How efficient does Asus M5A88-V EVO handle voltage spikes from CPU LLC?

[Stealth3si]

I don't have an oscillope to test this so any knowledable asnwer from experience would suffice. a tested formula to calculate overshoot perhaps?

i mean even if it weren't possible for spikes to occur from LLC, if there was woudl my mobo be up for the task to prevent damage to the CPU?

Do you think the mobo cna handle the possible spikes as a result of CPU LLC to keep it under the maximum safe voltage of 1.55V, or posibbly even 1.50V or 1.475?

Say for isntance with LLC enabled, the Load Vcore = 1.43V -1.45V.

At this point i'm not so much concerned about 24/7 OC stability issues as I am about possible CPU damage due to spikes or relatively signficant decrease in life span.

EIDT: Oops sorry posted in wrong thread...admin please move to appropriate thread. thank you.

 

[trents]

You can check those spikes you know simply by having HWMonitor open on the desktop while you run Prime95. They would show in the "Max" column for CPU vcore. Other than that, I'm not really sure what you are asking as your post kind of rambles.

 

[Stealth3si]

really? i thought software can't detect such spikes because they happen so quickly we don't notice it. what you said is what i would presumably think but i've read googled about LLC, vdroop, spikes, under/over shoots, etc, etc which seem to suggest that software can't read it. i will take what you say with a grain of salt though thank you.


as for the rambles, let me put it this way: under LLC, when the volts increase under load I get a constant voltage from LLC because it eliminates Vdroop. But when the system goes from load to idle state that load voltage will spike at the current voltage load level before dropping to voltage idle level.

for isntance, say i put 1.34V in bios but actual windows idle is 1.37V and load is 1.45V. Spkies occuur at 1.45V mark before it goes back down to 1.37V Idle. thsi means voltage level will land between 1.45V and 1.55V before going back down to 1.37V.

make sense?

 

[EarthDog]

If you would like your thread moved, report a post from within it to notify the moderation team. Otherwise, they will just have to happen to come across the thread.

Anyway, I wouldnt worry about it. LLC has been around for a couple of generations and I cannot think of one instance that LLC has toasted a chip due to overshoot. People run those chips at 1.55v for significant periods of time with proper cooling (talking ambient), so I would imagine that blips will not do anything. Non issue.

 

[Bobnova]

The LLC overshoot thing may or may not be true.
The Anandtech article that made it a Big Deal did no actual testing at all (or if they did, they didn't show any results from it), they read an Intel paper, whipped up some pretty charts showing a pretend voltage spike, and passed it onward as proven fact.

I tested it a couple years ago with the best scope I could lay my hands on and saw absolutely nothing.
I have access to a modern digital oscilloscope capable of seeing such a spike even with modern insanely high switching speeds (my analog scope will see it, but the image won't stick. It will on the digital), so I'll test it at some point here.


In short, don't worry about it.

 

[Stealth3si]

The LLC overshoot thing may or may not be true.
The Anandtech article that made it a Big Deal did no actual testing at all (or if they did, they didn't show any results from it), they read an Intel paper, whipped up some pretty charts showing a pretend voltage spike, and passed it onward as proven fact.

I tested it a couple years ago with the best scope I could lay my hands on and saw absolutely nothing.
I have access to a modern digital oscilloscope capable of seeing such a spike even with modern insanely high switching speeds (my analog scope will see it, but the image won't stick. It will on the digital), so I'll test it at some point here.


In short, don't worry about it.

Jumping jupiter it's the infamous bobnova!

In my experimenting with LLC and googled about this, vdroop, etc and eventually about cpu spikes thta got me worried.

then i read yoru article right after reading Master's Lairarticle.

so i just wante dto be sure you know thanks for the enligtening update on this issue.

 

[Bobnova]

Masters pretty much just copy/pasted the pretend "data" from anandtech

Realistically, even if you do believe the handmade charts, 40mv is going to do a huge load of absolutely nothing to the CPU.

 

[Stealth3si]

yea thanks for clarification bob..

heh looks like your article closed that can of worms..excellent job btw

and come to think of it lots of stuff changed since that anandtech aritcle...technology, computer products quality, etc, etc.

+1

 

[Bobnova]

Like I said I do plan to retest with a modern, better, scope. It may be a bit though, there's a lot on my plate.

 

[TsunamiJuan]

I think I am getting roped into this project. The Two of us are currently Talking strategy about how to conquer this under a Revised method of testing.

 

[trents]

really? i thought software can't detect such spikes because they happen so quickly we don't notice it . . ./QUOTE]

Then how do you know it's even happening?

 

[TsunamiJuan]

really? i thought software can't detect such spikes because they happen so quickly we don't notice it . . ./QUOTE]

Then how do you know it's even happening?

While the software is not really gonna be able to pickup sudden extreme over voltages of super short duration it can pick them up to a small degree if they show up while its taking a sample of the voltages.

After all it only takes a sample of the voltage every X seconds or so. Instead of constantly looking for voltage changes and showing them on change.