it''s not the MB makers, it's happening on 13/14 gen servers which DO NOT over clock/voltage or anything else since reliability and stability are primary over speed. intel is flat out lying, there's something wrong in their raptor lake architecture. that they're trying to blame board partners and users for.
I hope for a class action lawsuit against the motherboard manufacturers for over volting our cpus without our consent or knowledge and no easy or standard way to return the voltage back to intels specifications! This is disgraceful for all motherboard manufacturers to behave in such a disrespectful way towards their customers!
I have asus motherboard and it's super easy. Thers almost basic option in advance settings that you change from motherboard overdrive to Intel defaults. And its just 1 click. But ea it should be off by default setting no other way around
10:50: "If you are not overclocking, do not worry about it." From what I have heard, even workstation and server boards with no overclocking features whatsoever have stability issues with 13th and 14th gen CPUs.
Don't get me wrong this is a massive issue but one of the most wide-spread pieces of misinformation is this. It comes from Wendell at level 1 Tech, but it is not true, these W series motherboards were also running at a 4095w PL1 and PL2 just like "Gaming" Z790 boards. The same settings as these boards but they do not have the option to change them, so Wendell assumed they were 125PL1/253PL2. This issue is due to it's default VID table being too high mixed with a 1.1AC Loadline which causes overshoot to 1.5/1.6v, so it doesn't matter but this was misinformation being parroted. Pushing 415+ watts through a chip with no power limits does exacerbate the problem obviously. These servers are setup to run single-threaded workloads 24/7 (I.E Minecraft servers) which causes the 1.6v/6ghz boost to hit constantly leading to the higher failure rate than the average person running these out of box settings. Massive issue and Intel is at fault, but we have to be true to ourselves and once we are aware this is wrong we need to stop parroting it.
@@besim109AOG Because we now know that even conservatively clocked professional grade boards are seeing degradation. That implies that it's a fundamental problem with the chips, not (just) what board makers where doing. Not to mention, we already knew that at least some of the official specs from Intel say what at least some of the motherboard manufacturers where doing is in spec.
I’m only 2/3 of the way through the video and that was exactly what I was going to say. The fact that both the mobos are pushing these cpus past their specs, what the cpus are being advertised as being capable of is actually redlining them and if brought down to intel specs you don’t get speeds advertised, and lastly and most importantly, 50% or more new intel chips seem to be degrading with no overclocks whatsoever while being used for servers which is just damning. This guy is implying that the optimizations the mobo is applying won’t significantly cut the lifespan of your cpu which is not the case. They’re speeding up the degredation.
@@ahettinger525You understand motherboard has nothing to do with Intel's recommended voltage right? People talking about voltage doesnt necessarily mean they are talking about motherboards. You know that ye?
I had the same CPU as a daily from 2010 to Jan 2020, OC to 3.3, (just like my slot 1 P2-400) its now in the loft. Then upgraded to 3950X, will I get 10 years (hope so). BTW those CPU still work.
I was running an i7 920 from 2009 to about 2014, then upgraded to an i7 980X which I replaced just a month ago. During this time I had to replace 3 DIMMs, but both CPUs were (and still are) ok, even though I did overclock both and kept them running with close to 100% load and 24/7 for most of the time. The second and third DIMM I did buy used though, so there's that too. I am guessing the DIMMs themselves did hold up for about 10 years. CPUs were water-cooled though, so never really saw more than 75°C. The 980X was running with all 6 DIMM slots equipped for many years (as you may know the memory controller of the 980X was said to be a weak spot and putting full load on all channels on it was going to be more risky).
*_Battlefield 4_* was my go-to stress for several years on gaming PCs. It had ways of finding overall system instability that individual component-specific stress tests frequently seemed miss.
Funny because Battlefield 2042 is now the best game to test RAM stability, even catching errors that memtest programs don't see. The Frostbite engine uses your PC to the fullest 😄
I have seen a bunch of CPU's with 40 years working flawlessly, even today. Old RAM chips, specially from certain vendors, is another story, some last ages and are still reliable today, others degraded after a few decades. But CPU's, I never saw one just naturally degrade or stop working after a few decades, not with normal default use. So this conversation sounds a tad off to me, claiming that CPU's were made to last 3 years (or the guarantee period) is very strange, maybe today but that was certainly not the case in the 80's to the early 2000's, no one would buy a Intel or AMD CPU, if they only lasted 3 years. And big tech companies, will never accept "silicon degradation is natural" when buying huge amounts of CPU's for their computers or servers. Intel and AMD may wake up one day and see ARM occupying their place, if x86/64 CPU's start dying earlier than users expect or find normal, as being the case with certain Intel CPU's recently.
What he said about smaller parts does have some merit. Remember, many of these older CPUs were on a different process node. Intel for the longest time were on 14nm and finally got to 10nm. As foundries make parts ever so smaller, you have to start to wonder just how small is too small? This is one of the reasons why TSMC has been using machine learning to run these simulations not to just find the best way to tap out a design but to research on all the ways a design can fail. This is one of the biggest reason why Synopsis bought cadence because the tap out process is so complex and expensive, designers and manufacturers must work hand in hand to iterate their design to find one that will satisfy design requirements.
I had a 4770k, @4.6ghz 1.29v with 2133mhz ram. it then began to require more voltage to keep that overclock up until 1.318v, where I had to tune it down to 4.5ghz. It still required 1.31v to operate at 4.5, so still more than the OG 1.29 at 4.6. It never died, I sold it later in 2020 so I don't know if it had degraded further or what.
@@JMatrx mine just simply died, it was working 100% one day, the next day the PC did not boot, I tried the CPU in a different machine and it would not boot also, I placed a different CPU in the same MB and it worked fine.
My motto is to buy robust/stable enough stock hardware for the resolution that I'm aiming to play at (high/ultra quality 1440p in my case), so that I don't need to overclock anything & add stress/instability to the system. I see overclocking as more of a game onto itself for the bench-marking e-sport crowd.
I've had at least 2 components where i'm pretty sure i've experienced degradation, one CPU and one GPU. Both were ran with a custom OC including increased voltage, and had to be clocked down and/or voltage increased to keep stable after some years. So, i would actually recommend a lot stronger to worry about this. As soon as you touch voltages, always keep this in mind and think about if 1% more performance really is worth increasing the voltage by 5% or 10% with voltage having an exponential effect on longevity.
I've degraded one CPU, a 6800K. I was pushing it to 4.3GHz and Broadwell-E was terrible at OC. It took a lot to get that stable, and it was stable and I was running it like that daily. After about a year, it started to become unstable and I had to walk the OC back to 4.2GHz to be stable again. Eventually I just decided to run it at stock rather than risk further degradation. With the new CPU's, I'm less worried about stock voltage degrading the CPU than temps. Perhaps I'm not understanding how this impacts CPU's, but you can degrade memory with prolonged high temps, a serious concern when buying a used mining GPU. And this isn't an issue for most people, but if you're doing a lot of work that has the CPU running high usage in an all core workload, pulling down 250+ watts, and the temp is in the 90's. I feel like that's not a good situation if you're doing that on a regular basis. Seems like degradation would be a concern for someone like that. There have been a lot of Ryzen 5 3600's that bit the dust prematurely, and one explanation was the inadequate CPU coolers that were included could be causing the CPU to run too hot for extended periods of time. So, for me, I'm more concerned about the temps and I limit the PL1 and PL2 to keep the temps more reasonable in those workloads and shouldn't really impact single core performance.
@@delresearch5416 I mean, Devil's Canyon was a pretty good overclocker. Regardless, if the voltages and temps are kept in check, I'm not surprised it's still going strong. It's going to be interesting to see how these new CPU's hold over time.
If you have an Coffee Lake R desktop K chip or newer from intel, never let the motherboard control what's your voltage/frequency, always follow intel specification. you'll have lower temps, quieter fan and if you don't have a really good custom water loop, you'll have better performance as well by going stock. Extra points for slightly undervolting your cpu.
Would love to see motherboard manufacturers have an "AI" undervolting toggle where they continuously try small increments of undervolting until it becomes unstable and automatically sets it for you. Wouldn't that be way more advantageous to the average consumer who doesn't know how to adjuat settings, rather than cranking all their limits to the max out of the box that their cooler can't handle anywyas?
Probably but they’re not doing it because it’s best for consumers. They’re doing it so inexperienced reviewers (who often are the ones who will review a lot of motherboards) will run it against several and the one that blasts the power limit will perform better. Less true with the recent i9s but historically and with other SKUs you get an extra 100mhz and the reviewer concludes “asus has the best performance”.
It's not quite as sofisticated but AMD already has a feature to automatically find stable undervolt/negative offset curve optimizer settings in the ryzen master application now. It's not perfect but it makes it very easy if you want to quickly undervolt the cpu without spending a lot of time manually testing, just use that feature and set the values a tad higher. Doing a completely auto undervolt by default would probably still be too problematic stability wise and also require the board to take a lot of time stability testing (which is already a thing with memry training, especially on the threadripper workstation boards)
@@olnnnthey also got that in the adrenaline software for the GPU. Won't make a hughe difference, but i encourage everyone to do it. Literally 0 risk of damage, same stability as base voltages since it is quite conservative, and you save 1-4% power input for literally free for the exact same performance. (And in summer that also means a couple watts less heat, which is nice)
You don't need AI fuckin' garbage for that. It's no different than having a basic coded sequence to test stability the same way MSI Afterburner does it. Similar procedures are done through system controllers with DDR5 RAM training.
I still have a fx8320 system running in the house. Still running strong on a 4.4ghz oc. But I steer clear of crazy oc'ing. And I'm obsessed with low temperatures.
lol now I know why this channel is dead compared to my reputable outlets. It’s full of Intel fanboys blaming motherboard manufacturers when Intel encouraged this by openly stating to the press that having no voltage regulation would still be “in spec”.
I have a 4790K that I've been running stock in one way or another for the past 10 years, first in my main rig, and later in my HTPC. It has never been overclocked and has only been run for a few months with multi core enhancement on (default option on Z87 boards, against Intel spec). During the past few years, it has simply stopped being able to boost to the highest stock frequency (resulting in immediate BSOD/hard lock). This could, of course, be a motherboard issue, but my money's on the CPU. If It is the only case where I have seen such obvious degradation. Someone in another comment mentioned the i7 920 - I also used this CPU for around 7-8 years continuously, and with a slight OC at stock voltage on a X58 board. Never noticed any instability issues, until the board just up and died one day.
Depends on temperature. My Phenom II x6 lasted 11 years @ 4ghz OC @ 200W. The only reason it was viable for so long was an extremely aggressive 777-22 memory timing.
I’ve had a hell of a time trying to get my 12900k to run stable. The only thing left to replace is the cpu. Are there guides to help you know what to turn off in the bios to have a chip run stock? Is it possible for the consumer part of the industry to insist they add an “intel stock button” in the bios so that us dummies out there have an easy option to run tests at stock? At least have that as an option over the default no limits configs? Just a thought Cheers Rick
My 2500k was 8 years on 4.9ghz 1.425v and still works on the same voltages as the 1st day. All my CPUs are maxed on with everything including BCLK. So far so good.
You're lucky. I never did any OC on my 2500k, and my CPU + motherboard died after 3.5 years. Get them replaced using the warranty still works while OC without additional voltage . My R5 1600x showed IMC degradation after 5/6 years of usage. Again, everything was at default settings except for memory frequency and and latency that I had to setup manually because I bought overpriced fast memory with XMP profile for frequency that my CPU/motherboard could not reach. I tightened latency as much as I could but with default "auto" voltage settings. After 5 or 6 years, I had crashes and had to slow down a bit the DDR frequency.
I had a 9900K that held at 5.2Ghz for years, I got it the day it was released in 2018, around october of 2023 it started BSOD'ing and when I limited it to 4.8Ghz it was stable. It is running fine in my 14 year olds PC now and she doesn't even notice or care that it doesn't boost to 5Ghz anymore. I upgraded to a 14900K
Nice twisted damage control video. FACT is that the i7-s and i9-s are affected by it and probably potentially around 90 % of the CPU-s sold. The problem is that not everyone knows it and not everyone is going to link their BSOD-s to the CPU melting and not everyone is using it to play games constantly. If you take that in account, probably around 50 % of the examples are going to fail eventually and very soon.
I never overclock until ive had the cpu for a good number of years. My thinking is overclocking is not something you do until you need that extra performance usually near the end of the CPU lifecycle.
A good analogy I thought of regarding degredation would be like tires on your car, if you just do normal driving the tread will last for the warranty period, but if you are always pushing it by doing burnouts everytime you stop and start the tires will go bald more quickly (degradation)
All these newer consumer cpus are factory overclocked, unlocked and are allowed to run hotter to gain performance at high frequency to be stable and compete with other companies … Heat vs silicon degrades… nothing new not rocket science… the problem is cpu design and efficiency of programs… back to the drawing board!!! Because these systems can heat a room in the winter in a matter of minutes…Just need better design!!! Right now the cpu is like an incandescent light bulb that degrade overtime..
I am wondering about this problem. I am no expert, but the facts are, the CPUs use a lot of power, and then after a while they start crashing. I presume that something on the CPU is breaking. That sounds to me like the manufacturing process isnt creating a solid product. And it may well be that in combination with the huge amount of power going through them that could burn something out.
This coversation reminds me of an issue Greg Salazar was having on his 'fix or flop' series where a few of the same CPU was failing - I think it was the Ryzen 3600 chip, and I'm pretty sure the suspected issue was voltages being frequently ramped up to well over 1.4 by some motherboards when on PBO, no manual overclock. Not sure if he ever got to the bottom of the issue but I remember him saying he was going to flag it with AMD
This is why I run mine in a undervolted profile with a locked wattage with a +200mhz boost so if it doesn't mind running at a low power state and still wants to try to hit a high Ghz it will do it within the voltage and wattage I set, CPU seems to like it since Linux is reporting a 5.9ghz with a 7950X3D ..unless thats bugged, lol
After I built my 13700k I was pretty annoyed that a creator motherboard out of the box didn't have pl1/pl2 set to Intel spec. After setting that in the bios and undervolting I've been fine running the CPU on air.
They might have said those 70's processors were were good for a few years but given the number of ZX Spectrums with Z80s that are still merrily humming away they are probably good for 50 years or more.
So, did I get lucky? I use an i9 14900k with an MSI Z790 Edge and selected water-cooled in the BIOs overclock. My CPU doesn't go above 80c in HWiNFO. I use my system with a Gigabyte Aero RTX 4090 for Davinci Resolve Fusion/Blender and don't game. Originally, I had a Ryzen 9 7950X & ASUS X670E Pro WIFI and switched to the 14900k & MSI a month ago. Although the workload time was identical on both CPUs, when it came to rendering out my project in h.264 & h.265 it was taking between 30 min to hour with AMD, even with the 4090. Now it's less than 10 minutes. My system is stable now, but I don't want an issue in the future if I can dial back the voltages.
I think what causing people to talk about this isn't about the voltage degradation, but mostly temperature degradation because of the implementation of current CPUs to run at 95C / 100C most of the times without any thermal limiter / throttling
Well they're obsessed with the wrong thing, how many people have large workloads running all core all the time? A thermal limit is a way of limiting power and frequency. If you use a true temperature scale reflecting thermal energy in Kelvin you'll see the difference between 75C and 95C is actually about 6% Intel gaming laptops would constantly throttle reporting 100C and last for years so the change from power constraint to a temperature one really isn't large, it just confused people used to high end coolers giving lower temps rather than higher performance only. You can adjust settings in the BIOS anyway to limit power.
I always undervolt my cpu's and overclock them I've never had a problem. I'd never let the motherboard auto volt everything that's where the problems start!
a couple of years ago you only undervolted to run slower, cooler and maybe fanless. What you are doing is new to recent CPU designs. You are playing the silicon lottery in reverse so to speak. Getting as much efficiency as possible. But overclock and undervolt was basically impossible before Ryzen/Intel e- core.
@@KeithZimyou could do it - just not so "semi-automatic" as today. I ran my i7-8700k with a slight all-core overclock combined with a negative offset to the core-voltage. This resulted in a decent increase in multicore-performance without an increase in power-consumption. Was running this setup daily for ~5 years. Meanwhile have an AMD 7900X, which needed a different trick - I run it with the "120W" power-limits intended for the X3D CPUs (though it is a 170W non-X3D) which alone would lower the performance. But combined with a down-shift to the curve-optimizer and additional voltage headroom gained, by decreasing the SoC-Voltage (almost 10W) the 7900X is now running faster, than stock - in the 120W envelope and with good thermals (
Call me crazy, but I'm running my 7900x at 65Watts economy mode setting and 60 degrees C on the temp limit. 4.8 GHZ on all cores. LOL. Now that is effeciency. 10,000 on multi core cpu z. And 750 on single core. Love my new Rig.
@@ralphwarom2514 my mobile 13900hk* pulls 900 single core at 75w @ 1.24v, your letting your performance take a massive hit for probably no reason to be honest.
If you have an Intel 13 or 14th gen yes, lower the voltage and clocks if you want it to last. On 12th gen just keep an eye on the voltages, as the clocks themselves are often safe enough if it's not a KS anyways
Back in the day I had a i5-3570k all-core overclocked to 4.2ghz. Over time it gradually required more and more voltage to prevent crashing, until ~5 years later the system completely died. The only thing I didn't rule out was the motherboard. Good thing CO/undervolting is the way with Ryzen now 😁
I have ran my i78700k at 5.1GHZ ALL CORE on a simple 280m AIO for over 8 years now and the PC is on 24/7 running games or processing video stuff. NO LOSS OF LIFE AT ALL, BUUUUT my older systems from i7 3rd gen and 4th could not do this and after 2 years they shit the bed taking higher voltage over and over to maintain stability until death only the 3rd gen died but 4th got thrown out as it was doing same shit.... I have been running ryzen maxxed out and will soon find out if the chips last like intels newer gens or die like intels older ones....
Results vary, I have some that take no more voltage, and others that do. I have a 8086K build, that started at all core 5.0 undervolted, but now can only do 4.8, while never being allowed over 75C. I stopped overclocking customer builds after Athlon XP era because there's too much variability in binning and silicon quality, despite what the manufacturers will admit or show on paper.
8700k was the 2nd revision (I think) of Intel 14nm, so this is to be expected. Roclet lake chips are probably, being on the 3rd or 4th revison, even hardier.
If this was live I'd ask, is the expected lifespan assuming hours of use? What if you have an older or spare machine that's not turned on very often? I would expect it to last a long time until things like capacitors on the mobo die before the CPU ceases to function.
I degraded an i7 920 running it at 4Ghz continuously with water cooling. Over time it lost the ability to sustain that clockspeed and within a few months required excessive voltage to sustain its clock speeds. But that was a heft overclock with around 1.4V continuously and it lasted for 3 years before there was a problem. The chip still worked it required a bit more voltage at stock but it was degraded. I did a similar thing to the 3930k and it is still running happily with an overclock in a server without concern. Nowadays there isn't much to gain from overclocking since they super clock themselves so I don't bother anymore.
For my 13900k and 14900k I have done the following: MCE off, PL1 and PL2 limit to 225, limit P-core boost to 5.5 GHz and E-core boost to 4.3GHz, and use balanced power profile in Windows (although I do disable core parking to keep system highly responsive). Oh and just XMP on the RAM. I didn’t change LLC value. I have set voltage offset at a modest -0.010v. I have disabled the C6&C7 C states and EIST. Lastly I have locked AVX at 0 offset. I have tested on P95, CB R23 and CB R15. All great and in a mid 20 degree room, no workload exceeds 80c on package or cores. Very happy and benchmarks are very close to where they were before taming these beasts.
OK I have a solution that will see this platform out (of which I am invested as I have 3x 13/14th gen systems). So, due to the fact that Gigabyte hasn't been releasing Z690 updates and I am not convinced the microcode fix due next month will work for either my 13900k (x2) or my 14900k or that Gigabyte will release timely updates on their Z690 or Z790 boards, I have gone extreme. Initially I thought I would buy 3x 12900k and make peace with it. Then I thought, why not set everything to make my 13900k and 14900k effectively a 12900k? So I am now only running 8 e cores instead of 16, power limits, base and boost clocks are as per Intel specification for a 12900k. CPU-Z, 3DM CPU Profile and CB R23 show the system now performing at ever so slightly higher metrics than a 12900k (I guess due to process improvements and variation in systems). What is more amazing is that running the Cyberpunk 2077 and Forza Horizon 5 benchmarks at identical settings, yields a slightly HIGHER score! We are talking a single digit FPS improvement on CP and 20 FPS on Forza - even if that's margin of error, it's not slower. The system seems much more responsive in Windows and the games seem more smooth. Previously there was the odd stutter. Don't get me wrong, Intel are still on the hook in my eyes but I want to be productive and not worry about the systems until I decide to upgrade.
Sadly, no mention of undervolting with overclocking. I definitely understand that adding more voltage can result in higher frequency and higher wattage along with more heat. Those 3 things would lead to degradation. Also, i wish they would have talked about underclocking with unvolting and fixed performance mhz speed vs energy saver that fluctuates in mhz.
1.38V Sandy Bridge 1680 V2@4.4 all 8 cores running for years now, I do see a larger delta between die and cooler temp due to degradation of the die to lid contact., but this CPU is reaching the end of it's performance usefullness and will probably live in my NAS in the future clocked no more than 3.9 and undervolted. The only CPU that literally stopped working for me was an AMD A10, years ago I rebuilt/repaired a prebuilt my wife bought with a new MB and RAM, it ran for about 10 minutes, reached 100C installing windows and that was it, Error 00 on the MB.
Coincidentally yesterday my desktop got unstable again from degradation of the CPU's memory controller. It had slowly worsening stability issues from the moment I built it until I slowed the memory clock down from the XMP setting of 3200MHz to 3000MHz. But yesterday it started getting unstable again in the same way, crashing or locking up in memory operation heavy workloads. The problematic CPU is a Ryzen 3 3200G, so not much is lost if it does kick the bucket, hence I am leaving the memory clock as high as is stable, out of morbid curiosity to how long it last. My issues do track with the rumours I picked up of these early Ryzen APUs having a weak memory controller, though bad luck with the silicon lottery plays a part too as 3200MHz is considered slow for early Ryzen.
my almost 12 year old 3770k which is now in my secondary PC still runs like on day 1 @ 4.1 GHz with a slight undervolt and good air cooling. At 4.5 GHz with overvolting it would look different I guess.
"as long as you don't increase by 30%" ... in deed that is, what some manufacturers did on AMD 7xxx CPU with the uncore-voltage (memory-controller) when you enabled EXPO RAM. Just to get sure the RAM is stable they pushed the SoC up from 1.05V default to 1.3 or even 1.4V. Meanwhile it seem all to have settled at 1.3 as max (after esp. the X3D CPUs got roasted)... while in deed my DDR5-6000 CL36 RAM runs totally stable with
Yep. Small voltage increase can sometimes be an easy fix to gain stability. It's no problem the first time it's done, but another problem happens, and another. No one double checks are these voltages actually OK and BAM fried SOC. Can also be used to cover up poorly done(cheap) power supplies.
Most issues can be avoided by initially buying the product that delivers the performance needed/wanted running at stock settings and running the hardware at that point and not letting the MB set your voltages but set them manually to be within Intel specs. Do that and pretty much eliminate all problems being experienced by people running their chips at higher than specified settings either knowingly/unknowingly.
my i7 2600k @ 4.4 Ghz still kicking after 13-14 years although my voltage is around the 1.28 mark which is around 6.5% voltage increase from stock. My chip luckily don't require higher voltages and I've experienced Motherboard failures more than Cpu anyways. The point is take care of the system, don't push it too hard and try and get the best speeds & stability with as little voltage as possible & monitor temps every now and then. I'm sure I've overused my Cpu as well - 5000+ 4K Video renders & Crypto Mining for a Year straight as well as Unreal Engin Game Devs Compiling and such for 4-5 years not to mention around 6000 hours of gaming time on the system including emulation (RPCS3) which abuses the CPU (80% 100% utilization on 4.4Ghz OC) All that and I only paid around 125$ Back in 2013 for the whole system cause the business selling it didn't have a windows password for it so I naturally negotiated and to their ignorance they agreed. Best 125$ ever spent HAHAHA. Naturally upgraded the Gpu Ram and PSU Over time but still the same Platform (Now with RTX 3060 and 32 GB DDR3)
i7 2600k used 32nm lithography. Ryzen 7000 series uses 5nm. I predict more and more dead chips from small lithography mixed with thermal cycling and default 95C operation.
@@patsk8872 Indeed , hot chips + smaller lithography will have quite a Impact. I've already noticed a up tic on cpu failures more recently compared to previously however it's still not big enough for people to complain en masse. We are encroaching upon Heisenberg’s uncertainty principle & moors law will end. The time the failures will be more visible will be after a few years when software properly utilize high - end Hot chips like the 13900k 14900k & R9 7950x ect ... there is software now that utilize it as well but most people gaming and doing basic stuff are probably using 50/60% of the cores and threads so it's less utilized and basically running colder most of the time but games & software are catching up pretty quickly. Now recent chips with the constant 95C operation like you said will definitely also have a impact as they are boosting clocks unnecessarily as the moment you start utilizing the cpu it will drop clocks because the Voltage and Temp can't be sustained - the only benefit it has is making current single to quad core reliant software and games look good cause of a 10-20% fps increase.
Most of my computers are over 10 years old. I only ever over clocked one, a 486dx4 100 for a year, it locked up and I took it down to stock. Not worth the risk to stability.
I forget about this, and hear it again...man. I want know Dr. Cutress opinion right now. Remember me when I read Anandtech regarding AVX512. Dr. Cutress says it the great thing of the Intel processor. When Intel says that are not going to support in some consumer CPU's because is not relevant for final users. Reality hit hard.
I have a 5950x with no core OC and only a XMP profile enabled, and after 2 years I'm having to back off the RAM speed and very carefully manage the soc voltage (the difference between too low and too high is down to 0.0125v). This is the worst case of degradation I've personally experienced. I do worry about the fragility of these things as process nodes shrink.
There are already a lot of 13700-14900 processors cant run unreal 5 engine games due to using “too much power” and pn doping on silicon is broken over high voltage. It would cost a lot to find the problem and do rma over long periods ahead. Just limit them to 12th gen power levels and use longer
BINGO Well said. That;'s why I use AMD Ryzen 7800X3D. These CPUs are not so fragile and degrade so fast unlike Intel 13th and 14th Gen high end. The i5 and lower end degrade too but they have so much thermal and voltage headroom and require so much less for less cores and clocks that it is not noticed or an issue.
@@Wolverine607 i think intel put steroids on 13th and 14th gen highend cpus due to 5800x3d won against 12900k (no ram oc) i like my 7800x3d too for having 60w for most games
@@PatrickH-u8w Yes I agree. Intel was so desperate to beat or tie AMD they put these CPUs on steroids' and are going to have so many RMA claims in a couple of years and a lawsuit for degradation being so bad and chips being rendered so unstable right after warranty expires. Would not surprise me at all. Its pretty bad that a company as big as Intel cannot have better quality control and come up with a better solution. And yeah 7800X3D so power efficient and dumps so little heat inside the case. I mean it would be one thing if Intel CPUs did not degrade so easily and consumed max 150 watts instead of over 250 vs AMD 50 to 100. But 250+ watts and 150 to 200 watts in somewhat CPU intensive games is just too much power and thus heat generated in a room and your case its not reasonable. Unless you like all that heat in a frigid cold climate but yikes for spring summer and even fall and that's just part of it.
The thing with this whole drama about new Intels, Unreal 5 games and instability is due to people letting boards run them out of spec and exact culprit is here a current limit. Combination of unlocked power and current limits starts with instability and then degradation with living examples of CPUs being degraded after few months of such usage. EDIT: Stupid ideas in type of Multi Core Enhancement are also played here and for sure don't help.
my next build im running direct die water block on am5. i got gigabyte but im very interested to see just how much more boost it will give with only changing from air to water. from what i understand with asus - they were unlocking amperage which is a very easy way to blow any small traces in electronics. as far as ram goes i have found references of others running my kit at 1.75V but only some boards will let you go that high. strangely the higher density ram (single rank) seem to run easier, lower voltages, on the controllers. there are just so many things that have changed in modern day compared to 5 or 10 years ago.
I still run a 3950x machine and a fx9590 machine. End of year 2024 here. I will say the fx9590 machine makes my power bill go up so I don't turn it on much. I love redlining a machine all the time. However, the snappiest machine I run is not overclocked in anyway shape or from accept for I always water cool my computers. Every single computer I own and build for clients are watercooled.
My oldest cpu is an Intel I-9-10980xe it is still clocking 5.0ghz on all 18 cores so no degradation yet still same as when it was new with no voltage adjustment since I got it.
@@sawyiier4870 it’s custom liquid cooled and highly tuned, Under stress tests of many kinds, the cpu is anywhere from 60-75 c, power draw at 253-262 watts, as I said All core Oc at 5 ghz. Voltage varies 1.37-1.41… however I don’t run it all day everyday, it was first used in my studio, then it became a secondary/backup after two years of use , today it is my 4th system ,and barely used at all ,but as system was golden I refuse to get rid of it. It was my main video rig with 2/ Titan(rtx 2000 series/12gb vram).. Then switched to 2/Rtx-8000 Quadro cards(48 gb vram)..
In these days with the new chips its more recomended to uv and underclock your cpu a bit to get a stable perf or even better perf these new high end cpus are just really crazy with temps even with expensive aio.
My 4790k couldnt maintain 4.8 at 1.35v. Had to cut it down to 4.5 I want my new 7970x to last beyond a decade so im probably gonna keep it at a slight oc with pbo and leave it alone.
Some UA-cam tech channels are now calling out motherboard manufacturers whose firmware pumps crazy voltages to achieve higher clocks at stock config. The whole “factory” overclock situation is really nebulous though…does it void the warranty or not? I don’t think they want to give a straight answer.
The question I was waiting for was the effect of lifting clockspeeds but leaving Voltages intact. Does this also accelerate electromigration? Intuitively, I'd suspect it's the Voltage doing the damage, but also wonder if high increases to clock (while Voltage is kept the same) has negative lifespan effects.
I would say, yes indirectly raising clockspeed raises temperature which increases migration (regular thermal migration as opposed to electro-migration).
As an anecdote, I'm currently running a Ryzen 1700 (non X) bought in spring of 2017 that has been all core clocked higher than the single core boost (greater than 1ghz over stock - 24/7 - its locked in at those freqs - no boosting behaviour whatsoever), with a fairly significant over-volt for the entire duration. I do 3D rendering so I work it like a dog and it is still running great with no issues. Maybe they just don't make them like they used to?
I've killed cpus, gpus, had motherboards catch on fire while plugging in the power on the wrong socket, power supplys that pop like a gunshot, ram sticks that break while ramming into the motherboards. Just have fun man and don't worry to much about it. One advice is to have a dedicated gaming pc only to screw around with overclocking if your into that sort of thing.
It's an expected result of the undue popularisation of overclocking, I say "undue" because the modern CPUs are already very powerful and they do quite a good job of overclocking themselves within reasonable limits, and since hardware is reviewed by UA-camrs who routinely push everything to the limits, the more extreme hardware designs are always rewarded, and conservative/reliable designs are penalised. I essentially gave up on overclocking once I realised that I'm paying for imperceptible gains with heat, noise and instability, and it's completely not worth it.
In general I play around with staff to figure out what I can do before losing immediate stability (like, several hours of different stability tests) and end up with one or other taste of undervolting. My 3070 works at static 1800MHz, with VRAM and 8000, consuming 33% less energy than defaults at max 65*C for 0-5% performance cost. Even though I can get it to reliable 2010MHz, but it would add 20% power consumption (so, almost double from what I'm running) and will force me to replace thermal interface every 3-4 month. I might have been ok with water-cooling, but there is no water-blocks for 3070-s, so I can't do that.
Something that should be discussed: it’s not only degradation, but is an overvolted / overclocked CPU still providing valid & consistent outputs? We always have needed to deal with uArch errata, but recent events re: Raptor Lake and Raptor Lake refreshed may warrant a deeper chat. If decompression + integrity check crashes can be solved by dropping *below* stock multipliers, there’s something quite unusual here. Just as DRAM overlocking can lead to bit corruption (more than just DRAM degradation or failed boots / total instability).
The 300A was certainly a performer... I was Sr Tech at a local shop during those days and remember them well. The real clockers that few under the radar were the PDC e2100 chips... I was able to run one of those with a 100% overclock on stock voltage. No funky ram ratios, stock multiplier, just double base bus clock. Bam... 1800->3200MHz all day long on air. IMHO, Overclocking isn''t about the "fastest" speeds.... It's making the $100 chip run like the $500 one, just like the 300A did.
Straight up fraudulent data points, do not listen to a single word these scammers ever put out, how you actually have real jobs in the real world is the real question you should be asking yourselves.
YES! I worry about CPU degradation ALLL THE TIME! When our CPUs get to the point that they can verbally and physically degrade us, the human race is DOOMED!
I have a Core i7 9210 1st gen that's slow degraded. It just can't OC as high it used. If Linus Tech Tips or someone(not Gamers Nexus after his hit piece shenanigans) wants it for content, hit me up. Could be a good example of slow degradation.
Motherboards Manufacturers: Overclock your CPU by Default.
Also Motherboards Manufacturers: Fails to Detect and Automatically Activate XMP.
Buy an Intel motherboard and it will kill your CPU with overvoltage just for stock speed.
I think there are more manufacturers for RAM than CPU and hence more variation in quality control.
@@DipJyotiDeka barely.
Hynix samsung elpida and micron
it''s not the MB makers, it's happening on 13/14 gen servers which DO NOT over clock/voltage or anything else since reliability and stability are primary over speed. intel is flat out lying, there's something wrong in their raptor lake architecture. that they're trying to blame board partners and users for.
Wendell: I'm gonna slap this dude mraltoid a for blaming mobo manufacturers
I hope for a class action lawsuit against the motherboard manufacturers for over volting our cpus without our consent or knowledge and no easy or standard way to return the voltage back to intels specifications! This is disgraceful for all motherboard manufacturers to behave in such a disrespectful way towards their customers!
Good on ya Carey, I hope so too.
It's been about 20 years since I had multiple motherboard failures and now they want to destroy cpu's instead.
I have asus motherboard and it's super easy. Thers almost basic option in advance settings that you change from motherboard overdrive to Intel defaults. And its just 1 click. But ea it should be off by default setting no other way around
Until you researched the issue and realize that Intel have openly stated that having no voltage regulation is “in spec”.
Has these motherboard manufacturers been treating intel and amd differently? Any degradation complaints for amd?
10:50: "If you are not overclocking, do not worry about it." From what I have heard, even workstation and server boards with no overclocking features whatsoever have stability issues with 13th and 14th gen CPUs.
Don't get me wrong this is a massive issue but one of the most wide-spread pieces of misinformation is this. It comes from Wendell at level 1 Tech, but it is not true, these W series motherboards were also running at a 4095w PL1 and PL2 just like "Gaming" Z790 boards. The same settings as these boards but they do not have the option to change them, so Wendell assumed they were 125PL1/253PL2.
This issue is due to it's default VID table being too high mixed with a 1.1AC Loadline which causes overshoot to 1.5/1.6v, so it doesn't matter but this was misinformation being parroted. Pushing 415+ watts through a chip with no power limits does exacerbate the problem obviously. These servers are setup to run single-threaded workloads 24/7 (I.E Minecraft servers) which causes the 1.6v/6ghz boost to hit constantly leading to the higher failure rate than the average person running these out of box settings.
Massive issue and Intel is at fault, but we have to be true to ourselves and once we are aware this is wrong we need to stop parroting it.
well, this interview aged badly.
Can you explain why?
@@besim109AOG Because we now know that even conservatively clocked professional grade boards are seeing degradation. That implies that it's a fundamental problem with the chips, not (just) what board makers where doing.
Not to mention, we already knew that at least some of the official specs from Intel say what at least some of the motherboard manufacturers where doing is in spec.
@@ahettinger525 Thank you
I’m only 2/3 of the way through the video and that was exactly what I was going to say. The fact that both the mobos are pushing these cpus past their specs, what the cpus are being advertised as being capable of is actually redlining them and if brought down to intel specs you don’t get speeds advertised, and lastly and most importantly, 50% or more new intel chips seem to be degrading with no overclocks whatsoever while being used for servers which is just damning. This guy is implying that the optimizations the mobo is applying won’t significantly cut the lifespan of your cpu which is not the case. They’re speeding up the degredation.
@@ahettinger525You understand motherboard has nothing to do with Intel's recommended voltage right? People talking about voltage doesnt necessarily mean they are talking about motherboards. You know that ye?
As a person who had an i7 920 from 2009 to 2016, it feels like the instruction set and cache being left behind is a bigger issue for me long term.
I had the same CPU as a daily from 2010 to Jan 2020, OC to 3.3, (just like my slot 1 P2-400) its now in the loft. Then upgraded to 3950X, will I get 10 years (hope so). BTW those CPU still work.
I was running an i7 920 from 2009 to about 2014, then upgraded to an i7 980X which I replaced just a month ago. During this time I had to replace 3 DIMMs, but both CPUs were (and still are) ok, even though I did overclock both and kept them running with close to 100% load and 24/7 for most of the time. The second and third DIMM I did buy used though, so there's that too. I am guessing the DIMMs themselves did hold up for about 10 years.
CPUs were water-cooled though, so never really saw more than 75°C. The 980X was running with all 6 DIMM slots equipped for many years (as you may know the memory controller of the 980X was said to be a weak spot and putting full load on all channels on it was going to be more risky).
10:52 Ouch! Intel users down bad.
*_Battlefield 4_* was my go-to stress for several years on gaming PCs. It had ways of finding overall system instability that individual component-specific stress tests frequently seemed miss.
Battlefield 2042 is good for CPU. I have recently diagnosed a problem by running it while running XTU in the background and looking at telemetry
Funny because Battlefield 2042 is now the best game to test RAM stability, even catching errors that memtest programs don't see. The Frostbite engine uses your PC to the fullest 😄
I have seen a bunch of CPU's with 40 years working flawlessly, even today.
Old RAM chips, specially from certain vendors, is another story, some last ages and are still reliable today, others degraded after a few decades.
But CPU's, I never saw one just naturally degrade or stop working after a few decades, not with normal default use.
So this conversation sounds a tad off to me, claiming that CPU's were made to last 3 years (or the guarantee period) is very strange, maybe today but that was certainly not the case in the 80's to the early 2000's, no one would buy a Intel or AMD CPU, if they only lasted 3 years.
And big tech companies, will never accept "silicon degradation is natural" when buying huge amounts of CPU's for their computers or servers.
Intel and AMD may wake up one day and see ARM occupying their place, if x86/64 CPU's start dying earlier than users expect or find normal, as being the case with certain Intel CPU's recently.
What he said about smaller parts does have some merit. Remember, many of these older CPUs were on a different process node. Intel for the longest time were on 14nm and finally got to 10nm. As foundries make parts ever so smaller, you have to start to wonder just how small is too small? This is one of the reasons why TSMC has been using machine learning to run these simulations not to just find the best way to tap out a design but to research on all the ways a design can fail. This is one of the biggest reason why Synopsis bought cadence because the tap out process is so complex and expensive, designers and manufacturers must work hand in hand to iterate their design to find one that will satisfy design requirements.
My old i7 4790k is still alive after 8 years heavy overclocking.
5GHz all cores since release 2014
I had mine overclocked at 4.6Ghz since new in 2014, i used 1.29V, mine died Q3'23. loved that CPU.
I had a 4770k, @4.6ghz 1.29v with 2133mhz ram. it then began to require more voltage to keep that overclock up until 1.318v, where I had to tune it down to 4.5ghz.
It still required 1.31v to operate at 4.5, so still more than the OG 1.29 at 4.6.
It never died, I sold it later in 2020 so I don't know if it had degraded further or what.
I had it for 10 years and changed to 14900k set up. And now probably it will crush all the time
@@JMatrx mine just simply died, it was working 100% one day, the next day the PC did not boot, I tried the CPU in a different machine and it would not boot also, I placed a different CPU in the same MB and it worked fine.
@@dauletbaimagamabet
probably crush all the time? lol
@10:40 and there it is folks: "somebody's done a major screw up if it is" in regards to FACTORY SETTINGS killing CPU's. Looking at you, Intel.
Still running a Haswell Devil's Canyon!
My motto is to buy robust/stable enough stock hardware for the resolution that I'm aiming to play at (high/ultra quality 1440p in my case), so that I don't need to overclock anything & add stress/instability to the system. I see overclocking as more of a game onto itself for the bench-marking e-sport crowd.
Oh my, how the turn has tabled. 🙃
Voltage increase can cause electrons to jump through the insulation layer, and create a dead short.
I've had at least 2 components where i'm pretty sure i've experienced degradation, one CPU and one GPU. Both were ran with a custom OC including increased voltage, and had to be clocked down and/or voltage increased to keep stable after some years. So, i would actually recommend a lot stronger to worry about this. As soon as you touch voltages, always keep this in mind and think about if 1% more performance really is worth increasing the voltage by 5% or 10% with voltage having an exponential effect on longevity.
I've degraded one CPU, a 6800K. I was pushing it to 4.3GHz and Broadwell-E was terrible at OC. It took a lot to get that stable, and it was stable and I was running it like that daily. After about a year, it started to become unstable and I had to walk the OC back to 4.2GHz to be stable again. Eventually I just decided to run it at stock rather than risk further degradation. With the new CPU's, I'm less worried about stock voltage degrading the CPU than temps. Perhaps I'm not understanding how this impacts CPU's, but you can degrade memory with prolonged high temps, a serious concern when buying a used mining GPU. And this isn't an issue for most people, but if you're doing a lot of work that has the CPU running high usage in an all core workload, pulling down 250+ watts, and the temp is in the 90's. I feel like that's not a good situation if you're doing that on a regular basis. Seems like degradation would be a concern for someone like that. There have been a lot of Ryzen 5 3600's that bit the dust prematurely, and one explanation was the inadequate CPU coolers that were included could be causing the CPU to run too hot for extended periods of time.
So, for me, I'm more concerned about the temps and I limit the PL1 and PL2 to keep the temps more reasonable in those workloads and shouldn't really impact single core performance.
My 4960k been over locked from day 1. It runs at 4.6 for my nephew till this day emulated switch fine at a 1100 geek bench 5.
Now I had some overlooked gpus go to poo.
@@delresearch5416 I mean, Devil's Canyon was a pretty good overclocker. Regardless, if the voltages and temps are kept in check, I'm not surprised it's still going strong. It's going to be interesting to see how these new CPU's hold over time.
If you have an Coffee Lake R desktop K chip or newer from intel, never let the motherboard control what's your voltage/frequency, always follow intel specification.
you'll have lower temps, quieter fan and if you don't have a really good custom water loop, you'll have better performance as well by going stock.
Extra points for slightly undervolting your cpu.
fun fact 13th gen can be slightly undervolted and work just fine, and considerably cooler.
@@waynemacleod3416 that's basically every cpu
This is the way
Would love to see motherboard manufacturers have an "AI" undervolting toggle where they continuously try small increments of undervolting until it becomes unstable and automatically sets it for you. Wouldn't that be way more advantageous to the average consumer who doesn't know how to adjuat settings, rather than cranking all their limits to the max out of the box that their cooler can't handle anywyas?
Probably but they’re not doing it because it’s best for consumers. They’re doing it so inexperienced reviewers (who often are the ones who will review a lot of motherboards) will run it against several and the one that blasts the power limit will perform better. Less true with the recent i9s but historically and with other SKUs you get an extra 100mhz and the reviewer concludes “asus has the best performance”.
It's not quite as sofisticated but AMD already has a feature to automatically find stable undervolt/negative offset curve optimizer settings in the ryzen master application now. It's not perfect but it makes it very easy if you want to quickly undervolt the cpu without spending a lot of time manually testing, just use that feature and set the values a tad higher.
Doing a completely auto undervolt by default would probably still be too problematic stability wise and also require the board to take a lot of time stability testing (which is already a thing with memry training, especially on the threadripper workstation boards)
@@olnnnthey also got that in the adrenaline software for the GPU. Won't make a hughe difference, but i encourage everyone to do it. Literally 0 risk of damage, same stability as base voltages since it is quite conservative, and you save 1-4% power input for literally free for the exact same performance. (And in summer that also means a couple watts less heat, which is nice)
You don't need AI fuckin' garbage for that. It's no different than having a basic coded sequence to test stability the same way MSI Afterburner does it.
Similar procedures are done through system controllers with DDR5 RAM training.
I have systems from the 80s and no CPU has ever died.
a lot simpler and hence more robust back then.
And not overclocked either …
I still have a fx8320 system running in the house. Still running strong on a 4.4ghz oc. But I steer clear of crazy oc'ing. And I'm obsessed with low temperatures.
lol now I know why this channel is dead compared to my reputable outlets. It’s full of Intel fanboys blaming motherboard manufacturers when Intel encouraged this by openly stating to the press that having no voltage regulation would still be “in spec”.
Most modern CPU/GPU have aging mitigator which slowly increase voltage to keep stability.
I have a 4790K that I've been running stock in one way or another for the past 10 years, first in my main rig, and later in my HTPC. It has never been overclocked and has only been run for a few months with multi core enhancement on (default option on Z87 boards, against Intel spec). During the past few years, it has simply stopped being able to boost to the highest stock frequency (resulting in immediate BSOD/hard lock). This could, of course, be a motherboard issue, but my money's on the CPU. If It is the only case where I have seen such obvious degradation.
Someone in another comment mentioned the i7 920 - I also used this CPU for around 7-8 years continuously, and with a slight OC at stock voltage on a X58 board. Never noticed any instability issues, until the board just up and died one day.
Depends on temperature. My Phenom II x6 lasted 11 years @ 4ghz OC @ 200W. The only reason it was viable for so long was an extremely aggressive 777-22 memory timing.
"I'm not dead yet." ... "You're not fooling anyone, you're .3GHz slower. Do me a favour?"
I’ve had a hell of a time trying to get my 12900k to run stable. The only thing left to replace is the cpu. Are there guides to help you know what to turn off in the bios to have a chip run stock? Is it possible for the consumer part of the industry to insist they add an “intel stock button” in the bios so that us dummies out there have an easy option to run tests at stock? At least have that as an option over the default no limits configs?
Just a thought
Cheers
Rick
The best people to ask such questions is your motherboard's support.
the only degradation ive experienced was using liquid metal on direct die
My 2500k was 8 years on 4.9ghz 1.425v and still works on the same voltages as the 1st day. All my CPUs are maxed on with everything including BCLK. So far so good.
You're lucky. I never did any OC on my 2500k, and my CPU + motherboard died after 3.5 years. Get them replaced using the warranty still works while OC without additional voltage .
My R5 1600x showed IMC degradation after 5/6 years of usage. Again, everything was at default settings except for memory frequency and and latency that I had to setup manually because I bought overpriced fast memory with XMP profile for frequency that my CPU/motherboard could not reach. I tightened latency as much as I could but with default "auto" voltage settings. After 5 or 6 years, I had crashes and had to slow down a bit the DDR frequency.
Dammit just saw this. MORE IAN PLEASE!!
*big fan of his works in Anandtech btw
I had a 9900K that held at 5.2Ghz for years, I got it the day it was released in 2018, around october of 2023 it started BSOD'ing and when I limited it to 4.8Ghz it was stable. It is running fine in my 14 year olds PC now and she doesn't even notice or care that it doesn't boost to 5Ghz anymore. I upgraded to a 14900K
Just learned the 13th and 14th gen i9 have a 100% failure rate.
Nice twisted damage control video. FACT is that the i7-s and i9-s are affected by it and probably potentially around 90 % of the CPU-s sold. The problem is that not everyone knows it and not everyone is going to link their BSOD-s to the CPU melting and not everyone is using it to play games constantly. If you take that in account, probably around 50 % of the examples are going to fail eventually and very soon.
I never overclock until ive had the cpu for a good number of years. My thinking is overclocking is not something you do until you need that extra performance usually near the end of the CPU lifecycle.
And by the time you need extra performance no overclocking will be enough...
100% Intel failure rate. Should you worry about that? Of course not. I'm using AMD CPU....
A good analogy I thought of regarding degredation would be like tires on your car, if you just do normal driving the tread will last for the warranty period, but if you are always pushing it by doing burnouts everytime you stop and start the tires will go bald more quickly (degradation)
Yes but CPU's should last way way more than tires, even in heavy workloads and specially if you keep temps in check. Not just "naturally degrade".
This didn't age well 🤣🤣🤣
my 2500k lasted me 10 years, while OC'ed by 1 whole GHz, it's technically still good today(3 years later), i just don't use it anymore, bc i upgraded.
All these newer consumer cpus are factory overclocked, unlocked and are allowed to run hotter to gain performance at high frequency to be stable and compete with other companies … Heat vs silicon degrades… nothing new not rocket science… the problem is cpu design and efficiency of programs… back to the drawing board!!! Because these systems can heat a room in the winter in a matter of minutes…Just need better design!!! Right now the cpu is like an incandescent light bulb that degrade overtime..
This is a problem unique to Intel 13th and 14th gen chips. There's something wrong with them. AMD chips are fine and are happily burning along.
I am wondering about this problem. I am no expert, but the facts are, the CPUs use a lot of power, and then after a while they start crashing.
I presume that something on the CPU is breaking. That sounds to me like the manufacturing process isnt creating a solid product. And it may well be that in combination with the huge amount of power going through them that could burn something out.
Excellent video - thank you for covering this with an expert!
This coversation reminds me of an issue Greg Salazar was having on his 'fix or flop' series where a few of the same CPU was failing - I think it was the Ryzen 3600 chip, and I'm pretty sure the suspected issue was voltages being frequently ramped up to well over 1.4 by some motherboards when on PBO, no manual overclock. Not sure if he ever got to the bottom of the issue but I remember him saying he was going to flag it with AMD
I was always affraid of stress testing my GPU and now i found out why.. i always test my rig in my work programs and the games i play.
This is why I run mine in a undervolted profile with a locked wattage with a +200mhz boost so if it doesn't mind running at a low power state and still wants to try to hit a high Ghz it will do it within the voltage and wattage I set, CPU seems to like it since Linux is reporting a 5.9ghz with a 7950X3D ..unless thats bugged, lol
This is really good and more stable.
I’ve got CPU’s that are 25 odd years old that still work perfectly well…
After I built my 13700k I was pretty annoyed that a creator motherboard out of the box didn't have pl1/pl2 set to Intel spec. After setting that in the bios and undervolting I've been fine running the CPU on air.
They might have said those 70's processors were were good for a few years but given the number of ZX Spectrums with Z80s that are still merrily humming away they are probably good for 50 years or more.
So, did I get lucky? I use an i9 14900k with an MSI Z790 Edge and selected water-cooled in the BIOs overclock. My CPU doesn't go above 80c in HWiNFO. I use my system with a Gigabyte Aero RTX 4090 for Davinci Resolve Fusion/Blender and don't game. Originally, I had a Ryzen 9 7950X & ASUS X670E Pro WIFI and switched to the 14900k & MSI a month ago. Although the workload time was identical on both CPUs, when it came to rendering out my project in h.264 & h.265 it was taking between 30 min to hour with AMD, even with the 4090. Now it's less than 10 minutes. My system is stable now, but I don't want an issue in the future if I can dial back the voltages.
I think what causing people to talk about this isn't about the voltage degradation, but mostly temperature degradation because of the implementation of current CPUs to run at 95C / 100C most of the times without any thermal limiter / throttling
95 c is too high.
I set the temp limit to 60 c.
Granted I recommend 70 to 75.
Well they're obsessed with the wrong thing, how many people have large workloads running all core all the time? A thermal limit is a way of limiting power and frequency.
If you use a true temperature scale reflecting thermal energy in Kelvin you'll see the difference between 75C and 95C is actually about 6%
Intel gaming laptops would constantly throttle reporting 100C and last for years so the change from power constraint to a temperature one really isn't large, it just confused people used to high end coolers giving lower temps rather than higher performance only.
You can adjust settings in the BIOS anyway to limit power.
I always undervolt my cpu's and overclock them I've never had a problem. I'd never let the motherboard auto volt everything that's where the problems start!
a couple of years ago you only undervolted to run slower, cooler and maybe fanless. What you are doing is new to recent CPU designs. You are playing the silicon lottery in reverse so to speak. Getting as much efficiency as possible. But overclock and undervolt was basically impossible before Ryzen/Intel e- core.
@KeithZim Not technically correct, but point taken. I've written guides on undervolt OCs since Sandy Bridge.
@@KeithZimyou could do it - just not so "semi-automatic" as today. I ran my i7-8700k with a slight all-core overclock combined with a negative offset to the core-voltage. This resulted in a decent increase in multicore-performance without an increase in power-consumption. Was running this setup daily for ~5 years.
Meanwhile have an AMD 7900X, which needed a different trick - I run it with the "120W" power-limits intended for the X3D CPUs (though it is a 170W non-X3D) which alone would lower the performance. But combined with a down-shift to the curve-optimizer and additional voltage headroom gained, by decreasing the SoC-Voltage (almost 10W) the 7900X is now running faster, than stock - in the 120W envelope and with good thermals (
Call me crazy, but I'm running my 7900x at 65Watts economy mode setting and 60 degrees C on the temp limit.
4.8 GHZ on all cores. LOL. Now that is effeciency. 10,000 on multi core cpu z. And 750 on single core.
Love my new Rig.
@@ralphwarom2514 my mobile 13900hk* pulls 900 single core at 75w @ 1.24v, your letting your performance take a massive hit for probably no reason to be honest.
If you have an Intel 13 or 14th gen yes, lower the voltage and clocks if you want it to last. On 12th gen just keep an eye on the voltages, as the clocks themselves are often safe enough if it's not a KS anyways
Back in the day I had a i5-3570k all-core overclocked to 4.2ghz. Over time it gradually required more and more voltage to prevent crashing, until ~5 years later the system completely died. The only thing I didn't rule out was the motherboard. Good thing CO/undervolting is the way with Ryzen now 😁
Even with older chips uv can give you really good results.
I have ran my i78700k at 5.1GHZ ALL CORE on a simple 280m AIO for over 8 years now and the PC is on 24/7 running games or processing video stuff. NO LOSS OF LIFE AT ALL, BUUUUT my older systems from i7 3rd gen and 4th could not do this and after 2 years they shit the bed taking higher voltage over and over to maintain stability until death only the 3rd gen died but 4th got thrown out as it was doing same shit.... I have been running ryzen maxxed out and will soon find out if the chips last like intels newer gens or die like intels older ones....
Results vary, I have some that take no more voltage, and others that do. I have a 8086K build, that started at all core 5.0 undervolted, but now can only do 4.8, while never being allowed over 75C. I stopped overclocking customer builds after Athlon XP era because there's too much variability in binning and silicon quality, despite what the manufacturers will admit or show on paper.
8700k was the 2nd revision (I think) of Intel 14nm, so this is to be expected. Roclet lake chips are probably, being on the 3rd or 4th revison, even hardier.
I've been hearing about this I believe with the newer platforms. Am on Alder Lake(s) myself and reading about potential degradation.
If this was live I'd ask, is the expected lifespan assuming hours of use? What if you have an older or spare machine that's not turned on very often? I would expect it to last a long time until things like capacitors on the mobo die before the CPU ceases to function.
I degraded an i7 920 running it at 4Ghz continuously with water cooling. Over time it lost the ability to sustain that clockspeed and within a few months required excessive voltage to sustain its clock speeds. But that was a heft overclock with around 1.4V continuously and it lasted for 3 years before there was a problem. The chip still worked it required a bit more voltage at stock but it was degraded. I did a similar thing to the 3930k and it is still running happily with an overclock in a server without concern. Nowadays there isn't much to gain from overclocking since they super clock themselves so I don't bother anymore.
For my 13900k and 14900k I have done the following: MCE off, PL1 and PL2 limit to 225, limit P-core boost to 5.5 GHz and E-core boost to 4.3GHz, and use balanced power profile in Windows (although I do disable core parking to keep system highly responsive). Oh and just XMP on the RAM. I didn’t change LLC value. I have set voltage offset at a modest -0.010v. I have disabled the C6&C7 C states and EIST. Lastly I have locked AVX at 0 offset. I have tested on P95, CB R23 and CB R15. All great and in a mid 20 degree room, no workload exceeds 80c on package or cores. Very happy and benchmarks are very close to where they were before taming these beasts.
100% failure rate....
@@AthosRac ikr
OK I have a solution that will see this platform out (of which I am invested as I have 3x 13/14th gen systems). So, due to the fact that Gigabyte hasn't been releasing Z690 updates and I am not convinced the microcode fix due next month will work for either my 13900k (x2) or my 14900k or that Gigabyte will release timely updates on their Z690 or Z790 boards, I have gone extreme.
Initially I thought I would buy 3x 12900k and make peace with it. Then I thought, why not set everything to make my 13900k and 14900k effectively a 12900k? So I am now only running 8 e cores instead of 16, power limits, base and boost clocks are as per Intel specification for a 12900k.
CPU-Z, 3DM CPU Profile and CB R23 show the system now performing at ever so slightly higher metrics than a 12900k (I guess due to process improvements and variation in systems). What is more amazing is that running the Cyberpunk 2077 and Forza Horizon 5 benchmarks at identical settings, yields a slightly HIGHER score! We are talking a single digit FPS improvement on CP and 20 FPS on Forza - even if that's margin of error, it's not slower.
The system seems much more responsive in Windows and the games seem more smooth. Previously there was the odd stutter. Don't get me wrong, Intel are still on the hook in my eyes but I want to be productive and not worry about the systems until I decide to upgrade.
Minor degradation is not a bad thing but look at the motherboards that overvolted and killed ryzen 7000 CPUs. Lawsuit.
Sadly, no mention of undervolting with overclocking. I definitely understand that adding more voltage can result in higher frequency and higher wattage along with more heat. Those 3 things would lead to degradation. Also, i wish they would have talked about underclocking with unvolting and fixed performance mhz speed vs energy saver that fluctuates in mhz.
1.38V Sandy Bridge 1680 V2@4.4 all 8 cores running for years now, I do see a larger delta between die and cooler temp due to degradation of the die to lid contact., but this CPU is reaching the end of it's performance usefullness and will probably live in my NAS in the future clocked no more than 3.9 and undervolted. The only CPU that literally stopped working for me was an AMD A10, years ago I rebuilt/repaired a prebuilt my wife bought with a new MB and RAM, it ran for about 10 minutes, reached 100C installing windows and that was it, Error 00 on the MB.
Still running a 15 year old build, i7 860... Going stronger than ever. No reason to stop using it and will continue to do so for a few more years
Never encountered that problem in 40 years. Never overclocked anything since then either.
I'm hearing you in the voice of Miracle Max from The Princess Bride, "He's only MOSTLY dead!" 😂
Coincidentally yesterday my desktop got unstable again from degradation of the CPU's memory controller.
It had slowly worsening stability issues from the moment I built it until I slowed the memory clock down from the XMP setting of 3200MHz to 3000MHz.
But yesterday it started getting unstable again in the same way, crashing or locking up in memory operation heavy workloads.
The problematic CPU is a Ryzen 3 3200G, so not much is lost if it does kick the bucket, hence I am leaving the memory clock as high as is stable, out of morbid curiosity to how long it last.
My issues do track with the rumours I picked up of these early Ryzen APUs having a weak memory controller, though bad luck with the silicon lottery plays a part too as 3200MHz is considered slow for early Ryzen.
my almost 12 year old 3770k which is now in my secondary PC still runs like on day 1 @ 4.1 GHz with a slight undervolt and good air cooling. At 4.5 GHz with overvolting it would look different I guess.
"as long as you don't increase by 30%" ... in deed that is, what some manufacturers did on AMD 7xxx CPU with the uncore-voltage (memory-controller) when you enabled EXPO RAM.
Just to get sure the RAM is stable they pushed the SoC up from 1.05V default to 1.3 or even 1.4V.
Meanwhile it seem all to have settled at 1.3 as max (after esp. the X3D CPUs got roasted)... while in deed my DDR5-6000 CL36 RAM runs totally stable with
Yep. Small voltage increase can sometimes be an easy fix to gain stability. It's no problem the first time it's done, but another problem happens, and another. No one double checks are these voltages actually OK and BAM fried SOC.
Can also be used to cover up poorly done(cheap) power supplies.
My 14900k can't even do r23 at default settings lol rma time
Most issues can be avoided by initially buying the product that delivers the performance needed/wanted running at stock settings and running the hardware at that point and not letting the MB set your voltages but set them manually to be within Intel specs.
Do that and pretty much eliminate all problems being experienced by people running their chips at higher than specified settings either knowingly/unknowingly.
my i7 2600k @ 4.4 Ghz still kicking after 13-14 years although my voltage is around the 1.28 mark which is around 6.5% voltage increase from stock. My chip luckily don't require higher voltages and I've experienced Motherboard failures more than Cpu anyways. The point is take care of the system, don't push it too hard and try and get the best speeds & stability with as little voltage as possible & monitor temps every now and then. I'm sure I've overused my Cpu as well - 5000+ 4K Video renders & Crypto Mining for a Year straight as well as Unreal Engin Game Devs Compiling and such for 4-5 years not to mention around 6000 hours of gaming time on the system including emulation (RPCS3) which abuses the CPU (80% 100% utilization on 4.4Ghz OC) All that and I only paid around 125$ Back in 2013 for the whole system cause the business selling it didn't have a windows password for it so I naturally negotiated and to their ignorance they agreed. Best 125$ ever spent HAHAHA. Naturally upgraded the Gpu Ram and PSU Over time but still the same Platform (Now with RTX 3060 and 32 GB DDR3)
i7 2600k used 32nm lithography. Ryzen 7000 series uses 5nm. I predict more and more dead chips from small lithography mixed with thermal cycling and default 95C operation.
@@patsk8872 Indeed , hot chips + smaller lithography will have quite a Impact. I've already noticed a up tic on cpu failures more recently compared to previously however it's still not big enough for people to complain en masse. We are encroaching upon Heisenberg’s uncertainty principle & moors law will end. The time the failures will be more visible will be after a few years when software properly utilize high - end Hot chips like the 13900k 14900k & R9 7950x ect ... there is software now that utilize it as well but most people gaming and doing basic stuff are probably using 50/60% of the cores and threads so it's less utilized and basically running colder most of the time but games & software are catching up pretty quickly. Now recent chips with the constant 95C operation like you said will definitely also have a impact as they are boosting clocks unnecessarily as the moment you start utilizing the cpu it will drop clocks because the Voltage and Temp can't be sustained - the only benefit it has is making current single to quad core reliant software and games look good cause of a 10-20% fps increase.
Intel has Boeing
Most of my computers are over 10 years old. I only ever over clocked one, a 486dx4 100 for a year, it locked up and I took it down to stock. Not worth the risk to stability.
I dont think it really matters - most will upgrade before any failures occur ...
I forget about this, and hear it again...man.
I want know Dr. Cutress opinion right now.
Remember me when I read Anandtech regarding AVX512. Dr. Cutress says it the great thing of the Intel processor. When Intel says that are not going to support in some consumer CPU's because is not relevant for final users.
Reality hit hard.
I have a 5950x with no core OC and only a XMP profile enabled, and after 2 years I'm having to back off the RAM speed and very carefully manage the soc voltage (the difference between too low and too high is down to 0.0125v).
This is the worst case of degradation I've personally experienced. I do worry about the fragility of these things as process nodes shrink.
There are already a lot of 13700-14900 processors cant run unreal 5 engine games due to using “too much power” and pn doping on silicon is broken over high voltage. It would cost a lot to find the problem and do rma over long periods ahead. Just limit them to 12th gen power levels and use longer
BINGO Well said. That;'s why I use AMD Ryzen 7800X3D. These CPUs are not so fragile and degrade so fast unlike Intel 13th and 14th Gen high end. The i5 and lower end degrade too but they have so much thermal and voltage headroom and require so much less for less cores and clocks that it is not noticed or an issue.
@@Wolverine607 i think intel put steroids on 13th and 14th gen highend cpus due to 5800x3d won against 12900k (no ram oc) i like my 7800x3d too for having 60w for most games
@@PatrickH-u8w Yes I agree. Intel was so desperate to beat or tie AMD they put these CPUs on steroids' and are going to have so many RMA claims in a couple of years and a lawsuit for degradation being so bad and chips being rendered so unstable right after warranty expires. Would not surprise me at all. Its pretty bad that a company as big as Intel cannot have better quality control and come up with a better solution. And yeah 7800X3D so power efficient and dumps so little heat inside the case. I mean it would be one thing if Intel CPUs did not degrade so easily and consumed max 150 watts instead of over 250 vs AMD 50 to 100. But 250+ watts and 150 to 200 watts in somewhat CPU intensive games is just too much power and thus heat generated in a room and your case its not reasonable. Unless you like all that heat in a frigid cold climate but yikes for spring summer and even fall and that's just part of it.
The thing with this whole drama about new Intels, Unreal 5 games and instability is due to people letting boards run them out of spec and exact culprit is here a current limit. Combination of unlocked power and current limits starts with instability and then degradation with living examples of CPUs being degraded after few months of such usage. EDIT: Stupid ideas in type of Multi Core Enhancement are also played here and for sure don't help.
@@2MichalChojnowski2 last year am5 socket burn is fast burn, this intel ue5 issue is slow burn lol
my next build im running direct die water block on am5. i got gigabyte but im very interested to see just how much more boost it will give with only changing from air to water. from what i understand with asus - they were unlocking amperage which is a very easy way to blow any small traces in electronics. as far as ram goes i have found references of others running my kit at 1.75V but only some boards will let you go that high. strangely the higher density ram (single rank) seem to run easier, lower voltages, on the controllers. there are just so many things that have changed in modern day compared to 5 or 10 years ago.
If you overclock yes on air but liquid you still get a barrier
I still run a 3950x machine and a fx9590 machine. End of year 2024 here. I will say the fx9590 machine makes my power bill go up so I don't turn it on much. I love redlining a machine all the time. However, the snappiest machine I run is not overclocked in anyway shape or from accept for I always water cool my computers. Every single computer I own and build for clients are watercooled.
My oldest cpu is an Intel I-9-10980xe it is still clocking 5.0ghz on all 18 cores so no degradation yet still same as when it was new with no voltage adjustment since I got it.
Whats your temps and power usage this can tell why
@@sawyiier4870 custom liquid cooling cpu temp 65 to75 c max under stress load ,power is in the 250 watt range at last check
@@sawyiier4870 it’s custom liquid cooled and highly tuned, Under stress tests of many kinds, the cpu is anywhere from 60-75 c, power draw at 253-262 watts, as I said All core Oc at 5 ghz. Voltage varies 1.37-1.41… however I don’t run it all day everyday, it was first used in my studio, then it became a secondary/backup after two years of use , today it is my 4th system ,and barely used at all ,but as system was golden I refuse to get rid of it. It was my main video rig with 2/ Titan(rtx 2000 series/12gb vram).. Then switched to 2/Rtx-8000 Quadro cards(48 gb vram)..
In these days with the new chips its more recomended to uv and underclock your cpu a bit to get a stable perf or even better perf these new high end cpus are just really crazy with temps even with expensive aio.
My 4790k couldnt maintain 4.8 at 1.35v. Had to cut it down to 4.5
I want my new 7970x to last beyond a decade so im probably gonna keep it at a slight oc with pbo and leave it alone.
Miss the good old Intel 875 PBZ mainboard for Pentium 4 Prozessor. Absolute top mainboard 👍
Some UA-cam tech channels are now calling out motherboard manufacturers whose firmware pumps crazy voltages to achieve higher clocks at stock config. The whole “factory” overclock situation is really nebulous though…does it void the warranty or not? I don’t think they want to give a straight answer.
THIS is what makes me sad about the competition they after speed over the stability!
My i7 5775c BROADWELL is still kicking. Overclocked . Almost ten years in 👍
i wanted one of those sooooo bad, but i just wasn't confident how well it would work with DDR3 (IIRC) which i had a lot of
The question I was waiting for was the effect of lifting clockspeeds but leaving Voltages intact. Does this also accelerate electromigration? Intuitively, I'd suspect it's the Voltage doing the damage, but also wonder if high increases to clock (while Voltage is kept the same) has negative lifespan effects.
I would say, yes indirectly raising clockspeed raises temperature which increases migration (regular thermal migration as opposed to electro-migration).
As an anecdote, I'm currently running a Ryzen 1700 (non X) bought in spring of 2017 that has been all core clocked higher than the single core boost (greater than 1ghz over stock - 24/7 - its locked in at those freqs - no boosting behaviour whatsoever), with a fairly significant over-volt for the entire duration. I do 3D rendering so I work it like a dog and it is still running great with no issues. Maybe they just don't make them like they used to?
only if you're running Raptorlake at stock settings and well cooled.
I've killed cpus, gpus, had motherboards catch on fire while plugging in the power on the wrong socket, power supplys that pop like a gunshot, ram sticks that break while ramming into the motherboards. Just have fun man and don't worry to much about it. One advice is to have a dedicated gaming pc only to screw around with overclocking if your into that sort of thing.
oh man by experience, you have to remove that ln2 jumper on the board and mess with the settings bad to degrade a cpu
It's an expected result of the undue popularisation of overclocking, I say "undue" because the modern CPUs are already very powerful and they do quite a good job of overclocking themselves within reasonable limits, and since hardware is reviewed by UA-camrs who routinely push everything to the limits, the more extreme hardware designs are always rewarded, and conservative/reliable designs are penalised. I essentially gave up on overclocking once I realised that I'm paying for imperceptible gains with heat, noise and instability, and it's completely not worth it.
In general I play around with staff to figure out what I can do before losing immediate stability (like, several hours of different stability tests) and end up with one or other taste of undervolting.
My 3070 works at static 1800MHz, with VRAM and 8000, consuming 33% less energy than defaults at max 65*C for 0-5% performance cost.
Even though I can get it to reliable 2010MHz, but it would add 20% power consumption (so, almost double from what I'm running) and will force me to replace thermal interface every 3-4 month.
I might have been ok with water-cooling, but there is no water-blocks for 3070-s, so I can't do that.
This has aged like fine milk
Something that should be discussed: it’s not only degradation, but is an overvolted / overclocked CPU still providing valid & consistent outputs? We always have needed to deal with uArch errata, but recent events re: Raptor Lake and Raptor Lake refreshed may warrant a deeper chat.
If decompression + integrity check crashes can be solved by dropping *below* stock multipliers, there’s something quite unusual here.
Just as DRAM overlocking can lead to bit corruption (more than just DRAM degradation or failed boots / total instability).
DOn't overclock. Life is good. Extended. My 5720k is still rocking 24/7 after 10 years on my everyday machine. Still fast!
The Celeron 300A was famous with friends an family. having it running at 450 MHz was no big deal...
The 300A was certainly a performer... I was Sr Tech at a local shop during those days and remember them well. The real clockers that few under the radar were the PDC e2100 chips... I was able to run one of those with a 100% overclock on stock voltage. No funky ram ratios, stock multiplier, just double base bus clock. Bam... 1800->3200MHz all day long on air. IMHO, Overclocking isn''t about the "fastest" speeds.... It's making the $100 chip run like the $500 one, just like the 300A did.
Straight up fraudulent data points, do not listen to a single word these scammers ever put out, how you actually have real jobs in the real world is the real question you should be asking yourselves.
YES! I worry about CPU degradation ALLL THE TIME! When our CPUs get to the point that they can verbally and physically degrade us, the human race is DOOMED!
I have a Core i7 9210 1st gen that's slow degraded.
It just can't OC as high it used.
If Linus Tech Tips or someone(not Gamers Nexus after his hit piece shenanigans) wants it for content, hit me up.
Could be a good example of slow degradation.
It is a thing if you use it OC with big voltage. If you use it at stock is ok or undervolted.
My cpu wants to go for a walk.