Just in case 
Overclocking "Raptor Lake refresh" - 14900k,14700k, 14600k etc results, bins and discussion
31 reading
Nizzen
8,681 - 8,700 of 23,546 Posts
I think he's talking about the high LLC unless I am mistaken.
Most definitely but I do think they just couldn't do 16 P cores because of the 10nm package, Custom loops would have struggled I reckon.
Joined
·
2,224 Posts
Here is an Intel mod on the Intel forum telling a new user that under 1.72v spec is fine.
community.intel.com/t5/Processors/Voltage-of-I9-13900k/m-p/1563483#M68549
Joined
·
13,203 Posts
Absurd.
I guess degradation is a myth and doesn't happen unless you're FALKENTYNE, right?
You guys can choose who to believe. Someone who put their time in or someone who doesn't understand context.
ONE WEEK of 1.510v at LLC3 (1.1 mohm) LLC at idle and my 13900K can no longer loop R23 at 5.7 ghz regardless of voltage.
Used to be able to do it with 1.370v set + LLC6 for 30 minutes (after it was already degraded with Stockfish).
But you guys can do whatever the hell you want and believe what you want. Your choice.
Joined
·
1,889 Posts
But the CPU should never idle at 1.5v on auto voltages even with the power plan set to high performance on auto or adaptive voltage.
You would have to have manually set the CPU to run at like 1.55 "actual" Vcore voltage which I don't think a single person in here is gonna do.
Joined
·
3,724 Posts
1.55 in BIOS is totally in the normal realm for cold water benchmark runs. I ran 1.41v vmin on my 13700KF a few times.
And 1.60 set in Bios with a droopy llc (1.32 Vmin) on my Z790 Dark for 7+ months as a daily. Profile would consume ~350 watts under full load (I have hundreds of hours of y-cruncher on that chip at 330+W, and keep in mind it only has half the e cores). 300+W when running 8P with e cores disabled.
EDIT: forgot to mention with no observable degradation!
Joined
·
1,889 Posts
What I am saying is not many people would be using "Actual VRM Core voltage" in BIOS of 1.55 with LLC3 which is what yields 1.51 Actual "Vcore" voltage at idle which since it's on Actual VRM Core Voltage in the Bios then at idle it's just stuck at 1.510v Actual Vcore.
"Most" people doing cold water benching would be using something like LLC6/7 with 1.3/1.4 Actual VRM Core Voltage, This would yield much lower ldle voltage.
Joined
·
2,224 Posts
Having auto vrm vcore and adaptive on core voltage is a totally different story.
I can't even get that voltage unless I put a jumper across a set of pins on board with Red warning in the bios 😂
I guess that mod wants users to do that so they can fail their RMA's and get their stock price higher for buying new CPUs.
Its really the spec... voltage doesn't destroy the chip (unless its way to much and punches through structures that are not made to transport current). All the electro migration comes from current and temperature. Your 30 minutes 400W or so run in CB did more harm to the chip than some idle voltage.
If 1.5V idle would be a problem, intel would have massive RMAs from all the VID 1.501V chips on cheap boards. And mine would be dead - I run it easily in that voltage range, even under light load. Makes 150W at 1.45V... runs fine.
but who knows... we can not look inside that chips...
It's not temperature lol, All silicon that is designed to run at 95-110c can if everything else is stock run at 95-110c all day long.
It's mainly current with voltage.
You could run a 14900K/KF with a puny 120mm fan air cooler all day long bouncing of 100c and it will not degrade because it's designed to throttle back if everything else is stock.
When you are not stock then you are heading into the unknown of how much voltage is acceptable with what current.
Joined
·
1,889 Posts
Note what falkentyne said, he degraded his chip with 1.51v idling for a week, this to me entails the use of manual voltage since auto and adaptive will scale down the voltage at idle.
My 104sp 13900ks (1.484v 6ghz vid) hits between 1.5-1.55v at idle with an llc of 3, AC_LL=.41 (any lower was unstable) and DC_LL=1.14 (so my vcore matched my vid). svid set to adaptive mode and everything else auto, ICCMax autoing to 500a I believe. The die sense reported v core usually shows maximums of 1.51-1.52. During r23 my vcore/vid drops to 1.29v @ 5.6ghz p core 4.5ghz ecore with 320w of power draw, max temps of 95-97c, idle temp around 35-40c. No throttling.
If intel is shipping cpus with vid tables that high then my guess is that they consider 1.5v+ idle to be safe. I’ve personally switched to using llc 5 (AC_LL=.21 and DC_LL=.53) to keep my idle voltage just under 1.5v, but I think an llc of 3 would also have been fine and I had been using that for a while. Haven’t noticed any degradation. The all load voltage/power draw/temperatures is the same for me between LLC3 and LLC5. Only difference is lower idle voltage with llc5 (although with the higher overshoot this might not be reality regardless of what HWInfo says. Not sure)
Using tvb voltage optimizations reduced the idle voltage considerably for me (down to sub 1.4v) but was only stable at stock ratios. If I wanted to use the tvb +2 boost profile I had to disable it. The max reported voltage with +2 boost enabled/disabled with TVB optimizations off (so 6.2ghz max boost) was the same. Kinda wish there was more control over that feature to modify the voltage/degree reduction amount.
Manual voltage may be a problem tho... SVID has no chance to power limit... so the 1.5V manual is not the issue... its the 400A current spikes that you kick through the cores every time a process does something. anyway... I let you know if my CPU degrades... currently I run 2-4 core 7 zip benchmarks at around 1.45V.
Joined
·
1,889 Posts
Your running adaptive so your not running 1.5v at idle.
VID and vcore are 2 separate things.
Can you elaborate? Not saying you're wrong just trying to understand. The die sense reported vcore in hwinfo (not just the vid) for me at idle is over 1.5v at llc3 on adapative mode. Same thing when reading the raw vid reporting in asus octool tool. So 1.5v+ reported vcore at idle on adaptive mode is different and not unsafe compared to 1.5v+ vcore at idle with a manual voltage?
Joined
·
1,889 Posts
let the pc idle, what is the reported die sense voltage at idle? Not the max in HWinfo but the current.
This is at idle with LLC3, AC_LL .42 (lowest stable in R23) and DC_LL 1.14 (so vid matches vcore at full load) with adapative mode enabled, TVB +2 Boost enabled, TVB voltage optimizations disabled, C-States/everything else auto. I guess then the 1.5v+ peaks aren't cause for concern? Just as long as its not staying at 1.5v+?
![Image]()
Joined
·
1,995 Posts
Punch in 1.5V LLC6 in the BIOS and your dreams will come true in no time!i hope i experience cpu core degradation at some point. for now it almost always sounds like a fairy tale
Yeah, I think the problem might not be idle or even all core load, it's very likely heavy 1-2 core loads when using fixed vcore. When you set 1.45V+ LLC6 and start hammering one core, the LLC will pretty much do nothing, and the loaded core is getting almost the same voltage as idle. I degraded my 12900K very badly by just playing old single core games with 1.505V LLC4.
As someone who is too lazy to actually do it and will instead keep degrading CPUs, I must say it: we really shouldn't use fixed vcore with these chips anymore.
8,681 - 8,700 of 23,546 Posts
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
