would have to assume 105W as per the spec of the CPU.

The assumption was then if the CPU is using more current this could potentially shorten the lifespan of the CPU. So we know too much voltage will shorten the life of any CPU, my question is will running the high current draw also do the same thing?

• Stock is 1.45v *105W = 72A, Undervolted/PBO2/CO is 1.31*145W = 110A, so that in increase of 38A

So in a nutshell what I'm asking is -

 

- a referenceable answer (not anecdotal evidence) to 'will extra current (and how much) over stock spec, will shorten the lifespan of the CPU?'

- and lastly how are temps lower despite the increased current draw..?

[Kilrah]

You got the calculation mixed up. Power = voltage * current.

 

So 145W / 1.31V = 110A, 105W/1.45V = 72A.

 

What matters is power and temp basically, current is kinda irrelevant.

[b1k3rdude]

11 minutes ago, Kilrah said:

You got the calculation mixed up. Power = voltage * current.

Yeah spotted and updated.

[mariushm]

Ohm's law is V = IR ... it's easy to remember, just think "virus" ... voltage = current (I) x resistance

 

You can derive the other formulas from it.

 

The minimum voltage required to have all cores stable varies from cpu to cpu, so motherboards are often conservative and give cpus a bit more voltage than needed.

The minimum voltage can also change as time goes by as the CPU silicon degrades, if you overclock heavily... with normal usage the silicon should not degrade within a significant amount of time.

 

Higher voltage can make a cpu achieve higher frequencies, but also causes more leakage and therefore more heat.

 

Higher current is not really a big deal, just the vrm has to compensate harder for the voltage drops in the traces going from the vrm to cpu socket... and it may make the vrm slightly less efficient and therefore produce a bit more heat

Nowadays most motherboard spread the current across 8-12 power stages so 30A extra across 8+ power stages is not that big of a deal.

 

 

[Guest]

55 minutes ago, Kilrah said:

You got the calculation mixed up. Power = voltage * current.

 

So 145W / 1.31V = 110A, 105W/1.45V = 72A.

 

What matters is power and temp basically, current is kinda irrelevant.

105w TDP CPU = 142w PPT which is rated 140 amps... stock.

 

It's all relevant..... cause you can't consume power without voltage and amps, or I'm just not understanding what you are saying....

[b1k3rdude]

@mariushm, I am familiar with ohms law, I just wanted to come up with a title that would stand out and was relevant to the question.

 

So in a nutshell what I'm asking is -

 

- a referenceable answer (not anecdotal evidence) to 'will extra current (and how much) over stock spec, will shorten the lifespan of the CPU?'

- and lastly how are temps lower despite the increased current & power draw..?

[Kilrah]

1 hour ago, ShrimpBrime said:

It's all relevant..... cause you can't consume power without voltage and amps, or I'm just not understanding what you are saying....

I meant the actual current value isn't something you need to care about, it's just a result of the other things you control. You set voltage to get the CPU to run stable at the freq you want, you watch the power since that's what matters for cooling, and current will be... whatever it is, not important. 

 

10 minutes ago, b1k3rdude said:

- a referenceable answer (not anecdotal evidence) to 'will extra current (and how much) over stock spec shorten the lifespan of the CPU?'

You're not going to find any reference for that, nobody really knows, it's out of spec and like the freq you can achieve depends on the actual piece of silicon it'll likely be the same about how resilient to abuse it is.

But in the end the one parameter that matters and is mostly a result of everything else is temperature, and as long as that stays below the max spec temp it should work without premature degradation.

 

13 minutes ago, b1k3rdude said:

- and lastly how are temps lower despite the increased current & power draw..?

Different position on fan curve? Poor measurement? AMD is known for subpar built in power measurement and not very detailed temp reporting.

[Guest]

16 minutes ago, Kilrah said:

and current will be... whatever it is, not important. 

And current will be = not above stock 105w TDP of 140a unless the EDC and PPT limits are increased.

 

So I suppose in the case of not increasing power limits, the current does not matter. I can agree to that.

[unclewebb]

2 hours ago, b1k3rdude said:

if you lower the voltage then Ohms law means I will then be using more current.

What Ohm's law is your friend using?

 

V = I x R

I = V / R

 

If you lower the voltage then the current is reduced. 

 

P = I x V

 

Lowering current and voltage will lower power consumption which will make your CPU run cooler. That is why people undervolt their CPUs.

[Guest]

4 hours ago, b1k3rdude said:

 

So in a nutshell what I'm asking is -

 

- a referenceable answer (not anecdotal evidence) to 'will extra current (and how much) over stock spec, will shorten the lifespan of the CPU?'

- and lastly how are temps lower despite the increased current & power draw..?

The best referenceable answer would be; you will not decrease the life span of the Cpu significantly enough as long as when you manually OC that you stay at or under the FIT voltage.

 

To find the FIT voltage, Enable PBO All other CPU settings stock. Run Prime95 128k FFT with in-place unchecked. Monitor the SVI2 TFN (v-core) sensor. During load, what that reads is your FIT voltage. When manually overclocking, try not to exceed it and that will help prevent early degradation.

 

Other than that, temps only start to matter when you hit the throttle temp. Then you want to evaluate your cooling.

[b1k3rdude]

4 hours ago, Kilrah said:

• and current will be... whatever it is, not important. 
• You're not going to find any reference for that, nobody really knows
• But in the end the one parameter that matters and is mostly a result of everything else is temperature, and as long as that stays below the max spec temp it should work without premature degradation.
• Different position on fan curve? Poor measurement?

• but with respect, this is just an opinion. It looks like I will have email AMD amusing they will answer the question at all.
• AMD should.
• Well as you can see in the OP, I am much lower than stock and much lower than the majority of users. Fyi I am running my ALFII280 in push-pull (partly because the hoses on the AIO just aren't long enough to reach with the rad tubes down.)
• Nope, I never touched the fan curve in the bios. I had 3 different tools all show the same temps.

[Guest]

4 minutes ago, b1k3rdude said:

•  
• Fyi I am running my ALFII280 
•  

Another one of those deals. Buys big board and Cpu can't afford the big cooling. 

 

800$ CPU, you should look at spending 400$ on a custom loop with an all copper waterblock.

 

Anyways, good luck. Ohm's law isn't gonna help you. 

[b1k3rdude]

4 hours ago, unclewebb said:

If you lower the voltage then the current is reduced. 

 

P = I x V

 

Lowering current and voltage will lower power consumption which will make your CPU run cooler. That is why people undervolt their CPUs.

Unfortunately your incorrect on both counts, read the op and the post right after it. We know the wattage and the voltage -

 

P / V = I - (Stock) 105W/1.45v = 72A, (PBO2 & CO) 145W/1.31v = 110A

 

With Curve Optimizer set to 20, that reduces the vcore by 0.14v, the CPU is because of the lower temps turbo'd as high 'it wants' (I am not overclocking) and this increases the power usage, hence the higher current draw. Hence the purpose for the question in this thread.

[xg32]

19 minutes ago, b1k3rdude said:

.

i think you are using the wrong data, skipping steps and overcomplicating thing

 

105w isn't the "spec" of the cpu, it goes over all the time, and going above 145w isn't gonna automatically damage ur cpu.

 

1.45v is single core, not all core and it would not be 72amp (that's why temps are lower, it's single core vs all-core), you cannot assume that 1.31v is an undervolt when some zen 2 cpu could not handle anything over 1.3v

 

The only thing we can assume right now about zen 3, 1.45v single core and 1.3v all core is likely fine as long as you can keep temps under control.

 

 

 

[Guest]

18 minutes ago, b1k3rdude said:

Unfortunately your incorrect on both counts, read the op and the post right after it. We know the wattage and the voltage -

 

P / V = I - (Stock) 105W/1.45v = 72A, (PBO2 & CO) 145W/1.31v = 110A

 

With Curve Optimizer set to 20, that reduces the vcore by 0.14v, the CPU has due to the lower temps increased turbo as high 'it' wants and this increases the power usage, hence the higher current draw.

105w Thermally at max P-state. NOT boost states. It's not 110A, it's 140 amps. 

 

5900X under boost load all core is closer to 230w. 

 

Convert 230w thermally to BTU = 784 BTU an hour.

 

You're not even close on the numbers. 

 

105W (Stock) as you think, is only max P-state frequency and voltage. 

For a 5900X that's "BASE FREQUENCY" of only 3.7GHZ (not sure the voltage, but cpu-z can tell you) = 105w

 

784 bt/h on a 140.2 rad..... ouch. shitty little pump and cold plate. No wonder youz be under-volting. lol.

 

9 minutes ago, b1k3rdude said:

Er what an ill-thought out comment to make. I have to assume your not familiar with the concept of bang for buck?

• Firstly the ALFII280 was clearly demonstrated by several tech channels including LTT!!! as being very good cooling for the money.
• Secondly I was lucky enough to get this CPU at original MSRP.
• Thirdly I have no interest in spending £400 on a custom loop when my ALFII (also at original MSRP), gets me most of the cooling perf (Im sitting at 53c on all-core load in prime95...!!!) of a open loop system without any of the hassle or the risks (and no, not interested in a debate on that subject)

Well, no. That cooler is a POS. I'll gladly tell anyone that even Linus himself. Not everyone's environment will be the same as the reviewer's. 

 

Know the hardware. That means you're thinking this chip is only 105w. HAHAHAHA. Not even close. Turn off PBO and manually set Cpu multi to 37 and post it up. Should be running full load spot on with a 5% swing at 105w. With boosting, it's a completely different story. 

 

So the above, before you thought it's ILL, no please chill. I'm only going to put the truth in front of you. You can get mad, I will remain opinion free.

And as such....

 

I highly doubt an ALF II 280 would compete with a custom loop at any level while there are air coolers that can move just about the same BTU as your AIO. Of course it's great for a 5600X!! Maybe not a 5800X, those are heat monsters by themselve's. You can view the forum listings here at LTT for that issue in it's self. Then you come along with a 5900X, similar cooler as the 5800X dudes and just resort to under-volting and worry about degradation. 

 

The easiest test. 

Load 100% the Cpu at defaults. Run a 2 hour stress test ANY will do. If that chip ever reaches throttle point, it's no good. If it does not, then you shouldn't have the need to under-volt.

 

 

[Guest]

Oh, and the life expectancy of most modern day processors is MTBF 100,000 hours or roughly 10 years under 100% load. That is how your processor is designed to run at defaults. Hopefully that helps give you an idea of what to expect when you do minor changes. you could shave off a couple hours, or a couple of years. Just depends how much luck you want to dump off today. 

[b1k3rdude]

On 4/13/2021 at 10:29 PM, xg32 said:

• 105w isn't the "spec" of the cpu,
• 1.45v is single core, not all core and it would not be 72amp (that's why temps are lower, it's single core vs all-core),
• you cannot assume that 1.31v is an undervolt when some zen 2 cpu could not handle anything over 1.3v

• 105W is the TDP of the CPU, but your right. It isn't the power the CPU uses, so I will need to revert to stock and see what power usage is being reported.
• Ah yes, I will need to do the tests again and check the voltage.
• 1.31v is what showed in the bios, so I made an educated guess that this would be the voltage used for SC full load and AC full load. So as per the above point will run tests again making note of the voltage and power figures.