Overclock.net banner
61 - 80 of 94 Posts

·
Registered
Joined
·
807 Posts
I read this thread twice. Really great info here. I really wish there was an easy way of recording voltage overshoot like we can ( sort of) with vdroop when looking at vccin under load. With my 7900X, I didn't allow for much vdroop, but with this 10980 I set vccin to ~1.8-1.82 and lower my LLC as much as I can without causing issues. Currently at 1.8V vccin and LLC4 which allows the vccin to drop to ~ 1.74. So far everything seems fine. One thing I have heard is that higher vccin voltages can cause excessive heat. If this is true, it gives further advantage of using a lower vccin.
 

·
Registered
Joined
·
144 Posts
I've had this thread bookmarked for a couple years, I look at it every now and then. My 8700k is 3 years old and I've recently went from 4.9ghz 1.325v to 5.1 1.4v, I hope it holds up at this voltage, if not I'm gonna run this puppy into the ground
 

·
Registered
Joined
·
83 Posts
Happy pandemic everyone! rofl

Thanks for all your messages, I'm really glad so many of you found my experiment and the implications I tried to deduce worthwhile. Really appreciate your feedback. Hope it's been of any help for those who've been wondering the same but couldn't put their chips at stake.

Must say I regret being away for so long. Past couple of years have been extremely busy and this month actually marks my tenth year in the semiconductor industry. I reckon the level of stress my chips endured back then is no compare to what I've gone through since last year. Apparently the definition of telecommuting is being ordered to take care of doubled or trebled workload 24/7. Now that my workload recently got quadrupled on top of the fact that I will be forced to leave the country by the time they finally complete the vaccine rollout to pensioners, I decided to silently bow out of work and spend some quality time back on technology forums. My employers do not ought to know this. lol
 

·
Registered
Joined
·
2,175 Posts
Hi Guys,

I'm seeing a lot of "1.xx is the max safe voltage for [insert CPU]" and was wondering what your experiences are with this. Have any of you ever killed or damaged a CPU due to overvolting?

A lot of the numbers for max safe volts I'm seeing seem to be personal preference with little or no facts to back it up, the range of "safe" volts i've seen goes from 1.35-1.52 which is a little broad!!

I'm no expert in this field and my overclocking 'technique' is a little odd, I always go for voltages that many would deem too high, my 8700K is at 1.47 volts and my 4790K before this was at a high voltage too (can't remember off the top of my head) and both are running fine... the 8700K has only had a couple of months of use so there's plenty of time for degradation to show but the 4790K has been going strong for years with no issues.

So please share your experiences... I'd love to see some examples of degradation especially with recent intel cpus, I'd be happy to pump 1.55 volts through this 8700K if i was relatively sure it'd last 2-3 years
Not really. No don't pump 1.55 through a 8700k as it won't last and ontop of that you won't keep it cool at them voltages unless you are sub ambient aka -30c.

That being said I slammed 1.45 through my kaby lake quad core the other day on water lol.
 

·
Registered
Joined
·
39 Posts
Hi guys,

I have been studying the fascinating graph that audiotest kindly shared - thank you, it is much appreciated!

I appreciate that these results are liberal extrapolations; I understand the 5mv line from the experiment; however, I am having tremendous difficulty understanding the 50mv and 100mv lines. Maybe I'm just too stupid...

"Then if I assume that the amount of voltage shift degradation has also an exponential behaviour, I can predict another trajectory namely the 50 mV degradation curve."
Why? Surely the relationship is linear? If it takes 1 hour to degrade X, then it takes 2 hours to degrade 2X, or 10 hours to degrade 10x?

For example, at 1.45v, it takes 1,000 hours for stability to degrade 5mv. Logic then dictates that it would take 10,000 hours to degrade 50mv because it is correctly 10x as long. The graph however says it would take 100,000 hours which is 100x longer. 50mv divided by 5mv is 10, not 100.

The 100mv line is problematic too: At 1.5v, 50mv degradation would be 20,000 hours so 100mv degradation ought to be double that which is 40,000 hours. The graph is unfortunately showing 70,000 hours?!

I hope audiotest doesn't mind, but I have taken the liberty of adding logarithmic graph lines to the Y axis, and have put what I think ought to be the correct lines for 50mv and 100mv as solid orange/red lines respectively. Dotted lines are the originals by audiotest:

tty.png


As an example, it now 'correctly' shows that at 1.45v, if 5mv degradation takes 1,000 hours, then a 50mv degradation will take 10,000 hours; and a 100mv degradation will take 20,000 hours...

Maybe I have the wrong end of the stick and audiotest has used some clever logarithmic scale which isn't obvious to me...

----------------------------
The degradation projections of common vcore voltages on a 24/7 system (running v-ray?):

vcore = 1.400v, Degradation of 0.005v = 8 months until you need to up the voltage to 1.405v
vcore = 1.420v, Degradation of 0.005v = 4 months until you need to up the voltage to 1.425v
vcore = 1.440v, Degradation of 0.005v = 2 months until you need to up the voltage to 1.445v
vcore = 1.460v, Degradation of 0.005v = 1 months until you need to up the voltage to 1.465v

These results seem to beckon concern...
 

·
Registered
Joined
·
83 Posts
Hi guys,

I have been studying the fascinating graph that audiotest kindly shared - thank you, it is much appreciated!

I appreciate that these results are liberal extrapolations; I understand the 5mv line from the experiment; however, I am having tremendous difficulty understanding the 50mv and 100mv lines. Maybe I'm just too stupid...


Why? Surely the relationship is linear? If it takes 1 hour to degrade X, then it takes 2 hours to degrade 2X, or 10 hours to degrade 10x?

For example, at 1.45v, it takes 1,000 hours for stability to degrade 5mv. Logic then dictates that it would take 10,000 hours to degrade 50mv because it is correctly 10x as long. The graph however says it would take 100,000 hours which is 100x longer. 50mv divided by 5mv is 10, not 100.

The 100mv line is problematic too: At 1.5v, 50mv degradation would be 20,000 hours so 100mv degradation ought to be double that which is 40,000 hours. The graph is unfortunately showing 70,000 hours?!

I hope audiotest doesn't mind, but I have taken the liberty of adding logarithmic graph lines to the Y axis, and have put what I think ought to be the correct lines for 50mv and 100mv as solid orange/red lines respectively. Dotted lines are the originals by audiotest:

View attachment 2481086

As an example, it now 'correctly' shows that at 1.45v, if 5mv degradation takes 1,000 hours, then a 50mv degradation will take 10,000 hours; and a 100mv degradation will take 20,000 hours...

Maybe I have the wrong end of the stick and audiotest has used some clever logarithmic scale which isn't obvious to me...

----------------------------
The degradation projections of common vcore voltages on a 24/7 system (running v-ray?):

vcore = 1.400v, Degradation of 0.005v = 8 months until you need to up the voltage to 1.405v
vcore = 1.420v, Degradation of 0.005v = 4 months until you need to up the voltage to 1.425v
vcore = 1.440v, Degradation of 0.005v = 2 months until you need to up the voltage to 1.445v
vcore = 1.460v, Degradation of 0.005v = 1 months until you need to up the voltage to 1.465v

These results seem to beckon concern...
Thank you very much mate! That's a brilliant catch and you're absolutely right. There should've been a factor of 10 difference between 5 and 50 mV. That explains the four fold difference between 50 and 100 instead of two. Shame that I don't have access to my old Skylake build with MS Office anymore, in which I'd plotted the curves (yes, they are log functions). I used the blue log regression function Excel gave me the to render orange and red curves. I reckon I made a 10*log(x) vs. 20*log(x) error of some sort while creating those two other functions. Thanks for the correction, gotta love this forum.
 

·
Registered
Joined
·
83 Posts
Well, just read my post once again. It's either that or I actually implemented the 10x error margin into the 50 mV curve. Can't tell which, my memory's gotten really weak. I don't even remember what life was like before the pandemic. lol
 

·
Registered
Joined
·
348 Posts

·
Registered
Joined
·
928 Posts
Hi guys,

I have been studying the fascinating graph that audiotest kindly shared - thank you, it is much appreciated!

I appreciate that these results are liberal extrapolations; I understand the 5mv line from the experiment; however, I am having tremendous difficulty understanding the 50mv and 100mv lines. Maybe I'm just too stupid...
No you are not, and also you are not an expert at electrical test and measurement practices and facts.
There is no such voltage Degradation in electronics.
There is no power / energy calculations with out taking temperature as another factor of measured influence.
 

·
Registered
Joined
·
928 Posts
In my computing life since 1997, I did kill one CPU but not due voltage, we can not imagine back then that TEC CPU cooling this will cause 20C difference when CPU this is IDLE and or at LOAD state.
CPU this is no mercury thermometer, and 20C up and down, at every CPU usage, this damaged the metallic CPU interconnections.
I have this TEC as souvenir, keeping it for 23 years, as Never Use for CPU cooling again, reminder.
 

·
Registered
Joined
·
2,175 Posts
No you are not, and also you are not an expert at electrical test and measurement practices and facts.
There is no such voltage Degradation in electronics.
There is no power / energy calculations with out taking temperature as another factor of measured influence.
Are you ******ed. Electromigration is clearly a thing. Wanna step back to netburst aka sudden northwood death syndrome. That was pure electromigration killing cpus. It is a thing and will always be a thing. Ive degraded core 2 cpus running 1.5 vcore daily under water way back. Its a issue. You wont fully kill a cpu unless you slam like 1.7 vcore through it if its 45nm or lower.
 

·
Facepalm
Joined
·
10,014 Posts
No you are not, and also you are not an expert at electrical test and measurement practices and facts.
There is no such voltage Degradation in electronics.
There is no power / energy calculations with out taking temperature as another factor of measured influence.
You're either drunk, high or deranged. Or just an old person troll.
 
  • Rep+
Reactions: o1dschoo1

·
Registered
Joined
·
928 Posts
If I start debate's with all the ones that disagree with me, then I will damage the amount of free creative time that I could spend to research.
 

·
Registered
Joined
·
2,175 Posts
If I start debate's with all the ones that disagree with me, then I will damage the amount of free creative time that I could spend to research.
Or you just admitted you are wrong and backing down.
Know why intel recommends a max voltage? Cause electromigration aka voltage death will occur over a period of time. Aka you knocked 10 years off of the life of your cpu. I run my cpus well above intel recommended but replace them every few years. From personal experience electromigration is real. This is from someone who ran 1.45 vcore daily on a 45nm core 2 duo and benched it at 1.5+vcore. At stock vcore cpus last 20+ years anything over stock you start knocking time off so that literally contradicts what you said.

Know why intel couldn't put out 5ghz netburst cpus? Electromigration from too much voltage and heat.
Hell even turbo boost pushed the voltage on single core load for these new chips but it's not enough to kill your cpu in a few years. Id say 10 maybe but the pc will be outta commission by then.
Hell i slapped 1.55 vcore through my xeon w3520 for a 4.7ghz boot and that was super risky on water.
Edit #3.
If anyone knows about electromigration it would be intel. This is the sole reason they went from netburst back to a modified pentium m design for core 2 which is based on pent 3. More cores more efficient aka same tdp with more cores and less vcore/vtt and less vrm strain.
The elongated pipeline of netburst allowed the clocks to run high but 3.8 was the limit. 4ghz plus started to require high voltage and caused issues with electromigration and tdp. In 07 i had a 3.6ghz prescott single core kaying around i slapped in my pc for giggles to see what it could clock to. Guess what 4.5 was the max at 1.55 on water. I tried up to 1.7 for 5ghz and wasnt happening. And anyone thst knows the 65nm process 1.7 on anything but sub zero is a no no.
 

·
Registered
Joined
·
928 Posts
Or you just admitted you are wrong and backing down.
With out holding 100$ in your hand, ready to be send at my account, all your questions, they will stay unanswered.
Billions of people power on computers every day, this does not make them experts.
 

·
Registered
Joined
·
2,175 Posts
With out holding 100$ in your hand, ready to be send at my account, all your questions, they will stay unanswered.
Billions of people power on computers every day, this does not make them experts.
Big bet id drop 100 dollars down to watch intel themselves prove you wrong. Everything i said is purely based on why intel themselves backed the clocks back down from 3+ghz when going to first gen core 2.
 

·
Registered
Joined
·
2,175 Posts
With out holding 100$ in your hand, ready to be send at my account, all your questions, they will stay unanswered.
Billions of people power on computers every day, this does not make them experts.
From intels website. electromigration Definition. So you are saying intel doesn't know what they are talking about? Mr electrical wannabe engineer thinks he knows more than the people that develop the damm silicon.
 

·
Registered
Joined
·
928 Posts
Mr electrical wannabe engineer thinks he knows more than the people that develop the damm silicon.
The silicon developers they leave out for you crumples which they are not correlated with the reality.
They do not worry of a farmer keystrokes, but for their competitors.
 

·
Registered
Joined
·
2,175 Posts
The silicon developers they leave out for you crumples which they are not correlated with the reality.
They do not worry of a farmer keystrokes, but for their competitors.
Lmao is that a conspiracy theory? Id believe the silicon developer over some hot shot electrical engineer. You might wanna do some research cause its been discussed outside of intel bringing it to the light. Look up sudden northwood death syndrom. Cpus suddenly died when put above 1.7 vcore. That is voltage death aka electro migration. Or you havnt been around pcs long enough to know that or you dont know as much as you claim. Pick your poison. You are wrong. Read this thread. It clearly points out electromigration happening.
2515763
 
61 - 80 of 94 Posts
Top