The intel instability and degradation rant.

[@SM-rn3xy]

TLDR; i9-14900k - ships at 6Ghz but skilled overclockers can get it to push 5.5Ghz reliably?

[@sidsmusic991]

mighty intel lol

[@gabber_]

this is a really funny way to put it, kudos

[@fi0nn]

'2' cores @ 6 GHz not equal 5.5 GHz all cores.

[@Slenderman63323]

Actually Hardcore Underclocking

[@Hectorscarclub13]

I got mine to 6.4ghz delided Liquid Metal direct contact 0 problems

[@GhOsThPk]

This was one of the last strongholds Intel had which was their Stability, makes literally no sense to buy an Intel CPU at the moment. They're more than happy to hype their Arrow Lake new arch instead of dealing with this problem, that they KNOW have been happening since last year.

[@ActuallyHardcoreOverclocking]

You gotta really hope that whatever decision making process lead to this mess didn't also get involved in arrow lake development.

[@BelowAmbient]

weird, mines been running direct die at 1.4v 5.8ghz for 9 months now without a single problem...also running Asus optimized without a newer BIOS... this is related to people not cooling properly

[@stefaneni5321]

@@ActuallyHardcoreOverclocking yes but the fact that they're already abandoning the maintenance of a recent product is just shit!

[@Amberion]

@@BelowAmbientand the server farm hardware failures?

[@ActuallyHardcoreOverclocking]

@@BelowAmbient 1.4V is less than the max boost VID on most i9s

[@572089]

The ring makes 1000% sense. that explains why the CPUs fail cache tests or mem-tests running in cache. that explains the inconsistent modes of failure between cores that explains the IO errors and NVME (northbridge PCIE) errors it explains why lowing memory clocks would increase stability it explains why datacenter providers who are running these systems 24/7 would experience these failures as a dramatically accelerated rate; because these CPUs are purchased SPECIFICALLY for high single core load tasks like game servers where the P cores will be drawing large voltages the whole time. It explains everything. and it explains why specifically its the 13/14th gen i9s and some i7s, since they have the largest single core voltage demand.

[@mx5klein]

100% agree it has to be the ring. I've gotten PCIE errors and memory errors until I downclock the cache frequency. I think workloads that rely on the P-cores are the ones causing issues since that is when the cpu can really boost high. I doubt degradation happens when all the cores are loaded since the frequency and voltages decrease to keep power/heat in check. This is the first PC component purchase I regret and I owned a Radeon VII for a while.

[@ningyang2074]

Can't agree more. My 13900KS could run ring freq @ 5GHz last year when I bought it, but I can only run it @ 4.7GHz w/ even higher VID (ring down bin disabled)

[@mycelia_ow]

Glad to still be on Alder Lake.

[@rluker5344]

Datacenters with 150w caps like those w680 motherboards, and all core usage like those motherboards will not see high voltages very often. Maybe 1v?

[@NUCLEARARMAMENT]

@@rluker5344I run a W680 ASUS board with an i9-12900KS w/AVX-512 enabled in the microcode. I run 4X32 GiB Hynix A-die ECC UDIMMs from Kingston, and it's clocked at 6800 MHz with CL30 and tuned subtimings. It is an incredibly solid and stable setup.

[@ctrlaltdel02]

I think Intel perfectly knows what is wrong with chips. They are quiet about it because fix would be painfull. The question is it more painfull to reveal fix, or just not fix and release new generation in hopes someone will buy it.

[@Alonne1]

they are probably trying to delay their response so that it doesnt affect their new generation releases

[@jayvee8502]

They want to pretend that it is not a big problem. They hurried the CPU development cycle just to keep up with AMD.

[@gravesddd]

can you imagine the high amount of RMA's from enterprise & consumer the last 12 months? they absolutely know.

[@ConeJellos]

I think Intel is secretly run by lizard people.

[@whohan779]

@@jayvee8502 In Apple fashion: "A small number"

[@carlr2837]

I have a 14700k, and out of the box, I experimented with overclocking it, as I always do. I ended up setting it up to underclock it 6% to achieve what I thought was a safe operating status. I haven't had any crashed or instability, but I'm now tempted to reduce it further. CPU longevity is more important to me than a few % points of speeds. In forty years, I have never had any CPUs fail, except for one Cyrix chip in the 1990s which failed after about 3 years, and one Core 2 Quad Q9650 that failed after about 12 years. Seeing massive failures of a CPU in their first few years of life is shocking.

[@dethskullcrusher]

@@carlr2837 if you can archieve 1.2-1.25v at respectable speeds that'd be sweet

[@noname-gp6hk]

For real man, CPUs never fail... Except now

[@carlr2837]

@@noname-gp6hk Yep, almost never, unless they are overclocked and/or overvolted, or if the CPU cooler fails.

[@572089]

the big thing here is that if intel KNOWS now that the P-core voltage is causing degradation and is still advertising their products as 6ghz out of box despite the configuration being unstable, then they're open for a MASSIVE class action lawsuit for defrauding customers and suppliers. Thats probably why intel has been so so quiet; they KNOW whats happening, there IS no firmware solution or any solution that doesn't tank performance, and they can't afford to give anyone more legal ammunition. Fixing the problem by dropping the max power limit or neutering the power bid would HAVE to come with lower clocks, and then they'd be in defrauding customers/suppliers/MoBo partners with false claims/advertising.

[@N4CR]

There is no fix. Watch level1techs video and video with techjesus and wendell. They are running them stock/locked core speed/voltage under spec/ram speed well under spec, temp under spec and they are still dying.

[@es-br8ck]

It is a Rainbow Blue Team operated company. Rainbow Blue doesn't care about technology unless it can be monetized, and only in terms of how and when to monetize it. A rainbow blue led company will NEVER wilingly recall a defective product because it is defective, but ONLY when forced to by a lawsuit OR their lawyers have determined that NOT recalling the product is more expensive in the short term. And all rainbow blue team companies rely on marketing and public relations management aka psychology for at least 1/3rd of their operation or more, and with that, a hugely defective product lineup to them is not a technological disaster to be handled by technological solutions, but first and foremost a market mishap, to be handled by marketing or legal. Typical for this team is also redirecting the blame to whoever and whatever, but NEVVVVER admitting the slightest mistake. Unless you have fifteen certified experts testifying under oath with three pieces of concrete physical evidence in hand, live, at a court hearting, they will not admit a failure. If rainbow blue teams really, really f up something that is impossible to treat with marketing, they will sue everyone for "contributing" to it or even merely "talking badly about it", and if that is fruitless or impossible, then they shut up and eradicate the happening from all their collective memory. In the near future looking back, it will never have happened. It the far future, it will never have happened and you are at fault for bringing up "lies". Most likely outcomes are: - the problematic chips will be neutered via microcode / UEFI update, once they really know that it'll help, but repackaged into a security update fixing CVE-2024-thisorthat. - the next chips will be "redesigned" in some way and marketed via bought and paid for influencers to assure us that everything is okay now - AMD will be sued for some patent infringement to bring down their stock value, too - the media will overhype the risk that AMD and TSCM face because of the Taiwan / China situation, so cautious buyers and investors will still think twice about buying AMD Possible outcome: - intel will castrate their 13th/14th gen CPUs via microcode update when 15th gens have been out long enough for intel to be sure that this problem doesn't reappear. Users will then be faced with the choice of running their 13/14 CPU into the ground if they want high performance or install the update for stability but reducing their performance down to level potato. And everyone buys new 15th gen CPUs, yay shareholder. Highly unlikely outcome: - intel coming forward and explaining how they pushed their CPUs too far and damaged them, leading to a mass recall and exchange from all OEMs. - This will never happen. Never.

[@H33t3Speaks]

Right, but don’t stay focussed on Intel. All of everybody is in on this one. Board manufacturers also, advertised being able to deliver power safely to these chips: so we’re going to watch the industry march lockstep in with each-other, including AMD who may get to play savior but probably not; because this is looking more and more like a fabrication level problem because of the crazy physical demands needed to fab chips at 7nm, a process also employed by AMD… .. .

[@Savitarax]

I find it only coincidental that the end of intels 14nm era of “stability king” dies as soon as 14th gen comes out.

[@fattyguapo]

love your videos

[@MrSvetrkravata]

what do you mean, 14th gen CPUs are surely out already?

[@H33t3Speaks]

Raptor Lake is 7nm … .. .

[@lllXavierllll]

The 14900K chips are degrading insanely fast. My 14900K on day one could run at stock frequency with a -0.075 voltage offset on the last 2 steps of the V/F curve, without any WHEA errors either at full load or idle. Every 2 months or so, I would randomly find a couple of WHEA errors which were instantly fixed by increasing the previously mentioned voltage offset by +0.05v. This week WHEA errors began showing up again. Now the CPU needs to run at stock voltage as is unable to handle any negative voltage offset whatsoever.

[@KeyboardSavant]

I experienced my first WHEA error in June, and my system has been running with the 14900K since September (2023). Since that initial error, I’ve had the system freeze (WHEA error) occur 5-8 times. It occurred twice in one day on July 2nd. My system just seizes up, becomes totally unresponsive and requires a manual shutdown. Perhaps I’ll adjust my undervolt as well, a bit closer to stock frequency. Have your errors been minimized since approaching the baseline again?

[@lllXavierllll]

@@KeyboardSavant yeah you need to increase the voltage slightly. Degradation is real and in my case every couple of months I have to increase my offset by +0.05v to mitigate those WHEA errors. My 14900K is Direct Die cooled so I have tons of thermal headroom.

[@Joh_Lam]

Got my first 14900k in December 23. Gave me some "status access violation" in Chrome at the beginning. For the rest seemed fine. Then starting from March 24 degraded rapidly. Needed more voltage, while at the beginning worked fine with negative offset. Luckily my retailer sent me a new one the same day I sent the defective one to them. This one now has better SP rating according to Asus and needs lower voltage and so far stable and no issues. Hopefully that better SP, lower voltage will let it live longer. If not, Intel will have to replace again lol. Mobo wasn't cheap either. Can't switch like that.

[@fenix144]

This is insane. I had a 3770k back in the day that ran for 6 years at 4.8ghz all core (which was huge compared to the stock 3.7ghz all core stock). I think I ran something like 1.375-1.4 volts through it. It’s prolly still kicking around today somewhere

[@Jutastre]

The raw failure numbers of 14900's seems to have caught up with 13900's pretty quickly judging by some of the published data, so I'd imagine at this point they've gone far beyond it even.

[@richard-davies]

Problem with AM5 and Ryzen 7000 was just a simple case of board vendors shoving way too much voltage into the CPU causing it to die pretty quickly, CPU itself was perfectly made. Easy permanent fix was AMD capped the voltage via AGESA so it couldn't go higher and that was the end of that issue. This on the other hand is far more problematic as these chips are not spontaneously dying or blowing up due to overvoltage they are dying at stock and at low temperatures even in a server board where in some cases the Intel CPU has never passed 125w. This unlike AMD seems like a physical hardware defect in the CPUs which no amount of BIOS updates will be fixing this. Intel more than likely know the root cause of this by now as they have the tools for it, problem is do they announce that potentially all or a majority 1300k and 1400k CPU have a physical design issue scaring shareholders, regular customers and customers that buy these in the tens of thousands, because if they do then they will be doing one hell of an expensive recall or replacement program with actually fixed silicon. Whatever Intel do it's going to cost them a fortune as well as loose customers.

[@erkinalp]

it'll scare intel's engineers too

[@savagej4y241]

There's no way Intel will announce such an issue, they'll sweep it under the rug like Nintendo did with JoyCon drift being inevitable and do no recalls. MAYBE will issue BIOS update settings for motherboard manufacturers, new stock settings being an undervolt and underclock compared to original stock settings. But the damage is done. You'd have to undervolt and underclock on day 1 to prevent damage.

[@Afurai_]

@@savagej4y241 "to prevent damage" is completely useless when the CPU is practically a nice house that has an invisible landmine that moves around the house randomly and grows slightly every day... sure, the house (CPU) will be fine for a while, maybe even a year or so, but the one day you "step" on that landmine. It's over..

[@snakeinabox7220]

​@Afurai_ eventually intel will get sued . They cant just sweep it . They are going to be forced to eather pay out customer's or replace chips .

[@foobarf8766]

125w is WAY more than enough to bbq a CPU if the temp goes over 74-75 none of these videos show temp in safe margins, would be surprised if warranty support offered

[@wewillrockyou1986]

I think Wendell showed that it isn't only degradation, but also some kind of manufacturing/design level defect in the chips that is causing the degradation to be greatly accelerated. Alder Lake i7s and i9s don't run *that* much lower voltages than Raptor Lakes, yet dont show any comparable degree of degradation. Intel must have screwed something up when they did the hard design when they upgraded Alder into Raptor. Would also be interesting to see the rate of crashing/degradation relative to the VID table.

[@VioFax]

I don't think its an accident. I think its on purpose at this point. Some new artificially created-chip-rot-planned obsolescence BS. That isn't timed in properly yet or something...I call BS.

[@AleraKira]

@@VioFax Doesn't have to be planned, they designed Raptor Lake from start to production in 11 months which is unusually fast and the engineers bragged about this fact. They likely missed something and we get these end results.

[@winebartender6653]

​@@AleraKira There really wasn't any planning beyond binning tolerance and TVB adjustments. You can't go from design->tap out-> retail in 11 months. It's logistically impossible. Let's be clear, 14th gen is a pure refresh of 13th. As noted above, binning is 99% or the "work". There is a reason lower end P-core counts increased (and why plenty of brand new 13th Gen still rolls off the line). It's genuinely a voltage issue, pure and simple. We saw a 100mv increase maximal TVB, with every level of voltage increased on the TVB tables to a lesser degree. Because of this, the CEP lookup needed to be changed to deal with these changes. What Intel did was literally: "Well, we need to have some single core improvement over 3 generations. Find a way" and this was the outcome

[@cmiex]

36:50 ... but there lies the issue with the assumption that it could be "some kind of manufacturing/design level defect". Why only the high voltage parts? I'm looking forward to what information those investigative tech-KZbinrs will dig up.

[@winebartender6653]

​@@cmiexThere isn't anything to really dig up. So many parts of the CPU rely on the core voltage rail (cores, L3, ring, and others) that any of those areas could be the first to degrade. If it's any consolation, i'v been running a 12900ks for 2.5 years @1.4v/5.6/5.3(ring) fixed voltage (direct die cooling) and have had it degrade down to 5.5 within the past 6 months (@1.35v).

[@ShermSpinner]

39:50 regarding lowering the boost clock to fix it... wasnt there a whole thing about ppl saying the main/only "lasting" fix they found being just locking the multiplier at like 53? Either way, can't imagine that nobody at Intel saw this coming. Smells like a whole lot of engineers going "told you so" to the merketing people that made them push the silicon so much so they have something "new" to sell.

[@samiraperi467]

Locking the multiplier might not even be a fix, it's just a guess at this point. And by that point you're better off buying non-K. :P

[@marsovac]

marketing team cannot push anybody into a design decision that involves risk. first the CTO ask for an assessment and then discusses with the board of directors including CEO, and there the marketing director can try to justify the risk

[@tamarockstar45]

@@marsovac Are you on the board of directors at Intel? How the hell do you know?

[@hmmodi9052]

​@@marsovacHistory says otherwise

[@hughJ]

Feels like bizarro world that any crashes and application hangs is going to prompt me to think about possible CPU failure rather than something else. Given how relatively bullet proof CPUs have seemed in past decades, I have to imagine the quantity of RMAs that Intel have been having to process for Raptor Lake is several orders of magnitude greater than their historical average. I find it impossible to believe that Intel wouldn't have noticed this trend even just a few months after the 13th gen first going on sale.

[@pf100andahalf]

They were hoping that they could sweep it under the rug,

[@K31TH3R]

In over 20 years of building PCs, every time I've had a CPU go faulty (3 times), it was my own fault, because I was running too much voltage. That is until 2019, when I had a Ryzen CPU go faulty on stock settings after ~6 months. Both AMD and Intel are running their silicon too close to degradation territory. On one hand, it's nice to know you're getting almost everything out of the CPU you paid for right out of the box. On the other hand, failing CPUs are now MUCH more common than they ever have been.

[@VexxedSR]

Yea honestly I always thought CPUs were the most likely to survive the longest if not being pushed with overclocks and high voltages. Look at the amount of old core 2 duo/quads and Xeons still in circulation. Now fast forward 5 years into the future, we will probably have 0 of 13900k/14900k floating about with the failure rates

[@frantavopicka5259]

All thanks to management decisions at Intel to boost CPUs into overdrive to be at least somewhat competitive. I refuse to believe that any competent engineer would not shout loudly at execs saying that these levels of voltage are dangerous.

[@Aristoper]

@@pf100andahalf That or they were hoping it wouldn't snowball, but seems like that's what's happening.

[@giglioflex]

Wendell's video seems to suggests that lowering frequency only makes the degradation show up at a later point. The gaming server providers downclock the 13900K and 14900K out of the box now and they still show degradation regardless. I'm not sure what the problem is but this to me seems like a recall.

[@OdinAlgeron]

the problem is 2 cores boosting into a death spiral of heat + voltage the solution is simple, lock all cores to a set speed and see if you can undervolt problem solved, unless you are a moron and did it too late and now it has degraded and cant function with proper voltages

[@fepethepenguin8287]

Why u call a person using a product at spec a moron. How could they have known the product was defective in advance in order to bandaid it preemptively

[@Pslytely_Psycho_GreybeardGamer]

@@OdinAlgeron Moron? A bit harsh don't you think? The vast majority of users are not tech savvy enough to know that the MB manufacturers were overclocking the systems in the background. Hell, I'm not a tech but far from a rookie and I didn't realize what was causing my crashes immediately. Especially since 'out of vram' (on a 12GB 4070ti?) and nvidea driver crashes looked like I had a bad GPU. I figured it out just before the information dam burst, but most users expect the MB to run at recommended values out of the box, my ASRock MB was overvolting my system from day one along with a default TJ Max of 115C and it was over a month before BSOD's began and somewhat longer to figure out what the hell was going on because I did not do any overclocking. My only clue was a few games that either would not start at all or crashed quickly with nvidea errors. It was the CPU all along.

[@OdinAlgeron]

@@fepethepenguin8287 you are a moron because you didnt monitor temps and voltages is it the car manufacturers problem when the driver never checks coolant levels or changes the oil and blows his engine???

[@Grarlic]

​@@OdinAlgeron Yeah, you're so much smarter than millions of people now that you have the benefit of hindsight and immeasurable real-world testing. Running a processor at stock settings, rookie mistake.

[@rci-tf2zc]

intel hit a wall awhile back and they"ve been trying to dig under said wall but the grounds to hard so you got 13th and 14th gen crap.

[@CedricBassman]

I've been saying this for a year now. They're trying to stay "faster" than AMD and instead of actually improving their design, they simply brute force it, Customer be damned. But it seems to work, since anytime anyone brings up AMD there will be enough people going "just buy Intel CPU X, they're faster!" showing that big Numbers are truly all most people are interested in.

[@wildesage4172]

For sure. They were struggling to get >10% yields on 10nm manufacturing for so long. It was a shitshow. And clearly they didn't break through and perfect it. More like... flopped through. Lol.

[@noiprocsZ]

Or after digging under the wall they encountered soft ground that caved in and now they are in this huge hole which is 13/14th gen, the wall represented their corner cutting with gens and the ground is them not researching better

[@CedricBassman]

@@noiprocsZ It's actually crazy that Intel seems to have the same marketing strategy as AMD with their GPUs. They are taking notes from the wrong AMD branch.

[@TheTastefulThickness]

Intel is better and they will still be better next year and the year after that and no one buys AMD and that's for good reason. I won't even mention how touchy they are on RAM or how their gpus don't mix well will LLMs running local or quicksync or explosing cpus or ...stop...hes already dead!

[@adamksenior6669]

Well, I bought a 14900k back in February and tomorrow I'm getting the warranty replacement delivered. It had random issues for the first couple of months, but about 2 weeks ago it seemed to have degraded rapidly to the point I spent over 12 hours fighting BSOD while diagnosing the root cause and couldn't even reliably get into safe mode. I ruled out drivers, software, and most of the hardware before taking it to a shop that eventually put the CPU into another system and the BSOD issues persisted. The one time I went top tier in building a PC and I get this experience. 😑

[@puffyips]

Eh you made a coin flip of a mistake, one side 14/13900k and the other 7800X3D

[@Corrosion37]

Thats how my 1st 14900k was. Took months to get a refund. got another on amazon for 500ish dollars. Been running good this one.

[@freesS3]

@@adamksenior6669 Please tell me what mainboard you use and exactly what RAM, really want to know this

[@adamksenior6669]

@freesS3 I am running an ASUS ROG Maximus Z790 Dark Hero with Corsair Vengeance DDR5 6400 CL 32, 2x 16GB sticks. I orginally was running 4 sticks, but I had to learn the hard way 4 sticks of DDR5 doesn't seem to want to run at the advertised speeds, so knocked it down to 2 sticks. That helped rule out RAM sticks during my troubleshooting at least.

[@adamksenior6669]

@Corrosion37 I'm glad you're running smoothly now and I hope it stays that way.

[@Torbjorn.Lindgren]

The company I work for has lost 4 of the 8 earliest 13900KF we bought, all to CPU failures - the last one was recent so tried loading ASUS latest Intel Safe settings with no luck, so I'm definitely on team "it's degradation". These were all running ASUS stock settings (at time of purchase) on ASUS Z690-A, no overclock and using 240mm AIO. A number of later 13900/14900 (all K or KF) has had partial failures where updating BIOS and loading the safe settings has brought them back - for now at least. I don't think I've seen it with any of the i7's yet, and obviusly they're now all detuned to standard VID/CEP/IccMax. The jury is still very much out if running with those settings merely extendes time or actually avoids degradation - it's certainly not enough if they've degraded too far. I really would have liked Wendell to check and confirm that at least the SuperMicro workstation boards actually didn't disable CEP and/or IccMax - would hope they wouldn't but you always want to verify your assumptions and he didn't. Not that AMD 7xxx has been a cakewalt, memory stability at 2x32GB has been variable (seems to depend on AGESA version but newer can be better or worse) and 4x32GB is a HARD on AM5, but is easy on AM4/Intel DDR4 and Intel DDR5. Which is a pain when people need that memory....

[@madclone84]

Your problem is running an i9 14900k on a 220 cooler when it needs a 420MM and your motherboards probably aren’t up to the task of taking the I9. It’s recommended you have a new z790 MB designed to take these chips.

[@rustler08]

​@@madclone84 First of all, no it does not need a 480mm cooler. Unless you're running full on benchmarks, the CPU is not going to be maxed out. And even if it is, the design of the CPU is to hit a target temperature and fluctuate power to keep the highest clocks without exceeding that target temp Second, even a 480 cooler is not enough. Hell, you can have dual 480s and still thermal throttle. The difference is just how fast will the AIO loop saturate.

[@madclone84]

@@rustler08 yeah I’ll take my advice with no issues rather than this guy burning up his chip

[@macicoinc9363]

@@madclone84you are buying dogshit tier AIOs if you need a 480 to cool 250 watts lmao.

[@madclone84]

@@macicoinc9363 do you even have a 14900k. You need a 360 or higher to cool the chip especially if you have multi core enhancement enabled from the asus MB. It’s going to get hot and you need to cool it

[@CAPiiX7s]

I'm on my 3rd 13900KS. Each chip got worse over time and I highly suspected degradation. It's crazy how this is blowing up now, when I first encountered a chip failure back in August 2023. My current chip occasionally blue screens, but it isn't frequent enough for me to deal with RMA'ing at the moment.

[@asdf_asdf948]

Just go with amd, can't imagine rma a CPU 3x

[@kimjongpoontv69]

Been threw 2 14900ks myself, and both only stable with HT off and fixed vcore. About to just jump to AMD at this point

[@rapamune]

Why would you be bothering with Intel when x3D chips are out lol.

[@kimjongpoontv69]

@@rapamune Well I didn’t have any problems with my 13900k that I sold, but there were problems day one with 14900ks and thought my shit was faulty but the second one I got wasn’t any better. No matter what I changed in bios. Only thing that stopped the crashing was disabling TVB and HT and using a fixed Vcore. But wasn’t happy losing out on performance that I was paying for. I built an entire 4 rad loop and direct die for this thing. But I have ordered a 7800x3d, heat killer block and X670e board, won’t touch intel again for a while.

[@asm_nop]

@@kimjongpoontv69 Yeah, that's a nightmare scenario. Glad you're on your way to a better experience. I'm running a 7800X3D, and since the AGESA updates to fix the wacky DRAM training, this platform has been amazing.

[@rapamune]

Gamers Nexus hinted at it being related to cache/ringbuss (in their recent interview/talk with Wendell). They should be releasing a video on it soonTM.

[@winebartender6653]

It's all on the same voltage rail (L3/ring bus/core) so whichever goes first will be cpu dependent.

[@Zarcondeegrissom]

@@winebartender6653 is clock distribution also on that, esp the NVME/PCIe IO section, and core clock distribution? There are a few things it can be that the entire CPU runs on, ring bus, clock distribution amplifiers, cache/latches, etc. Something is sensitive to silicon quality and atom migration.

[@mx5klein]

Playing around with mine that is mostly stable and at stock 5ghz cache frequency prime 95 crashes instantly. Lowering the frequency down to 4.5ghz has it solid. I was using smallest FFT's so it should only really hit L1 and L2 cache.

[@snowwsquire]

@@winebartender6653yes it’s all the same rail but that doesn’t mean each piece can tolerate the same voltage

[@cemsengul16]

Can't wait to watch their follow up video when it comes out since Intel won't speak on the matter.

[@joeykeilholz925]

Intel bragged so much that amd wouldn't be able to create a compelling and stable platform when ryzen was new. Now look at Intel! They have failed in every conceivable way - except arc, you can stay.

[@renevandenbosch9967]

That's not a majority opinion yet, that arc can stay. Do you like it for gaming?

[@benjaminoechsli1941]

And Arc is only staying because they need it for iGPUs, if the rumor mill is to be believed.

[@iiisaac1312]

Intel makes good SSDs as well

[@MissingNumb]

AMD listens to the engineers, and follows a very old school methodology of incrementalism. Intel hasn't really taken as much care of their CPU engineering dept, and shifted into a modern "chase the market" methodology since the "lake" generations. Arc is likely a part of this debacle as far as drawing resources and oversight away from other areas. Intel also chases a lot of other revenue streams (ie govt subsidies) and have also taken on the juggle of shifting/building fabs. Industry as a whole is kind of pushing a bit too hard to meet the "goals" they all set for the market as far as performance. We're reaching limits of material conductivity (at least modern consumer priced) for the data bandwidth we are already pushing past. Copper and silicon will soon be a thing of the past in your whole system.

[@Equ1ne]

@@iiisaac1312 Intel no longer makes SSDs afaik, they sold that business to SK Hynix a while back. What was the Intel SSD business is now Solidigm.

[@spyder256]

The main thing I always heard people say as to why they were going with Intel, because they were more stable becasue Ryzen was too new or whatever. Welp, so much for that

[@macicoinc9363]

AMD still needs to get ram and Expo sorted out. I have yet to work on an AMD system without ram issues. You shouldn’t have to wait 10+ minutes on first post (if it even works)

[@KitsuneKiera]

@@macicoinc9363What kind of ram issues out of interest? I have an AM5 build with 5600 XMP DDR5 and I personally have never had a single issue with it, although I know 5600 is hardly pushing anything hard there.

[@beasttitan8747]

​@macicoinc9363 Yea AMD has problems too, took me 6 days to get my 5900x running. It never rejected the ram or anything, just needed more power or else it would go crazy 🤪. Had to up the voltage which I wonder how many beginners know this?

[@Geralt26]

In the meantime I had 0 problems with AMD 5xxx and 7xxx PCs :). My decision become better and better with time:)

[@churblefurbles]

I dunno, before this Ryzen was known for failures, perhaps due to chiplets.

[@erkinalp]

I had a few instabilities in my 5950x, but that's because i've undervolted the cpu (~75mV down from -30 PBO 10x) and overvolted the ram (slightly beyond XMP levels) at the same time

[@Geralt26]

@@churblefurbles I think this opinion was because of this curve optimization + early age bios things + too high memory settings (mine have not problem to work with 4 bank 32 GB memory from the box with declared speeds and even with forced T1). And about curve: I think it can be tricky to test full stability after changing curve as it can work even 2 months without crash and after 2 months you finding a one game when you can see crash sometimes (even if it work great in 50 others). I prefer to have full stability and after this situation with one game I just reverted to default curve (as it does not give a big impact anyway) and from that time I did not saw even one time instability. Ps. i also prefer to have temporary early age bios issue corrected then hardware problem like with intel 13 and 14th generation, when intel try to say nothing is wrong :).

[@Geralt26]

@@erkinalp Yes. It was also my case on the very start. Undervaluing can be tricky as it can work longer time and in many games/applications but then you find that it is not 100% stable and in one sometimes crash :). But it is outside specifications and you do it knowing it. With intel it is different story. It happens right from the box and become worse and worse with time.

[@erkinalp]

@@Geralt26 in my case, idle and full load are fully stable but the transient just above the idle is metastable

[@topaz987]

My 3950X was and my 5950X currently is 100% stable. Looks like my next CPU is gonna be either Zen 5 or 6

[@arthurbesnard1536]

5950X is the most efficient and stable CPU I ever used. This thing is just out of this word

[@dolphhandcreme]

Intel already had CPUs that degraded and died, unfortunately they were used in firewall appliances. Apollo Lake LPC degradation. These CPUs where unable to boot because their LPC-bus died and they couldn't read out the BIOS anymore EDIT: this was 2019. 2 years earlier, they had a similar problem. (Atom C2000 LPC)

[@razorblade7108]

Today I learned Apollo Lake also had the same issue lol

[@stage666]

FG-91E

[@arthurbesnard1536]

In servers too, Scaleway still use a lot of them

[@l3lackoutsMedia]

​​@@arthurbesnard1536 Im feeling more and more confident about my dual socket sandybridge-e server. That thing will probably never break.

[@jakubmi9]

Oh it's much worse than that. The LPC failures were in all Silvermont chips, continued into Airmont (Silvermont die-shrink), *and* somehow got into Goldmont (completely new architecture) as well. Tick-Tock and another Tick, all with the same degradation problems. And it wasn't just LPC - SD Card, RTC and USB were also degrading on all of those chips.

[@jamescampbell6728]

Moore's law is dead said intel engineers were bragging about how fast they got out raptor lake and beat zen 4... How impressive of them to raise voltages to an unstable level in a mere 11 months!

[@madclone84]

Intels official statement is that it’s not the chips, it’s the motherboard manufacturers and game’s not optimizing their products for this new chip

[@MrSvetrkravata]

@@madclone84 I wish it would be true, however, it is bullshit

[@The_Man_In_Red]

​@wikingkrig5801 I wonder if the via and substrate layer density is so high that voltage is bleeding through layers due to the fact how many cores are shoved on these relatively small chips? 14900k = 24 cores @ 257 mm² Threadripper 7965WX = 24 cores @ 4x 71 mm² (284mm²) Processor dimensions for Socket sTR5 is 4410.9 mm2 Processor dimensions for LGA 1700 is 1,687.5mm2 So Intel shoved the same amount of cores onto a substrate that's roughly 2.5x smaller. I know Intel e-cores are not the same at all as a Zen 4 core and the architecture is vastly different, but still.... Makes me wonder if LGA 1700 needed to be nearly twice the size to accommodate 13th & 14th gen. I'm probably wrong on all accounts, just speculating and sharing an interesting bit of info I noticed 😅

[@whale27]

It is indeed Unreal Engine games. The engine uses the oodle compression toolkit, which is very sensitive to errors during packing/unpacking and will fail upon building a game for shipping or launching due to CPU instability. I spent hours upon hours trying to fix this as it seriously impacted my work.

[@The_Man_In_Red]

Not just CPU instability, I got this error in Remnant 2 once or twice. Replaced my 9 year old dying PSU and all errors were fixed. Ryzen system. (Critical Kernel -Power Failure, I forget the event code)

[@dorkasaurus_rex]

the fix for this is to use XTU and downclock all pcores to 53x or 52x, will fix this in most if not all games where you have this issue

[@dabombinablemi6188]

This whole debacle has further cemented my original choice in having Ryzen be the upgrade path from the i7 4790K.

[@LeadRakFPS]

Well, you made the right choice. I just upgraded from my i7 4790k to the i9 14900k and it was problems from day one.

[@Psythik]

Do it! I went from a 4670K to a 7700X and the difference in performance is mind-blowing. At least two orders of magnitude faster, no joke. Paired it with 32GB of DDR5 6000, a 4TB PCIe gen 4 NVME, and a 4090. Programs open the second I click the icon, and loading screens in games are pretty much a thing of the past now. 4K gaming at 120Hz is amazing!

[@snowleopardd]

I also upgraded from a 4790k to a 14900k. I cry in my sleep every night wishing I got the 7950x3d instead.

[@macicoinc9363]

Funny, I upgraded my 4670k to a 7950X3D. Wonder how many people were on 3rd and 4th gen intel until this gen.

[@dabombinablemi6188]

@@macicoinc9363 I myself had the 3700X as an intermediate between the 4790K and the 5800X3D. The initial thought (and result in benchmarks I ran) was double the performance at well under half the power consumption - 4.6GHz had my i7 drawing close to 120W (which combined with a 200W GTX 970 that boosted abnormally high at stock...toasty).

[@nempk1817]

The intel silence is what i don't understand.

[@PorscheRacer14]

So, I'm going to date myself when Intel was silent on my Pentium MMX. Anyhoo, many years later I received a cheque for $11USD and the chip still lives on with that issue. Unless government compels them or any large corporation, they have no real reason to communicate.

[@drCox12]

Anything they concede would cost them a lot of money.

[@xmlthegreat]

They have a lot of money to lose by coming out and admitting something is wrong. If they keep mum for as long as possible, the only thing they lose is customer trust which most corpos don't care for anyway

[@PorscheRacer14]

@@xmlthegreat Yep, not like AMD can supply the market anyways. They know they are the first or sometimes only choice.

[@RFC3514]

They've done it before. There was a data corruption issue with 800-series chipsets back in the Pentium-III days, they knew about it, and kept selling them. It's the old "is it cheaper to deal with potential lawsuits than it is to lose sales?"

[@insu_na]

My own personal hypothesis for what's causing the issue is the internal clock generator failing. Every error that I've seen come out of unstable 13th and 14th gen intels looked like an unstable BCLK to me. Some reports like everything becoming extremely sluggish before a total crash, etc. I don't know if they used ol' reliable Quartz in their internal clock gens, but I think it's more likely they used MEMS or NEMS oscillators as the clock generator. AFAIK intel has no greater expertise with MEMS/NEMS so it's quite possible that they created and implemented a design that's particularly susceptible to degradation. An easy way to test this would be to plop a known unstable CPU into a mainboard with an external clock generator to let it override the suspected unstable internal BCLK with its own stable 100MHz BCLK.

[@saricubra2867]

If a CPU degrates itself it's an absurd amount current or voltage, same thing happened with Bulldozer and Piledriver, Pentium...

[@insu_na]

@@saricubra2867 That's one possibility, yes. It's not the only possibility.

[@TheBackyardChemist]

Intel had a major issue with the clock gen inside Atom CPUs degrading, like a bunch of really expensive Ethernet switches, NAS boxes, etc. have been bricked by this

[@ActuallyHardcoreOverclocking]

@@saricubra2867 Piledrive is basically indesctructible in my exp. The motherboards die before the CPUs do. I ran some of them at like 1.6-1.75V on air/water cooling

[@timothygibney159]

My IO is what was going out and things like nvidia drivers reinstalling themselves and recognizing my monitors and usb keyboards and mice reinstalling themselves at startup or not working until I unplug and plug them back happened first. My ram was fine in testing with memtest86. It is the IOMMU portition that looks like it is glitchy first which would impact clocks and video ram paging second

[@ZeroHourProductions407]

Shoot, I still have Athlon xp cpu's that work as well as they did when they were new. The notion of intel pushing things to this point, _from the factory_ on top of it, is insane to me. Feeling like I dodged a serious bullet, both in the fact that my main pc is still on a 9th gen but also that my laptop is a Ryzen cpu instead of an intel.

[@mistkeyblade]

The only reason I stuck with Intel is because it was stable. I always ran into the odd application that would hiccup on an AMD CPU but work flawlessly on Intel. Looks like Intel just basically killed the best thing they had going for them after scrambling to catch up to AMD's competition

[@samiraperi467]

My Zen 3 Ryzen has been rock solid after tuning voltages (y-cruncher and whatever else AHOC uses). Downside is I can't update BIOS if I can't be bothered to re-tune, because Asus has a habit of screwing around. Last time I updated Cinebench performance went up but I got bluescreens, so did a rollback. (And I get 50 MHz above promised single thread boost, just with undervolt and higher power limits, which are used only when benchmarking and transcoding.)

[@denvernaicker8250]

i based my decision on the same reason and from hearing youtube tech channels say the same thing (jayztwocents), i have a 14700kf and its running stable so far, but early on undervolting caused issues

[@JasonMendoza-hd3ce]

@@denvernaicker8250 It basically comes undervolted, you can't get more out of it lol. It's shipped so near the max that there's nothing left and is why one person in an airconditioned room is fine and the next in a warm room isn't, they didn't build in the normal amount of headroom to allow people in Arizona or India or the arctic circle to all be able to use it, they basically made it for the arctic circle only and everyone else has an unstable chip.

[@giglioflex]

Can't say my 5800X3D or 7800X3D have a hiccup in any application. X3D CPUs are the best you can buy for consumer workloads.

[@timothygibney159]

@@denvernaicker8250 I heavily undervolted mine trying to keep it stable. Now I was told that my heavy under volt could have killed. I am confused as I was told degradation is always the cause of voltage changing the structure on a nano scopic level until connections can't occur anymore without mroe volts. Oddly ramping this up didn't help either as the failure already started

[@dankvader420]

I find it funny that voltages we used to think as "too high" and "will degrade your chips in the long run" are now the default voltages from the manufacturer…

[@savagej4y241]

What makes it worse is that these newer i9's will only take around a -100 to -135 mV undervolt. That's nothing, compare that to the 5800X3D that can undervolt all the way from 1.35V to 1.025V but still hit 4.4GHz overclock on all cores with PBO. Its not the king but the 5800X3D is still a legend lol

[@Dave-ct1jk]

Buildzoid puts out a video on 13/14th gen failures, sees that its an hour long. Grabs popcorn. Lets goooooo

[@Wasmachineman]

not only a feature length KT3406E video but now also a 1 hour rant by BZ, fuck yes

[@rkan2]

Damn, 10 likes on this tweet too lol 😂

[@erwinvb70]

There a two groups of people who watch this, the first are Intel owners who want to know what’s wrong with their CPU and the second are AMD owners who enjoy some confirmation they made the right choice

[@JCDentonCZ]

What confuses me more is that people DIDN'T expect this exact scenario when Intel announced these massive TDP using heaters.

[@killswitch8493]

the raptor lake i9 is just desperation. they're trying to get the top spot on the charts and they're using brute force - to the detriment of everything else. even the raptor lake i7s are blown up to the max, the i9 is just over the top. even at the darkest days of amd they never released processors in that kind of state. edit: also note that the 12900k stock clock speed topped out at 5.2ghz. the 12900ks went up to 5.5ghz. this is still considerably lower than the raptor lake chips.

[@mehck2]

>even at the darkest days of amd they never released processors in that kind of state. One of AMD's first releases back in the day literally caught fire in the socket. Please do not fanboy a corporation anywhere that I can read it, so that I do not have to ridicule you for it.

[@sunlbx]

@@mehck2 don't forget to ridicule Intel for NetBurst pushed to 3.7ghz out of the factory, which was the best they could do to counter dual core athlons at the time

[@iLegionaire3755]

Clear that you never used AMD’s Bulldozer platform.

[@sunlbx]

@@iLegionaire3755 that's a different era where the tables were turned. let's not be cherry picking here, both the companies had ups and downs

[@michaelmalenchek4575]

AMD FX 9590 would like to enter this chat.

[@scpdatabase69420]

I was on the edge between getting a top line intel 14th gen or Ryzen 9. I am glad I stuck with my Red-Green mix for my new build rather than switching back to Blue-Green. The R9 seemed to have higher lows but slightly lower max fps. Felt much better suited for the type of performance I was after. This is a surprising benefit of my decision.

[@PowellCat745]

If you only game, the 7800X3D is a better CPU than 7950X3D.

[@p_bitty]

As a i9-14900K haver, you have no idea how much I appreciate your videos, especially for the settings to stop my pc from crashing.

[@Vivicect0r]

Intel keeps their mouth shut cause fixing the issue is probably impossible by software. They must do a total recall, but they can't afford it. So they make it look like nothing happened. I am so happy I have decided to jump the AMD bandwagon this time.

[@forog1]

Intel just needs to recall the whole 13th and 14th gen they are done. Their best "Stable" highest performing CPUs are the 12700k and 12900k/kf

[@FreelancerVideoDump]

my 12900k was so shit it couldn't even run xmp, sigh

[@asdf_asdf948]

@@forog1 do they even have the money for that scale of recall

[@darkkingastos4369]

They won't do a recall unless the government forces them to do it and that won't happen unless consumers file complaints

[@SirSomnolent]

Their market cap is like, 150 billion. Their revenue from chips something like 50 billion/yr? which is almost 1:1 with their operating costs. Even if you only recall 10% of past few years... bad. 50%?...

[@user-ew8lt6yi2i]

@@asdf_asdf948 u serious? this isn't a random shop near your house where u buy cigs and soda. Intel revenue for the quarter ending March 31, 2024 was $12.724B, a 8.61% increase year-over-year.

[@carl4889]

If there is actually a hardware flaw with Raptor Lake (be it suicide voltages degrading the silicon, or an actual fault with the ring or I/O controller) do you think that would explain why you had such a hard time getting stable memory overclocks with it? ie., were the errors you were getting caused by general instability, or were they actually caused by the early stages of whatever problem is affecting RL?

[@Saturn2888]

My wife said they're gonna have to send someone to Mordor to deal with the ring of power. You can tell she doesn't care about the Intel controversy.

[@DeepInsideZettaiRyouiki]

Which is the most retarded stupid silly and non-empathetic approach ever. That you flex her approach also says a lot of bad about you. From both... horrible policy and operation of today's corpos against all consumers to personal lack of empathy... ...but sure, she would Cry-Karen like a Roaring Kraken if some non-existing woman bollock was affected.

[@collingtech1]

class action lawsuit incoming?

[@stevetheborg]

in 1996 i replaced my clock crystal on my ibm 80286 and overclocked it(TOO MUCH). it rapidly degraded , blowing chunks into the ceiling. it needed a heatsink.

[@parbs198]

So i've told BZ my LLC findings before, but while overclocking my 13900k on z690 unify-x i've ran into a situation where medium droopy llc has resulted in lower idle voltages, lower volts / power consumption during gaming / medium workloads, because you dont achieve maximum 'droop' until your cpu is being hit with maximum loads. But here's the thing - the medium llc also resulted in greater stability in my extensive testing. Now i did separately calibrate the ring vid / ecore vid - for instance i ran ring at 5ghz with the ecores and pcores underclocked, so i could isolate the vid requests to be ring dependant, and found the lowest voltages at which they caused crashes for a given frequency. I also did the same isolation on the ecores, and then used AC/DC loadline values which resulted in a medium droop, and at which they were both able to run plus adding a small amount of headroom. It's only now that i setup my all-core OC (i'm not using single core boosts at all, its all core OC all the time), and to maintain the VID requirements of the ecores and ring, I simply used adaptive oc by modifying the V/F curve with offsets. The end result was 5.7ghz all-core oc, at about 1.4 max idle voltage, and max load voltage around 1.32. Note i have also enforced a 253w long term power limit, with 288w short term. This has been 100% stable after endless amounts of y-cruncher / OCCT / gaming / video rendering which it has been hammered with. So essentially by using a less droopy LLC, i have increased stability, the ring and ecores dont get unhappy when the p-cores downclock from hitting TJMAX (which i have set at 90C) or by hitting the power limit. I have my own theories about how this is the case, but im wondering it the undershoots are actually more desirable than having an overshoot from say, a droopy LLC sitting at 1.5+ voltage running single core boosts, then when the system suddenly hits a load there is a delay in the VID request where the power is smashed into the cpu at 1.5v with high watts before cpu has time to downclock to an all-core load and send the subsequent lower VID request. I would be very interested to see if the all-core clocking and less droopy LLC is actually preventing degradation from the above situation. Any thoughts?

[@keggerous]

I've been having issues for about 3 months. I fixed the issue after a month, went on with my life. Recently, I started playing Alan Wake 2 again. I already set my power limits to the suggested levels and my computer WAS working fine. Well, turns out Alan Wake 2 didn't want to work anymore. Apparently my 4090 doesn't have enough video memory? Hmmm... Anyways, Corsair (who I bought my computer from) is replacing my CPU. Meanwhile intel ignored my inquiry to their support.

[@ChumlyFernando]

I genuinely appreciate the outrage and coverage about this. Chasing numbers for the sake of bragging rights and then ignoring the issues that come from it is inexcusable. I have recommended 13700k and 13900k cpus to friends over the last few years bc I /knew/ intel would be stable and they have great performance. Now im following this so I can ensure i know how to support them when they break

[@StevenSheridan31416]

My gut feeling is that we have to think beyond the top-level numbers, eg Vcore - that this phenomenon involves microscale variations. That is, the answer will involve physics. Specifically, some phenomenon which was previously negligible rearing its ugly head. Eg some effect that used to cause 0.01V transients, which, when no one was looking, grew to now causing 0.2V transients. How much will the actual voltage across the silicon vary from point to point at a given Vcore? If the P-core power throughput is ramped up to an unprecedented level, could the load line end up delivering 1.6V to the cache and E-cores? Or perhaps there are high-voltage transients from inductance as the P-cores switch down from mega-power surges? I don't know how much inductance CPU circuits have. Or perhaps it's just the VRM inductance, saturating the CPU's capacitors and flooding the ring with a transient 1.7V? These are just a few obvious thoughts, but I have no experience with the fairly extreme time-space-current-voltage regime involved here. All I'm saying is, I think the answer will be outside the box of the usual physics and data points that are usually measured and contemplated.

[@dabriudabriuiutubiu]

AFAIK Intel redesigned the ring for Raptor Lake to support more e-cores compared to Alder Lake (8 vs 16). Ring bus implementations get way more complex when you add too many points to it.

[@kotekzot]

As far as I understand it, there's nothing complex about adding more nodes to a ring bus, it just creates more latency as the average number of nodes a message has to pass through before reaching the destination is increased. IIRC E-cores sit on the ring bus in cluster of 4, so adding 8 more E-cores should have added 2 extra nodes to the ring bus. With 8 P-cores and 4 E-core clusters, that's 12 nodes vs 10 for the 8+8 models and 8 nodes for 8+0 models, or a 20% and 50% increase in the number of ring bus nodes respectively. I can see why they felt increasing the frequency of the ring bus was the only way to maintain gaming performance, which is very latency-sensitive.

[@PAIN-ot4cj]

so is it just a matter of time then till our 14900k and 13900ks break down ?

[@N4CR]

The silence and long-running issue this is, tells me intel is lawyering up for trust action defense. Big problem when you have 50%+ failure rate at conservative/stock settings in professional server environment, eliminating poor end users knowledge/hardware issues/etc.

[@Wintelburst]

secondhand market of these chips might be a mine field for unwashed masses that are not tech litterate

[@cemsengul16]

Yeah I am finished with Intel. I have never purchased an AMD processor in my life but now I will. I can't accept this betrayal from a company I spent so much money on over the years.

[@karlstenator]

My last AMD was XP2000+ and then Intel with Core 2 Duo, and finally my i7 2600K, which has prompted me (after almost 15 years 😂) to look into upgrading - and after following this drama for the past few weeks, I'm going to go with AMD.

[@N4CR]

@@karlstenator I upgraded from 2600k to 7800x3d end of last year and it was an amazing jump and worth the wait. Get CL30 DDR6000 it runs stock.

[@UngoKast]

13900k and 14900ks. I thought it was my RAM, then my motherboard, then my SSD, then Windows. I have replaced and reinstalled all of those and both CPUs have crashed nonstop

[@vasudevmenon2496]

Over the last 20 years I've been an Intel guy. I still have an Intel CPU from ivybridge and in the last few years have been switching to ryzen for my family and friends due to thermals out of the box. Intel is too hot for subcontinent during summer

[@UFitzIt]

This is not just limited to 13th and 14th gen CPU's. I've been running an i9-12900KF on a ASUS Z690 MB for almost a year, no overclocking, out of the box BIOS settings, with no issues. Recently I started experiencing lock ups and blue screens. Any form of benchmarking ALWAYS froze and blue screened. After a lot of diagnostics I nailed it down the the MMX benchmark. After discussing with the shop I purchased the system through the guy suggested I try the Intel default profile and boom, fixed!. Haven't had an issue since.

[@nossy232323]

Degrading of the CPU's seems like at least part of the cause. Intel will try not to say much until their next gen CPU's are out. And then they will issue a patch nerfing the performance of the CPU.

[@WillFuI]

I heard a rumor that it’s an issue with how the ring connects to the cache. Apparently that changed a lot between 12th and 13th. But that was a ton of salt. But it’s also voltage related

[@Mcnooblet]

My 13900k was fine for a while. The BSOD started when putting the CPU under 100% load which pulled a lot of power. I think there likely is an issue with how much power it can pull at 100% load for long durations. In my case, I was doing repacks (20-30 minutes at 100%). I thought it was the RAM or XMP or something so I wasn't really paying attention to the CPU. I think the damage was being done at this point, as it got to the point where I couldn't get but 10% into decompressing the repack before a BSOD with the same circumstances trying to install a repack or get passed shader compilation of Forbidden West. Eventually even setting to 253w/253w/400a would result in a crash (brief second of power limit throttling when the crash occured). Same thing on shader compilation as that also puts the CPU at 100% load and would crash the game. Important to note, with limits in place (253w/253w/400a) the apps would crash. With unlimited power limits, I was getting BSOD. Eventually I could turn on IA/GT CEP (current excursion protection) which now limited my CPU to not being able to consume more than 200w at 100% load. Installed multiple repacks that were crashing like this, and could finally finish shader compilation of games like Horizon Forbidden West to actually play the game. Just did an RMA, and Intel sent me another 13900k but this time I decided to install a Thermal Right bracket which I have my own suspicions of now vs their stock bracket. My CPU doesn't even get hot anymore, and I mean like 20 C less under 100% loads. I remember with the old CPU seeing it pull as much power as 283w (and remember 400a is not to be exceeded says Intel) and it did in fact hit 100 C before on some cores, I think damage may be coming from them staying this hot or remaining there too long as well. Ironically, anything less than 400a on a 253w/253w would result in a EDP current throttle for me. So I can only assume when it went above 253w in the early days that it likely took the 400a above as well. New one is working well though, I put the 253w/253w/400a in place in which it hits at 100% load and I can tell it clearly wants more. But it's the only configuration that doesn't result in some kind of throttle, so I'm happy with it. I got my performance back in Bannerlord and can now run 1000 unit counts again without lag (was down to 600 with my old 13900k). One other thing, the 125w/253w/307a will result in an EDP current throttle under any kind of load, only idle will it not have a throttle. Also, Gigabytes new default settings on their BIOS f12d also results in a constant EDP current throttle. Sweet spot with no throttles is 253w/253w/400a for me, it may just be the 400a that is where it is suppose to be, but not above it, from my personal experience.

[@giglioflex]

20-30 minutes at 100% is not that long. I've been encoding videos for on my 7800X3D for a year and 1/2 straight and have zero crashes. This is with tuned memory too.

[@MarioAPN]

Fitgirl?

[@PAIN-ot4cj]

It’s clear that your cpu was hitting temperature tjmax and erroring you could have easily fixed everything by enabling a 90c limit

[@shodan6401]

I think that this is really insightful. It's the exact same thought that I had, and here's why: I used to be a QA Engineer at Xerox and Kodak, first with software, then firmware, and eventually end-to-end systems regression testing. Wendell said two things that made my ears perk up: He said that some logs recorded an M.2 error. And then he said that systems that employed four memory slots were represented much higher than those that only used two. And going way back to some of the first problems reported, systems were throwing a GPU memory error. This is all of your high speed IO. And that all points damningly to the ring, and voltage degradation of the ring and its inability to support the IO demands. He also said that sometimes the chip would just slow down for like, a minute before a crash, though not conclusive, this would coincide with the start of a communication breakdown (Led Zep) between the processor and its external components. So I think that this is correct. The ring is being burnt up as demand increases and asks for more voltage. If this is the failure in its architecture, perhaps there is a way to isolate and test for this, I don't know. I don't test CPUs nor am I aware of the tools available. But maybe someone smarter than me can do this. But as a former professional bloodhound, it seems like this is where the blood is....

[@Iscandelt]

I have a sample of 10pcs of 14900k systems runing SoC Simulation for 24/7, which I've made ~1month ago. I've already seen that this CPU is right on the edge of being stable and memory controler on two Machines couldn't ran 6600MHz 2R 2x48GB DIMMs. Just either plain crash or crash after a mintue or so in y-cruncher. I'm quite curious if/how quickly those would degrade. I undervolted each of those by around 60mVs.

[@raized943]

I give it 8 months until you have to add the -60mV back 😂 Just kidding, I wish you good luck man...

[@CuevadelRaton]

One of the friends i helped to get the 14900k stable got a pretty fast degradation, first was a shutdown every week, then days, then hours, and finally crashing into bios, the CPU is probably just dead at this point. When i helped him, told him to use a set voltage around 1.15 or 1.1v and adjust the frequency to try preserve the chip from more damage he opted for intel defaults fix and i don't blame him.

[@Eduardo1007]

The long silence after "what a massive fail" at 1:03:17

[@JonathanRLight]

Good rant. Voltages would be one difference, and continuous current draw would be another, higher on the higher end, where migration and degredation would be accelerated if there is a defect.

[@B-26354]

I've been AMD for the last decade, I'll be staying team red. Intel much like Nvidia have lost their way.

[@JJFX-]

What's crazy is we could assume Intel simply didn't have enough time to test longevity at these "send it" voltages but then they released 14th gen and made the problem even worse. As much as I'd like to assume AL-S had a lot more thought put into it, I'm skeptical about their priorities. I'm also suspicious about whether supposedly using TSMC for i7/i9 is really just about capacity.

[@Riyozsu]

Intel's priority is snake oil marketing and higher performance than ryzen even if it comes at cost of power consumption, thermals, silicon degradation.

[@tbreeze79]

what gets me is "PC World" has been slient. They are hardcore Intel fan boys. If this was AMD, they would do a whole podcast on it.

[@Riyozsu]

Intel bribes OEMs to not talk about this issue. If it was opposite, the entire prebuilt industry would complain to AMD.

[@astir0412]

Intel fan boys are dead quiet about it What does Jayztwocents say?

[@GTRPT]

I work in electronics design. My first thought was also that the higher voltages for 6+ GHz would be the cause. Intel/AMD are pushing the limits of what the manufacture processes can provide. If I had to take a guess, and oversimplify by a factor of 1 million, I would say they're using transistors flavored for lower than 1.5V and then feeding them 1.5V. Why would they do that ? Because the higher the voltage a transistor supports the thicker the channel and the slower it is. This is oversimplified of course and also I never worked with FinFETs but it also should apply (someone correct me if I'm wrong). I just find it weird this wasn't caught in the stress simulations of even in the voltage/temperature stress simulations of the actual silicon...

[@bobby_luo]

I now call this generation Degradation Lake

[@FatihSultanMehmed]

In my opinion, the last really good Intel CPU was the 11th (Rocket Lake/Tiger Lake) series. With the 10th (8 cores 16 threads) series we got twice as many CPU threads as with the 9th (8 cores 8 threads) series. With the 11th series the biggest innovation was the AVX-512 instruction set. With the PS3 emulator (RPCS3) we can see how well it works and makes AVX2 look old. What did the 12th series get in comparison? These downgraded efficient cores, but the AVX-512 instruction set is gone, what a fiasco and when I hear all the problems about the 13th and 14th series, which for me don't really have any new instruction sets either. AMD introduced the AVX-512 instruction set with Zen 4 and with Zen 5 the floating point pipe is doubled. Intel is in danger!

[@mickmoon6887]

Anyone with a brain knew CPU degradation had been real especially with silicon quality below bronze People wouldn't complain about degradation and frequency loss in the technical forums without any reason

[@fetus2280]

I knew about this and was Confirmed with my 4790k OC. I didnt need to Question that after .

[@BelowAmbient]

burn in yes, degradation, no

[@fteoOpty64]

Only with very high temps over 80 DegC......if you cool them to below 35 DegC, they could last for decades or longer...

[@fetus2280]

@@fteoOpty64 I had mine at 50c constant full load. I wish I had sub zero but water did well. lasted 4 years until i started to have problems. But i went in with the understanding I was on borrowed time. Now trying to get a replacement but i cant find one for a respectable price.

[@samiraperi467]

@@fteoOpty64 The current i9 K SKUs fail *below 70*. It's not the temperature unless the silicon is somehow bad enough to degrade at those temps, and it's not.

[@valentin3186]

Something tells me Arrow lake development includes the idea that led raptor refresh to fail such tremendously

[@DevilbyMoonlight]

I am so glad I went AMD this time around, the 1st time since 2001

[@jeroen5736]

I just went back to intel ;-/

[@DevouringKing]

My last Intel was in year 1999 :D

[@Psythik]

​@@DevouringKingYou stuck with AMD during the Bulldozer years? You are a much better than me. So glad Ryzen was a success. The late 2000s and most of the 2010s were some rough times if you were an AMD user.

[@DevouringKing]

@@Psythik Jeah i was hyped in September 2011 on Bulldozer release i got a Bulldozer on Day One. I keeped it until 2016 and then switched to a Vishera (Piledriver) for that 8% more IPC, same Mainboard, same Windows. This system runned from 2011 until Zen + 2018 for me, and then for my girlfriend. Runned 13 Years.

[@Trikipum]

I remember when cpus were basically inmortal and the part of a computer that rarely failed.. you bough it and it lasted forever or was screwed up from the start and you had to RMA, never seen a CPU failing in my life like this

[@LandonDean-zk4wn]

AMD had issues with Exploding CPUs. These things have happened before.

[@dirkjewitt5037]

I think Intel is fine with their I5's, they just need to fix I7 and up. At the least, if you bought an I7 and up, you should be financially compensated. They would have to first fix the price accordingly to the performance. If it's not getting close to the performance in the box, then they're ripping people off.

[@moonasha]

I have a 13th gen i5 and I'm pleased with it so far. If it doesn't degrade like the other ones, I'd say it's a very good value chip that seems to be severely underrated

[@dirkjewitt5037]

@@moonashaI know, it's the budget go to right now. Cheap, lots of cores, plays nice.

[@jeroen5736]

I hope you are right

[@Riyozsu]

This is generation issue. If the i5 uses the same architecture as i9 of 13th and 14th gen. It is going to face instability. Maybe not this week maybe next month. But definitely it's not free from it. I have read about 13700k and 14700k issues with recurring BSOD.

[@dirkjewitt5037]

@@Riyozsu yes, the I7's and up seem to be the issue. The I5's really don't seem to have any issues. I'd still get one of those. They also said that the 12th gen is doing just fine. Recalling the 13th and 14th gen I-7 and I-9's seems like the only logical step.

[@nempk1817]

Maybe they don't want to talk about it because of the zen 5 release, if they compare the zen 5 with an 10% weaker of 14 gen just makes zen 5 looks good

[@markmorrison3090]

What I had noticed after testing my i9 14900kf was that one of the two favoured cores was not stable at 6ghz ever after increasing the voltage to 1.55V. Since reducing the max clock speed to 5.9 GHz I haven't experienced any issues. I assume the degradation is due to trying to reach clock speeds that it can't run constantly.

[@sanchezmiki2006]

Framechasers did a video yesterday that the reason intel was having this issue was because 2 cores where trying to boost too high

[@timothygibney159]

Your CPU is WAAAAAY overvolted. That won't last long. I never would go over 1.35v on my cpus. Your voltatge is probably what fried that core. I would lower them significanly if you want it to work by fall.

[@Rroff2]

@@timothygibney159 Intel specs VCC upto 1.72V for the 14900 (actual internal voltage will be lower) - comparable nodes are 1.2-1.5V - I highly doubt the Intel process is that much better.

[@markmorrison3090]

@@timothygibney159 i doubt it but i only pushed it that far to see if it would work. I lowered the max voltage to around 1.501V after that

[@ThunderingRoar]

​@sanchezmiki2006 Framechasers is literally a meme tier channel theres no way people are taking whatever that guy says seriously

[@Slava4243]

It seems that oxidation MIGHT be the issue. Even 13700T, yes, the CPU with a base power of 35W. Things are getting more and more interesting. GN made a good video - "Intel Needs to Say Something: Oxidation Claims, New Microcode, & Benchmark Challenges" Any thoughts?

[@alberts4541]

My 14700k has a vid of only 1.289 Volts so not buying a i9 was probably a really good choice.

[@saricubra2867]

My 12700K is 1.17 volts. An old i7-4700MQ that i still have averages 1.05 volts.

[@denvernaicker8250]

@@saricubra2867 I have a 14700kf as well, and using intel extreme utility and undervolting and slightly overclocking early on had negative effects and restarts with games. I was wondering the impact of this news from Wendel and the impact for 14700 owners as well. Thanks for sharing there is some comfort in your comment.

[@saricubra2867]

@@denvernaicker8250 Don't use total overclock, stock turbo clocks are enough. If you want more perfomance, match the all core turbo clock to the single thread. NEVER increase voltages. With my i7-4700MQ i did that, if one core turbo clock is 3.4GHz, i set all core to 3.4GHz as well. Still stable many years latter with that partial OC, improved the frametimes on games.

[@erkinalp]

@@saricubra2867 don't do that on AMD processors though

[@darkkingastos4369]

The problem is the lower end cpus may still be degrading just at a slower rate. A few more months may show more of the lower end chips malfunctioning

[@alexmihai22]

People owning Intel CPUs don't wanna believe these CPUs are degrading.. but believing is one thing, while reality catches from behind. I wouldn't buy anything more than i5 13600(K) if going with Intel and I would consider no upgrade path from there.

[@Multimeter1]

I had the same issue with my 14900K day 1

[@TheNaitsyrk]

I can confirm that this is a thing. I had 13900KS. P CORE 124 E CORE 102. E cores can sort of do 4.9Ghz for bench. IlInitially P cores could do 6.1Ghz. I cant remember thr voltages i were running but it was probably LLC6 1.45V which is 1.33V real voltage as per CPU-Z or LLC7 1.35V +-. I was maintaining 1.33V whej running CB23 as per CPU-Z. If I am not wrong, it was running 1.43V at idle. After couplenof months i decided to OC again and 6.0 on P core was no longer possible (stably)

[@urmensch12]

At this stage of the intel instability game i almost understand nothing. But now i am too invested in the story to stop

[@XiaOmegaX]

Intel tried to make their toyota try to keep up with a lambo (AMD) and burned out the engine in doing so. Had they kept to the speed limit, they'd still be reliable for decades of use.

[@darreno1450]

I suspect Intel knew about the issue and released them anyway hoping it would go unnoticed long enough. Even if that wasn't the case, given the way they handled this situation, why would anyone buy another processor from them? That would be it for me.

[@ijustsawthat]

I've been waiting for this since this topic started appearing everywhere

[@AlaaSalehLE]

Nothing bothers me in this matter more than that intel isn't speaking about it. Like c'mon say something!!!

[@Hamentsios10]

You think i7-14700F with slight undervolting would be relatively safer?

[@AdmV0rl0n]

Truth will out at some point. The longer Intel 'silence running' this issue, the larger the scandal. Its going to be I think (because at this point, speculation is where we are at) some form of design or manufacture defect, and its going to run in the billions. A very large part of trust with intel is that they used to have a rep for the validation, and reliability to a good degree. Each day this progresses with no real response from them, is one day of more damage. And this reputational damage sticks. It doesn't wash off. Buyers of the next chip family are going to be extremely wary if they were bitten on this and had a deaf vendor. Some - and I think these are tip of iceberg info - info coming out has people stating just how bad things are, and how deep a hole its put them in. Imagine being a Dell, or a Lenovo, or HP, and staring down a barrell where your customer base has been sold bad processors that are steaming forward towards a class action suite, and the supplier is playing public laaa laaa laaa laaa. This is going into the folklore bin. Along with Dell Caps, IBM Deathstars, Nvidia substrates. The only question now is 'how big' its mushroom cloud is going to get.

[@OdinAlgeron]

this is user neglect 100% intel did what they were supposed to do: gave you a high spec machine people are not properly taking care of it you ALL know that dude, the one that bought a sports car and NEVER changes the oil or monitors coolant levels and runs it balls to the wall 24/7 and then wonder why its constantly breaking down

[@mduckernz]

@@OdinAlgeronBruh, this is happening even in servers, on server motherboards. It’s not the users.

[@OdinAlgeron]

@@mduckernz did you not read what i stated? whos fault is it when the oil and coolant is never changed and the engine blows?

[@MichaelDeHaven]

@@OdinAlgeron That makes no sense. Wendell's data is from server farms on server motherboards. You could argue that they don't know what they're doing, I guess. But then why is it just happening now? Do you honestly think people suddenly got dumb? Why are they then charging more for Intel and even some switching to AMD? Isn't it more reasonable to suggest there's an issue with the chips?

[@Riyozsu]

​@@OdinAlgeronnot everyone is techno god and can tune each and every atom of silicon on the chip. Intel created a bad generation. Just accept it. It fails if you use minimum clocks, it fails when you overclock it, it fails if you leave it idle.

[@DLMyth2]

If I would of paid attention before building my 14900kf system I would of never of gotten this crappy CPU. From the start I had a lot of stability issues & high temps. The only way I was able to "fix it" a bit is by updating to the latest bios and making sure Intel default is on. I also had to set PL1 & 2 to 210W because there's no way I could run 253W without my system crashing every 30 mins to 1 hour. I still had stability issues with my games like Warframe crashing on me for no reason. So I had to lower the settings from 57x to 55X and this somewhat tamed my issues for now. Given I had to gimp the CPU settings which I find Ludacris because you spend so much money and you need to nerf the settings. I lost a lot of respect for Intel after this and if they don't get their stuff together I'll be going back to AMD. If the next CPU lineup is stable and better I may swap my MB & junk 14900kf to the new CPU line up. But confidence wise I doubt Intel will fix their issue for a couple of more CPU lineups.

[@dionysus6081]

My choice of cheaping out and getting a 13700kf instead of a 13900k seems like a great choice in hind sight, I've had 0 stability issues. I have a -95 mV undervot.

[@Horendus123]

Also in this boat

[@N4CR]

THey are on the chart and also fail.

[@Prophet650]

Great points. Massively distracted by the crazy amount of times you kept saying "like"

[@10100rsn]

The main problem is that the failures are not all identical. Each CPU has been degrading in similar but slightly different ways. The question is, has anyone investigated or poked around on the CPU interposer? Checked all the tiny capacitors and resistors on the interposer board and tested if they are actually connected to something like they should be? It could be a design error on the interposer and not the silicon itself.

[@ThunderingRoar]

If that was the case the cpus would be showing issues on day1 and not after 6months like what Wendell talked about

[@10100rsn]

@@ThunderingRoar Not necessarily. It could be wrong components that take time to fail in circuit. Or It could be something as simple as cracked solder joints because of insufficient solder, wrong solder or solder that has impurities. It could even be that the solder profile is wrong during assembly and the solder paste isn't reaching the melting temperature for long enough. Perhaps glue is being used to hold the components before soldering and it is getting under some of the pads during the solder process making an incomplete connection on the pads. Or a combination of the above. Anyone with one of the effected CPUs and a $40 microscope can inspect it close up and maybe do some tests with a proper multimeter. Maybe even go over every component with some solder paste and a heat gun as a test. Anyone try it? Nope.

[@ThunderingRoar]

@10100rsn Wouldnt all of those things also apply to i5 cpus and we dont see them having issues?

[@10100rsn]

@@ThunderingRoar Not necessarily.

[@razorblade7108]

​​@@10100rsnIt's highly unlikely because reducing clockspeeds seems to mostly resolve the issues, which points at silicon degradation. People who ran stock clocks at lower voltages had to just increase their voltages every now and then to fix stability issues until they eventually became too unstable to run at all, which points at silicon degradation.

[@arrow207]

P = I * V though... raising V will lower current for a given power demand. Do we know for certain that voltage is causing degradation more than current? Perhaps lower voltage is driving a higher current which in turn is causing problems.

[@Azureskies01]

Intel couldn't stand the fact that AMD beat them in games with the X3D chips (most of the time). They would NEVER allow AMD to possibly ship a CPU that can boost to 6ghz before them. Also their IPC would probably be around the same as AMD's non-X3D chips in most scenarios if they clocked the CPU to 5.4 or 5.5. Long story short, intel doesn't know how to be number 2, and it is going to get them sued.

[@brunodinis7454]

frequency is irrelevant for the IPC... Intructions per CLOCK...

[@Azureskies01]

@@brunodinis7454 What effects ipc? could it have anything to do with frequency? frequency of the Ring bus for instance or interconnect between cache and cores? what sometimes dictates those thing? Just don't post if you have zero idea of what you want to talk about.

[@astroidexadam5976]

except AMD's X3D cpus were clocked slightly lower than their stock non X3D counterparts

[@stevewatson6839]

Not wasting my time, mate. You can off on one as often as you like for as long as you like; it is both informative and hugely entertaining.

[@waynetuttle6872]

The reason for the lower LLC is because Intel is trying to run the clock out till 15th gen, they think it will get swept under the rug if they can just make it to 15th gen

[@waynetuttle6872]

And, the ecores aren’t degrading but if the power rail is failing then the ecores can be sucking down power that the pcores could use

[@waynetuttle6872]

But it isn’t power, it isn’t voltages causing the degradation or game servers being pampered wouldn’t be dying

[@waynetuttle6872]

My theory is that they keep making them thinner and thinner and this has caused exposure. They keep honing and honing these binned cores to the point of craziness. They started this with 11th gen, but now it is nuts. They are leaving thickness out of the manufacturing process, maybe this is leaving circuits exposed to heat, especially the power ring

[@rluker5344]

@@waynetuttle6872 Then it isn't CPU silicon degradation. It must be something else that is failing. If these continuous multicore loaded, 150w capped game servers running probably around 1v are failing at much higher rates than the general public running these chips on Z790 boards with intermittent single core loads that are often hitting 1.4v-1.5v that points towards the cause for silicon degradation not being the cause for these failures. The other significant difference is using w680 motherboards that have, by typical i9 usage standards, utterly pathetic VRM systems. If there is a difference in stability percentages with a difference in hardware, it might be the difference that is the cause of the instability.

[@waynetuttle6872]

@@rluker5344 possibly, but I doubt it. Take any other architecture as a standard, aside from 11th gen which also has failure issues due to thinning of the substrate. Given enough time, despite the vrm or chipset, every last 13th or 14th may fail under ultra low voltages. Some of the server chips are dying day one, this is clearly a production issue when there are currently 4790k chips still chugging away at 5ghz from 2014 with zero issue going back to stock settings. There are 3600-3950x chips that have been gut overclocked and super high voltages since release that will function perfectly at stock settings and run stock voltages and clocks with zero issues. Half of all server chips are literally useless from one to thirty days of use.

[@sirius4k]

These were not datacenter CPUs. They were 13900K/14900K CPUs in barebone datacenter-ish motherboards. Running at stock speeds and massively underclocked memory, which still failed.

[@Bound4Earth]

Amd isn't on par with instructions per clock, but Intel isn't stable so Intel isn't on par with themselves.

[@mariusoneo]

I have 14700kf for a year now.After stock testing and high temp form day one i undervolted and underclocked 5,3ghz -0.100V and not a single bsod ycruncher stable occt etc.I have noctua u14S and max temp 79C in ycruncher.So goodluck with stock settings,intel pushed to hard frequency and voltage. Only fix is underclocked from day one and undervolting.Some cpu are more resilient so degradation will apear later,but no one will escape without reduced clock.

[@dethskullcrusher]

Bro what's your highest vcore? The problem is the extremes, i might be the case that even 1.4v could be just too much voltage for the ring of the cpu.

[@mariusoneo]

@@dethskullcrusher with undervolt and 5.3ghz is 1.24V maxim low load,all core is 1.15V

[@xmlthegreat]

I think you finally reached the truth of the matter towards the end, Intel have been quiet because they don't want to tank their stock prices.