[SPON: Use "brokensilicon“ at CDKeyOffer to get Win 11 Pro for $23: www.cdkeyoffer.com/cko/Moore11 Win 10 Pro ($17): www.cdkeyoffer.com/cko/Moore10 Win10 Home ($15): www.cdkeyoffer.com/cko/Moore10h Office 2016 ($29): www.cdkeyoffer.com/cko/Moore16 Office 2019 ($48): www.cdkeyoffer.com/cko/Moore19 25% software discount code: brokensilicon 3% OFF EVERYTHING discount code: dieshrink ]
Support you with pirated software that could get your video a Community Guidelines strike? It's already happened to another RU-vidr who had to edit out their sponsor.
AMD has learned from the 7900xtx issue sure but they're seeing what's happening with intel just burying their head in the sand. They don't want that bad press when it can just be avoided entirely by delaying the launch while they recheck the CPUs (That's what AMD are saying they'll do). They also don't need to rush Zen 5 out the door. What the hell has iNtel got to challenge them with right now? Zero products can compete with it right now. New users are not going to gamble and buy intel 13th/14th gen at the moment unless they don't follow news at all.
A little note if I may: I do not, but if I did own an effected chip that was not causing me enough grief to want to outright use a different system, I would wait probably 3-months and see what people say, what reports like in this video says about those updates. "If" there is no real net loss and no new problems then see what others say about being able to roll back your BIOS, if that is not possible that is alarm bells, if it is then proceed with the updates after a period of time and 3-months is the blink of an eye in the age of a motherboard during which there will be very few BIOS updates to do with CPU security and they always hit performance, the rest is all to to with the motherboard or adding support for a new line of CPU's, this kind of thing hardly ever happens, and that alone freaks people out.
*TOM !! Question: What processor are you running, and if it is a 13th or 14th gen, what are you doing to go about the detrimental effects it could have on your processor (if you ARE using a 13th or 14th gen)?*
Shoutout to Buildzoid of Actually Hardcore Overclocking for being 100% spot on with his speculation about it being a ringbus failure due to the shared rail. He basically reasoned and came to this exact conclusion as soon as Wendell released his video on the subject.
Much respect to Buildzoid, but anyway we need to wait for some independent confirmation or a presentation of the facts from Intel that at least looks plausible.
I mean 1.5v doesn't help... but yes the ring bus instability is absolutely a thing. It affects my 13900k and I can't rma the thing. It'll pass any stress test you throw at it but still weird intermittent and most infuriatingly SILENT errors. Even tried grabbing a spare mobo to test it on (massive pain to swap) but didn't really change anything. alderlake 12400 chip and no issues that I can detect after 2 weeks of usage.
AMD Athlon 1400MHz (0.18um) is x200+ times slower than Intel 14900K. Athlon 1400 had only 37M transistors. They were really simple in comparison to modern CPUs.
@@Fataha22 Ain't no way any dumbass is actually trying to lie that AMD is stable in literally anything they do lol. It's like going from a fresh dog shit of intel to a nice bucket of collected dog shit in amd. People already forget about all the issues of 7000 launch. How fast people are to forgive and forget in the tech world. Then again the second 15th gen intel is out and 14th gen is old news people won't remember this mess either. No wonder these companies get away with murder.
@@alexanderbelov6892I'm confused as to why you picked an 1400mhz Athlon. Even the 12th gen was reliable, and the worst recent issues with the 7950X was a single motherboard having a bug
@@alexanderbelov6892 I had a Phenom II X6 OC'd at like 180w for far too long, and it's still running in the parents computer (at default clocks), Plenty of people will comment about how their Sandy or Ivy Bridge is still holding strong, I would gamble that a modern Core 12th gen or any Zen 4 (perhaps excluding the X3D chips) will last ages too. They really shouldn't have any issues within the voltage spec of the silicon, degrading happens due to excessive voltage physically moving atoms around, there might be some increase in fragility as we get smaller and smaller transistors though. It really does feel like Intel just got cocky, rushed to market, doubled core count, increased clocks by 400MHz, It's somewhat impressive they could get 6.2GHz out of that the older 10nm process, when AMD on TSMC 5nm only does 5.7 or so.
RIP network engineers. My first big network job at a bank was in 2015 and it was replacing all routers/switches at 550+ locations. As soon as we got them all replaced, we had to do it again, because the Intel Atom CPU's had a hardware bug that made the units worthless after about a year of uptime. So we had to do it all over again with first-in-first-out priority. There is a ton of logistics behind getting that much hardware at that many locations swapped out. Cisco AND Intel ponied up cash for that FUBAR.
Intel for decades: "don't push VID beyond about 1.43V without exocitc cooling" Intel the past 2 years: "nah, 1.55V is a typical operating voltage" - bruuh
The actual Intel spec is 1.72 Volts which is insane. Without liquid nitrogen, not a single CPU known to man can go to 1.72 Volts. The Intel spec has been this way since 8th gen.
Also 90% of 13th/14th gen cpus using default bios because they don't update which had the chips running at 1.3v "why is my cpu dying" The root cause is not voltage champ. god help everyone believing this bs when they have been using an undervolted cpu by default since day 1 they just don't know it because they never checked the bios or any monitoring software to see their voltage. There were legit more claims or degrading cpus in older bios that were running around 1.3v than there have been in newer "intel default settings bios" which run the cpus at 1.5v "voltage is the cause guys that's why less cpus are breaking at higher voltages and why cpus still break even at extremely low voltages trust us guys not gaslighting you to delay for 15th gen please eat it all up" I need to start selling magic beans to everyone who believes this stuff I am going to make a fortune.
This is playing out just like the FDIV bug that forced Intel into a recall after a disastrous response to a catastrophic architecture design flaw in 1994.
Exactly! I think it was Andy Grove's arrogance that caused so much harm with the FDIV bug. Intel is even more arrogant today and is no where near the top of their game they were back then. Sad to see such an iconic company go to hell.
except back then they basically had no competition. This issue will blow over into OEM's and they will reduce their risk with blue chips in the future.
disastrous... given where intel has been for 30 years, i dont think you know what that word means. Disastrous would be Cyrix and its issues with gaming.
No, it's playing out _exactly_ like the "Sudden Northwood Death Syndrome" (S.N.D.S). Go on, look it up, this very very similar, except now, Intel themselves are the cause of the rampant disastrous overvolting and degradation of chips over time due to electromigration.
So basically the intel engineers knew about the problems but probably told the higher ups : yeah it'll work but dont clock it too high cause you'll need too much voltage and the chips will go kaput. Intel highers ups : it'll be fine! Those engineers are weaklings and scaremongering! I've worked in billion dollars corporations long enough to know that the bosses most of the time dont listen to anything that theyre being told by the people under them that actually know WTF they are talking about... I know I've worked 2 levels under the top of a 30 billion $ corporation for a decade and the number of ``smart`` people at the top that never ever ever did any job except being in suits and going golfing is quite something.
The suits are almost always r*****ed they only know how to look at financial charts. Tech companies need engineers to lead them like with Lisa Su in AMD.
@@hellcoreproductions The problem is that Gelsinger is trying to rapidly make Intel leapfrog AMD within a few years. Intel pretty much stopped properly innovating for like a decade, and now Gelsinger is trying to save his company and reestablish dominance in like 5 years. It's still quite surprising that this seems to affect only Raptor Lake and not any of Intel's other new architectures, but CPUs are extremely complex and intricate, likely one of the hardest things to ever design. So trying to cram 10 years of development into a couple of years is likely never going to end well, as proved in this case.
Executives don't make those sorts of decisions, engineers do. The leadership follows the advice of their technical people, that is what they are paid for after all. As far as the executives are concerned, they want a profit, and a product that generates large numbers of RMAs destroys profit margins, so they are not going to push for that despite what popular opinion might want you to believe.
@@BobMar1964 no ring bus is a part of CPU it's meant to connect cores to each other and their caches Memory heavy workloads stresses ringbus because there is more data going around
@@arianamirgholami9555 For some reason I was thinking 13-14 gen was a Mesh connect. But that makes sense cuz my 10900X dont have the latency issues on a Mesh architecture that these 13-14 gen parts. Guess they are cooking the chips or there is some deep physical issue with these. Shutting the fab down until the HVAC was repaired should have happened.
I almost feel sorry for Intel, but ultimately I don't because they sold the same quad core for 10 years and then when ryzen started getting good they decided to do 10 years of r&d in just a few years to try to catch up. It obviously didn't work. Reaping what you sow, etc, etc.
@@calvitocalvon1711 Way to miss the point. It's not about core count. Intel's r&d stagnated for 10 years and then they tried to catch up too fast and now they're paying the price.
People keep pointing to the Pentium FDIV issues from 20 years ago, but this honestly seems much worse and encountered by users far more often. The FDIV bug was there in the hardware, but the odds of encountering it were miniscule. Intel needs to get out in front of this fast because it's cheaper to replace the CPUs than it is to repair company reputation.
The problem is Intel didn't get in front of it fast enough. People have been on the record talking about these problems for a year and nothing happened until now when it was exasperated. Even now, their response is pretty lackluster as they're making it difficult for people with chips they received from prebuilts to get new chips (you basically have to go back through the company that built your PC and then they have to get you a new CPU from Intel and that's if Intel will actually do it since the chips didn't come direct from them).
@@originalscreenname44 Not exactly thins problem...hybrid architecture was a mess to begin with, premature failure due to ring bus being overloaded on high clocks is just icing on a cake. Tho if this is the reason, new intel lineup (e-chips without e cores) should solve the issue. Benchmark scores and productivity results will take a nosedive however...
The real issue is that Intel's decision to do a Socket every two generations means they cannot offer a bugfixed replacement without pulling down massively multicore Xeon designs (that don't use a ring bus) and releasing low core count versions of them. That's extremely expensive, as it's gotta be a new design to use a smaller die or it will be unprofitable as hell, and it still will get you lower performance than Raptor as the grid bus design has higher latency below 24 cores than the ring bus, meaning they will still face lawsuits for offering a worse product as a replacement. I'm not even sure if the grid bus is separately voltage controlled and if it isn't it will need a redesign or the cores won't be able to reach Raptor clocks without burning out the bus again. They are stuck without a way out, and the company will go bankrupt if they don't sell current wafers too. Intel are well and truly screwed when this gets to court. They are just playing for time now so that they have a new working product out when that happens and hopefully can absorb some of the losses with the profits from that.
I've been telling people on Quora for months after doing my own testing, but the army of people that believe Intel can do no wrong and it's mobo manufacturers fault for overvolting etc etc is unbelievable. I genuinely believe some of them are taking money under the table to post FUD and protect Intel from lawsuits.
Oh, the greed. Intel knew that the architecture was flawed and would likely (and did) disintegrate when higher voltages were used. All this to compete with AMD who simply continue to do their own thing by achieving higher performance using half the power and consequently less heat and no chip degradation. And Intel kept this knowledge hidden from the public and it's business partners who are supplied these faulty cpu parts. So now the worm has turned. For years AMD struggled as a result of Intel's meddling by bribing the likes of HP, Dell, Lenovo, Asus, Acer and others to not use any of AMD's offerings. They made billions upon billions of dollars and when they were finally caught and found guilty by different courts, were fined and forced to pay AMD a token amount as restitution. Now it is AMD's turn to finally be allowed to compete in an open market which outed Intel and where business is conducted legally and properly. Yes, the worm has most definitely turned. I have no sympathy towards Intel and finally that Karma train has come for them.
Weren't it for AMD, there would be no Intel today. Weren't it for Intel, there would be no AMD today. This is merely a tirade of "give and take" in the x86 world.
If it is indeed a ring bus failure then Intel needs to recall the fucking CPU. To do otherwise is license for a class action. 13/14th gen are turning out to be Lawsuit Lake.
Class action lawsuit isn't really a thread, they'll have to pay pocket change to US customers only, and even then only people that claim it. It will probably be $20 compensation or something.
I'm looking forward to Intel's partners suing the company for industrial sabotage, breach of contract and fraud. Not that it will happen but it'd make for a nice chapter in a novel.
Intel owes every Raptor Lake buyer an LGA 1700 CPU of equal performance to the one(s) we bought that works reliably in the motherboards be bought. Or an equivalent next generation CPU and the motherboard to run it. Anything short of that earns Intel a class action. Telling us we can't have what we paid for because we have to nerf it to keep it working is simply not acceptable.
No, they only thing they owe you is a replacement CPU for three years if yours fails in normal use and you have not overclocked it. Spoiler alert: every Z class board overclocks CPUs out of the box 😂 So no, they don't owe you anything. Take it up with MB vendors. If Intel sends you a new CPU it is because of goodwill.
@@hellcoreproductions is right though they sold a cpu in an ecosystem that is dead and it still isnt what they said it would be. Either prolong 1700 or the class action could be for financial damages too.
@@Tugela60 That is some first class Intel fanboyism you've got on display. I'm impressed. You'd be a perfect fit for Intel upper management where you would proudly proclaim that the customer isn't important and instead it's only about the bottom line and shareholders. Back to reality: the actual real bottom line in the real world is that people who buy an Intel cpu deserve a stable cpu that doesn't unalive itself at stock settings.
@@grigorioschristodoulou5229 I remember when they swapped from the liquid metal to a cheap TIM to stop people clocking quads to the moon (2600k the GOAT)🗿🗿
So its a ring bus overvoltage failure... wow. This is such an important video! The micro code update will only prevent future damage, not erase the damage that's already done to the chip! I think for 13th gen owners we just need a path forward, if we're not ready to go completely AMD, maybe throw away the chip, and buy 12th gen.
Of course you can't "fix" damage... I've seen this a few times and I don't get why people would think otherwise. If your chip is damaged then return it under warranty. You damage your chip every second it is turned on and applying voltage... It just should happen over a long time. But when the combination of microcode and bad motherboard bios development and apply extremely high voltages things can be damaged quickly. Supply to little and you also create stability issues.
@@JustSomeGuy009 Problem. If you bought directly from Intel, they've been denying warranty claims. If you bought from HP, Dell, etc... Many warranties are only for a year, and this issue has been going on longer than that. This is actually a lemon law issue at this point.
In what world would a microcode fix a chip that's already been damaged? This is how it's going to work. If you have a chip that doesn't suffer from the degradation then you're good. If you have a chip that's brand new and you apply the microcode before using it, then you're good. If you're one of the people with a damaged chip then you exchange it. It's not rocket science.
@@alexcardosa8079 This isn't even slightly true, the largest source of reliable data for this are server providers who are running the most conservative settings.
The ring bus design wasn't meant to scale above a dozen of cores. With Haswell and Broadwell all 12+ core CPUs had to implement dual ring bus bridged together to avoid current overload on top of the signal latency issues. That's why Intel moved away from the ring bus interconnect with Skylake server CPUs.
Well, an 8+16 RPL is the same size as a 12 P-core Bartlett Lake. It sounds like the size is less an issue than the voltage being high because it's shared with factory-overclocked cores. If they split the ring into two but kept it at >1.5V it'd still get fried.
I'm far from an Intel apologist, but comparing 32nm process to sub 3nm process is apples and oranges. Electromigration and the associated issues were theoretical back then, very real today.
think about how many people have gone through RMA hell / been denied until this became mainstream and they couldnt hide it anymore. man.. they really did their loyal fan base dirty. a dwindling group...
I think they tried to stay relevant at the time with AMD, and with all those so-called "geniuses," working up there at Intel said ship it. They've been in this game way too long and the equipment we have now days, knowing good and well, they were shipping a product that in no way was passing tests. Almost reminds me of when VW sold vehicles that had code in the PCM's to lie to the emissions test computer about pollution. Don't think for one minute any business is squeaky clean. If they recall 13th and 14th and replace them with Bartlett Lake-S that would be a step in the right direction to possibly hold onto some customers. RMA and get the same defective product defeats the purpose. If they were a private company probably wouldn't of happened. But they have shareholders and got to keep that revenue coming in by any means necessary, right or wrong (apparently obvious now)
I think even the laptop chips are degrading very fast. I have a Dell with 13850HX and starting to see random glitches where everything freezes for about 10s and audio glitch, then back to normal. Laptop only 6 months old.
Intel said laptops aren't affected and the only data we got was that warframe error report which showed no laptop chips. So I don't see how anyone can claim laptops are affected when we have no evidence...
@@WaterspoutsOfTheDeep I think my own experience and the experience of others is sufficient. My 13850HX laptop started off with random glitches but now crashes randomly every few days (the entire system hangs and the caps lock LED indicator starts blinking which is an indicator of hardware fault, not an OS issue according to the laptop documentation).
@@WaterspoutsOfTheDeepAlso the same laptop has some random compression failures. If I run 7z in benchmark mode I eventually get nonsense errors given enough time (roughly 10 minutes) which is an indicator of a problem with the CPU. If Intel genuinely says that mobile CPUs aren't affected, I believe they're not being honest.
@@WaterspoutsOfTheDeepoften the laptop chip is made in a different fab line. This fab line may not be affected. So that tells me they do know more than they're saying if they can confidently say mobile isn't affected. They know root cause and they found out too late in testing and didn't tell anyone. They just waited for it to look like a random defect. We now know different and it's big especially for stockholders.
Fuck Intel! I was a lifelong customer who never bought an AMD processor in my life and I really regret buying a 14900K for my current rig that has been crashing since day one. I just didn't know it was because of the CPU. I will RMA and be done with them.
Tom, you need to get TechYesCity back on for another interview to discuss this Intel issue. He's been talking about issues with the Ring Bus adding latency on 13th & 14th Gen system for over a year. In fact, I think he said he downgraded to a 10th Gen Intel processor because it was just snappier!
What if the big 14-stop ringbus on 13/14th gen on his CPU was having errors from the get go. It was correcting them, but adding random delays every time it needed to. And what if, as the error rates go up as they degrade, the correction logic cannot keep up and let broken data through? Some golden sample won't have any issues (for quite some time anyway), but there may be CPUs with a lot of corrected errors out of the box, only to get worse overtime.
95% of casual home user will never learn about the Intel CPU degredation issue on their own, let alone update their BIOS+microcode in consequence. So Raptor Lake CPUs will continue to die in masses in the coming months and years.
Yup, same here... I mean, not like I'm doing it willingly to prove of a point or something... I just don't like to spend money when what I have kind of satisfy my needs already...
My plain ol 3770 still rockin at 4.3GHz, 2700k still hits 5Ghz they were built to last, I prefered when we had to overclock them ourselves instead of buying these pushed to the limit CPUs
There is no 14th gen, they have lower voltages because it is a more mature 13th gen silicon. No microcode can fix toasted silicon, lowering the voltage will make damaged chips more unstable
On damaged cpu you have to reduce both voltage and clocks by like 200-300mhz, then it will be stable again. But thats not the advertised product spec anymore...
Unfortunate yet unsurprising as Intel has been behind in IPC for some time; echos of the Pentium 4 shenanigans when Intel had to push clocks to compete. Thank you for the update.
Zen 4 clocks almost as fast as Intel 13th gen, to compete with 13th gen and right now Intel is only slightly behind AMD in single threaded IPC. AMD is better for heavily threaded APs, but this is only at the top end of the product stack. When you get into the range of 13700K Intel does better most the time because AMD has nothing with the same thread count, and when you drop down to the lower end same thing. When Zen 3 came out AMD had an IPC advantage, but Intel basically caught up and Zen 4 had to rely on faster clocks to compete, and higher power consumption, so both companies did the same thing dude. Intel's platform is better too. AMD has to use two chipsets when the CPU-chipset link is ONLY 8GB/s, SAME AS IT WAS for X570. Intel runs that link at 16GB/s and only needs one chipset. And no I'm not an Intel fanboy, I haven't bought an Intel based system in over 25 years. You just have the facts mixed up.
@@johndoh5182You are probably interacting with a member of reds marketing team and nothing more. After Raptor Lake killed AMD on 1% lows, Red has been swirling the drain. The 7900XTX did not compete with the 4090, literally at all. Nobody wants a 7900XTX. The 7850XD is put of sight for most integrators and builders. AMD was honestly going to really start to catch Intels heals but… Blue played them again and here we are. The company is screwed at the moment so they’re playing dirty on social media.
@@johndoh5182Intel fanboy vibes….. 😂 I use Intel myself but some of AMD’s chips are better. The next gen for AMD might leave Intel in the dust for gaming at the very least. We will see how all this plays out.
Speaking as a long time fan of AMD products (not necessarily the company) - supporting them even when they didn't have the best stuff out there - we shouldn't just assume that they'll do the right thing for their customers (a.k.a. us). However, we should make sure to let them know how much we DO appreciate when they're open and transparent with us. As for Intel, it's not the first time they've hid things like this, and I doubt it will be the last. It's just one of many reasons I choose to not purchase their products.
Makes sense being a ring bus failure. That's why some CPUs come back when you disable the E-Cores. And the general behavior seemed like it was an uncore problem, too.
The amount of people that will do a BIOS update is going to be marginal. Thinking about the general consumer population, barely anyone knows that something like this exists at all, then it's not easy to a typical consumer, then there's some risk involved that will turn some of them away. Those CPUs are going to cook real fast.
@forgingfire2850 only one developer. I can't find any data on what laptop cpu's are crashing. No reddit post, nothing. If it is happening, then it's not nearly as prevalent seemingly.
@@forgingfire2850 one game dev said they tested a few laptops and saw issues and gave no evidence. One persons word that we may not even be able to trust. The error report data from warframe showed no laptop cpu's, intel said laptops not affected. I just haven't seen any evidence laptops are affected, aside from maybe 13th gen from oxidation perhaps.
12:07 I feel personally called-out. I was running an overclocked E5-1680V2 @ 4.5GHz for quite a while (and playing most of my games at 4K), then upgraded to a 7800X3D. I went from an 8-core Ivy-Bridge @ 4.5GHz to an 8-core Zen4 @ 5GHz. DDR3 straight to DDR5. Year 2013 straight to 2023. Say what you will, that was an amazing architecture. I honestly wonder what that CPU design would be like today if it were directly ported to something like TSMC N3.
It would be much better on the current nodes, but not as performant as the current architectures, since they evolved a lot as well. One example would be much more efficient branch prediction, which is super sophisticated these days when compared to, say, a simple one-byte predictor that remembers only the last branch decision.
This is the same management that let Raja Kaduri run their graphics unit out of control while developing Alchemist too. I get the feeling that there's a lot of buddy-buddy group-think the Intel C-Suite.
"Intel 7" 300mm wafer costs them as much as Model X?! That car starts at $80k. TSMC sells its leading edge stuff for ~$20-25k. I just find it hard to believe it's *that* bad for Intel.
Also improving the longevity = Can we make the chip survive long enough that it's initial warranty expires. If Intel wanted to do the right thing they could come out with an extended warranty for 13th and 14th gen chips starting the minute the microcode is released. There needs to be a class action lawsuit and a full on recall of these chips.
Intel has been making CPUs with a ringbus since around 2011 (Sandy Bridge). L3/Ring has also been a separate voltage domain throughout most of the Core i series. They should have known. Actually, they must have known. This is ridiculous!
Considering that they dont use the ringbus on their highest end (HEDT and high core count xeons)...they surely knew that, and how to avoid that problem.
The engineers knew. Some business executive ordered them to do it this way to try and beat AMD. He knew the risk/guarantee, and bit the bullet. Now he's gone from Intel and the engineers are left saying "told you so" to the board that trusted this guy to do a good job.
@@alexturnbackthearmy1907actually, the reason they don’t use a ring bus on the last few HEDTs and as of late _ALL_ Xeons, is the same reason that AMD doesn’t use monolithic on their high end parts: it just doesn’t scale. The E-core clusters share one ring bus connection for each cluster of four cores, and so Intel has only scaled it from 4 cores on a ringbus with 7th gen to 6 cores with 8th gen, 8 cores with 9th gen, 10 cores with 10th gen and effectively 10 ‘cores’ with 12th gen, and now 12 “cores” with 13th gen. Whereas Intel has 56 core Xeons and is allegedly planning to offer 128 (performance) core and 288 (efficiency) core CPUs in servers this generation. And the reason that they don’t use the mesh in desktop CPUs is it’s terrible for desktop. It’s only “acceptable” for servers, because it’s trade-offs of lower performance for desktop workloads don’t typically apply to servers.
The ring issue makes sense if you account for the 13600 serie being less affected then 700 and 900 serie: less MHz, less P-Cores, Less E-Core, pushing electrons through that ring.
This actually matches very well with what Intel said in a PR statement (part of it below): "We have also looked at it from the instability reports on Intel Core 13th Gen desktop processors and the analysis to-date has determined that only a small number of instability reports can be connected to the manufacturing issue. For the Instability issue, we are delivering a microcode patch which addresses exposure to elevated voltages which is a key element of the Instability issue. We are currently validating the microcode patch to ensure the instability issues for 13th/14th Gen are addressed." Notes: 1) Intel doesn't say they have _fixed_ the issue, just that they have "addressed" it. 2) Oxidation, according to them, is not a major part in the issue. 3) The over-voltage isn't the root cause of the problem, it is "a key element". So to me, it seems that flawed design is the root cause, oxidation made it worse and microcode update mitigates the isssue by lowering voltages. All fits to what you said on the video and what Intel has stated. They are just tiptoing around the communication to not directly lie, but still to make it seem as if they actually fixed the issue that pretty much was just a bit too high voltage. So I think you're right, once again.
The casual player part is imo the most concerning to me imo. In the last few weeks i saw many people complain about the memory issues in three different games and most people simply didnt knew about those issues. And even less will be able to do a BIOS update. (Lets be honest here) Not to even mention the people that buy Complete PCs. 90% of all players with 13th/14th Gen CPUS will not know whats going on or be able to update anything. And the biggest problem that i see here... AMDs prices will probably go up significantly. They are just waiting to see HOW BAD the Intel situation gets and find out by how much they can push those numbers. And to be honest... as someone, who probably has to upgrade the whole shtick of CPU/Mobo/RAM in the foreseeable future, due to games with all that UE5 Nanite/Lumen crap sucking all life out of my CPU (12400F), i dont see myself going for Intel either. Thats a really shitty situation. Not just for the people with the problems, but also for the competition in the following years.
“Your game is crashing my $3500 gaming PC with ‘out of memory’” - And after several ppl hint at the Raptor Lake issue, thread starter would still not disclose which CPU model they have. They did update BIOS in the end though and their system no longer froze up while crashing the game.🤘🏼
13:30 Windows automatically updates microcode, but only in Windows, so every time a user starts the system it's going off the old microcode that's killing these chips, most laptops can update the bios through windows update, but DIY/system integrator desktops are a problem since those can't update bios through Windows update
Buildzoid at Actually Hardcore Over Clocking predicted it was the ring bus getting burned up weeks ago. It was his best guess but it's interesting that he got that detail right.
I undervolted from day 1 and my 13900k chip still failed I blamed myself, but to find that out later 2nd chip is doing alright after limiting to 262 watts, 6 months later is still good, fingers crossed
@@MooresLawIsDead Mine is not boosting, it's fixed at 5.5 on P-Cores and 4.3 on E-Cores, with Intel Speed Shift and Intel Speed Step turned off. 5.8 ghz boost is turned off. So, it stays fixed at 5.5 and 4.3 always, even when doing nothing. Meaning, that even in my case with undervolt and cores limited, it still failed. 300 watts limit that I set originally could be a factor, but not in classical sense. That 47 watts was still PL limited to 300 watts, and voltages did not jump around on V-Core, so where did the wiggle room of 47 watts go? Memory Controller? Ring bus? It wasn't going to V-Core. Chip failed even at 1.28 volts that I set to Load Line Level 6, meaning I am in range of 1.28-1.35volts at all times, with an occasion of 1.37 burst at load (not more). Chip still failed. What does it tell you? It tells us that the boosting exuberates the issue, but is not the lead cause as in my case I am limited across the board. Voltage to ring bus may be sporadic as well. Voltage to Memory Controller could also play the role. How about something more serious: integral architecture's transistor failure, not because of voltage to the core, but flawed interactions in voltage to the ring bus or even defects on manufacturing level that is then multiplied by millions of chips. Flawed design, perhaps. You said it yourself, they released it in record breaking time. My chip failed at 1.28volts, which is an extreme undervolt on 13900k. Literally 0.01 volts next to unstable chip. But, undervolt does NOT degrade the chip so rapidly, it can (despite common myth it doesn't), but not at such a quick rate. I know you are not running 13900k or 14900k, and I am running 13900k since launch, so I had time to examine the chip for what it actually is, and that's no offense, it's just experience with it. I undervolted for years, and the nature of the degradation could be similar to undervolting at first sight, but it's more common with overvolting the cpu in how it locks up, because before bluescreens I would just freeze on Win 11 screen randomly (never in games), nature of too much undervolt. Raising voltage back did not solve the issue, and even stock setting did not solve it. Meaning, it was like I overvolted the CPU for 5 months, which I didn't. The fact that Intel's sources are speculating means they cannot identify the issue. Oxidation is an excuse. Voltage is likely an issue, but I speculate architecture of E-Cores and P-Cores, cache, MC, ring bus, and perhaps something more could be at play. In my case first red flag was when my memory began to fail on stability tests, perhaps Memory Controller on 13900k degraded first. You see where I am headed? No one truly knows where integral failure resides. I trust Intel engineers on assessment, but so far they are speculating. This is the biggest chip disaster in a 25 years of owning PCs (to me). Because that people that know are hiding the truth. That is upper management and engineers that signed NDAs and will never speak until they are not under NDA (aka fired and in 2-5 years after NDA expires). I also don't have a horse in the race, I owned Intel for 12 years exclusively, and 13 years before that was AMD CPUs. Look up how much debt the Intel is in right now, it is in double digit billions of dollars. Any badly placed public statement will bankrupt their company, and reliance on government bail out is almost the only thing that can save them, unless they choose to borrow more money and be in deeper unsustainable debt.
@@MooresLawIsDead My CPU was watt limited from day 1, it was also heavily underclocked to 1.28v, and was not boosting to 5.8ghz. It still failed. I speculate voltage is jacked up in other facets of the chip besides of VCore, I did it on Day 1, tested in AIDA64, Intel Burn In, Intel Extreme software, and TM5. Memory began to fail first, so I would say Memory Controller was the first to go. I know how undervolt and overvolt failures look like since I've been doing it for 18 years or so. The issue is not just boost. It wasn't boosting at all. As I also pin my cores without Intel SpeedShift and without SpeedStep. VID table reads max of 1.37volts to V-Core. Also, RU-vid deleted my previous comment which was way more in depth on how and what I did. The engineer source that gave you information is only part right. My stuff failed without boosts and without overvolting. Mark my words, Raptor Lake is flawed on manufacturing/structural level, it's much more serious than just the voltage as we can see. They botched the manufacturing design of the chip. Somewhere, as you can see even lower end chips (even below 13700k level) are affected. What does that tell you? That architecture on 13th/14th gen is messed up. The real source that would release that statement is under NDA, and they wouldn't risk releasing that info until they are absolutely certain it won't be traced to them. No one is gonna risk their job on higher level to condemn their own position and increase chances of potential bankruptcy of said company. Take this scenario, I am a high level engineer at Intel, do you think I would release a statement to RU-vid channel, even for pay, when my whole livelihood is based on said job, and my salary is in $200k+ range? Absolutely not. I don't expect real truth to come out, until everyone is safe with the decision of leaking it. People's heads will roll if they spill it now, to you, or to anyone. But, trust me, it is way worse than some boosting behavior.
this story resemble in some way the infamous ps3 ylod failure that literally EVERY PS3 with a 90nm and a great amount of 65nm RSX chips WILL die and only the one that were lucky with a golden sample will have their ps3 phat and fat/slim that survives without going to the frankestein ps3 route
90/65nm chips from that era do have a design defect, however the defect isn't this bad. It's an underfill issie IIRC and the main problem is that it can't handle thermal stress as the material starts expanding past 70°C. The higher, longer and more frequently you push the temperature beyond 70°C the quicker the chip dies, which is also why Nvidia 7000/8000 series laptop chips also had extremely high failure rates. On the other hand if you're able to keep the temperature below 70°C the chip won't break as the undefill material never expands. Thus it is actually technically possible to have a PS3 from that era never die on you, but it would require that it was never used to play any 3D games without homebrew fan control keeping the chip cool, which is extremely unlikely given that this was discovered only quite recently.
@@giserson2 I bought 'the' PS3 back in the days, 60gb, 4 usb ports, backwards compatibility, Linux mode, avoided the patches etc (avoided entire games due to this). I went through three of them in warranty until I returned it and put the money towards a ATI I believe still at that time it was during the merger 6970 in a new build, 1440 IPS 14+ years ago was a riot. Motorola V10s also have this flaw but you can reball them with leaded solder to fix it. Unleaded solder caused far more environmental damage through failure, than leaded solder ever did.
@@giserson2 infact i have a ps3 model cechg that, even if i delidded it, changed 4 nec tokins ( was giving error 1002 and 1001) it still died on me ( error 1601,1701,3034,4432)
I smiled when you said Sandy Bridge. My HP Z620 which is what I watched the video is still cranking dual E5-2667, pretty much the best computer I have ever owned. I hope my 14900k stays alive, its been stable and reliable so far but I locked the cores right away and never pushed it.
In this case admitting mistakes at this point in time would also sink the company in damages payouts Which would be fine for the investment bankers who are trying to sink it and asset strip it
Coverup of a mistake of this magnitude usually comes from the top. It's time for Pat to resign. It's not like Intel is doing gangbusters since he came aboard as CEO.
I am swapping out my broken 13700k processor tomorrow. Could I take a picture of the lid of the CPU and somehow email it or send it to you so you can look up the batch number information and see if it came from the Arizona fab, and or add it to like a spreadsheet or a data collection document that you might be keeping? I've been using this CPU since Feb 2023 if that matters.
You're definitely right about the AMD delay trying to avoid PR problems. Just look at the X3D failures that happened at launch, there was only a small handful of chips to experience that failure. Yet the issue blew up and was talked about everywhere, despite it being a tiny hiccup in reality.
That failure wasn't caused by the CPUs themselves, it was a couple of motherboards that weren't applying the correct voltage (showing one voltage in BIOS but actually sending a much higher one). And I don't think they're delaying Zen 5 just "to avoid PR problems". They're doing it to ensure maximum PR problems for Intel. Not only do they keep the media's focus on Intel's failures for a bit longer, but when Zen 5 does go on sale, it'll be against patched (and therefore slightly slower) Intel CPUs. Plus they come across as the "responsible" company, that didn't hesitate to recall all CPUs due to some minor performance issues (while Intel still hasn't, despite tons of crashes).
Funny how AMD downclocked (and undervolted) X3D to avoid frying the V-cache, and now Intel's apparently frying their chips in that very way, though admittedly in a different area of the chip. Just goes to show how hard it really is to make a chip.
Nothing got downclocked in chips bro, what are you even talking about. It was a SOC voltage limit issue when using ddr5 expo memory overclock😂. They limited soc voltage to 1.3 V because ASUS pushed it to far with their faulty mobos. Stock chips work without any problems. On the other hand Intel chips are dying even undervolted. It's completely different issue.
@@dmitrijssmeiss2737 You're talking about something entirely different. 3D V-cache equipped CPUs have lower boost clocks than their standard counterpart because the cache is voltage sensitive. The 5800XT goes up to 4.8, the 5800X3D only 4.5.
@@Eidolon2003 still nothing to do with dudes comment. Out of the box these chips work flawlessly, whereas Intel chips even downclocked and undervolted run into issues. Couple of chips burnt because of AMD and mobo manufacturers communications issues regarding EXPO and that's all. Problem was fixed within an hour and everyone got replacement mobos and chips. Dude said it's hard to make a chip and not to fuck up. On STOCK settings they work perfectly fine, issue is not in the design. First problem was enabling EXPO with to high voltages limits and second pushing ddr5 memory speed out of the 7800x3d specs to 6000mhz which led to increased voltage on soc controller. Nothing would even happen if guys run the memory on supported speed of 5200mhz because at that time asus pushed voltages to far. These clock downgrades which you are speaking about are already built in the design, you know what you are buying, whereas Intel just suggested to drop CPU performance for 10-15 percent after you already bought super expensive cpu thinking that it will run at advertised speed but apparently not, which is concerning. Its like buying a lambo and guy says you can ride in it only at 10 mph.
@@dmitrijssmeiss2737 You're right, but the original comment had absolutely nothing to do with the Vsoc thing. He was only talking about X3D having lower clocks than standard out of the box. The word "downclocked" is comparing the stock clocks between the two, not that they ever dropped the clocks after the fact. AMD "downclocked" X3D intentionally from the factory because they knew the cache couldn't handle the voltage requirements. AMD had the foresight to alleviate the issue, unlike Intel.
Interesting that you bring up Sandy Bridge (which was a great arch on a great node) because that was back when intel was on top both in terms of process *and* µarch. These days it's pretty obvious that they don't have a lead in either. They may be closer in architecture with AMD, but TSMC seems to have them beat in process. It is my opinion as a guy on the internet that intel got overrun with finance guys and MBA types during the 2010's when there was not a soul even close to them in any way. No competition. IIRC the performance gains were single digits pretty much every generation during this time. And let us not forget, 4 cores for a decade. This is the result. I could be wrong. It could be my dislike for intel and their priorities over those ten years. Mostly I feel like they put desktop on the backburner and we lost a ton of potential performance during that time. They could get away with it then. What were enthusiasts going to do, buy Bulldozer? Anyway, the chickens have come home to roost and I agree with Tom that a lot of these chips, even the lower bins, are going to just fall over eventually. In ten years a working sample might be a collectors item. Who knows.
Buildzoid explained that very well in its last video a day ago. Every part of the CPU calculate a "desired voltage (VID)" and at the end, the highest one is applied (Vcore). Exemple; when one P-core wants to go the 6 GHz, the VID may be over 1.500v. This 1.500v will be applied to all the other cores including the ring bus.
one thing I found strange is I set my (14900kf btw) cpu SA voltage to 1.1 in bios (was auto at 1.4) and cpu vddq to 1.35 (was auto at 1.41) and i have better stability. ive also gone through the trouble of disabling the 6ghz 2 core boost (all p cores at 5.7), and I finally have (so far) complete stability. ive tweaked a couple other voltages down and so far so good. i think they screwed the pooch on it big time. I also disabled Hyperthreading as a precaution... things are fine now no more crashy crash. that was the worst and biggest pain of my life. Should mentions I’m in the latest bios of an Msi z790 godlike (major regret btw, good board just way overpriced)
The problem is you'll never know how long it'll last. Whether you've already damaged the CPU or not. I'd be inclined to RMA the CPU so you can do this with a fresh one.
@@leebutler9470 Hi, that's why i do that , 12th Intel gen was ok, more i checked intel was making with newer gens, less i liked. so i was going for AMD: i upgraded from i7 8700k to Ryzen 7 7700, i hope the choice was better, except for idle consumption with X670e motherboard :/
I think Arrow Lake does also change something related to the ring bus, organizing the P and E cores into seperate clusters, so hopefully Arrow Lake does actually fix all the issues that has been plaguing Raptor Lake.
@@auritro3903Problem is if Arrow design development cyle was started way before the Raptor lake problems started poping out then then Arrow could have designs could inherit the same problems. The bright side is if Meteor lake or Lunar lake doesn't have these problems of severe degredation as Raptor Lake then maybe this problem has been solved.
Less than stock. Also, if you have say 8 P cores and 12 E cores a lot of the time you really only have 8 cores because a lot of resource hungry programs aren't even coded to use both types of cores, only P cores. Windows doesn't even know how to manage them properly. Intel have created a mess and need to go back to the drawing board. None of this is an issue with the AMD processors.
Pretty big bummer i still have a 2700k that ran over 5ghz for 10 years even had it at 5.3 for benchmarks. It still runs today in my living room. Meanwhile my 13900ks is running at 5.2 just so it wont kill itself
As someone who seemingly has a "golden sample" 13900K, I've been running it avg 10 hours a day both gaming and professional full-core maxed out use (3D rendering), and has performed flawlessly for over a year now. I run it on a Strix B660 board so no overclocking which surely is a theoretical benefit, but still if it was a faulty sample I'm sure I would have experienced some degree of crashes by now. I must add that I'm located in the EU, and I wonder if our chips came from a different fab so some of the potential risk factors related to manufacturing were limited compared to the ones shipped to large parts of the US. I can't say though, mine is just one sample out of 1000s, hardly worth extrapolating any data from. But I do wonder if there's actually a measurable difference in various global regions and the reported failure rates. Could be an investigative avenue that could potentially lead to isolating the main issues and where they stem from -besides the general chip architecture itself which is shared among all samples worldwide.
June 4th I bought a Lenovo Slim 7 Laptop 2023 with the AMD Ryzen 7 6800HS-CE, 8-core. I never liked the high voltages in the intel chips even without knowing about oxidation or massive failures to come. My 'ol 10th gen gaming desktop is the last Intel i'll ever own. It was a good run up 'til then.
mine was reeaaaallly bad when i brought it, it started having issues just after two days. i couldnt undervolt it BY ANY value, in fact i had to overvolt it to make it stable by +0.030v. How did that work? im really curious and want to understand that, any theories? i think it might have just been really sensitive to vdroops? because i ditched that overvolt for a extreme loadline calibration this made me really stable. Either way, mine is just 100% garbage silicon
This is all on Pat Gelsinger. Came in 2021 as CEO, probably ordered the Raptor Lake architecture to make a big splash. Raptor Lake as you said are not even in the radar of your contacts and was developed less than a yr. Released on Q4 2022. Meteor lake was the true successor but Pat wanted to please the board of directors. Way to go Pat. I thought Bob Swan did a bad job but you seem to have outdid his bad leadership even though you came from a technical background.
Yea, it's sad. Those bad decisions always come from the top. I would not be surprised in a couple of months to hear from some engineers at Intel that they knew about this problem and their warnings were ignored.
So much fud on the Internet. Gelsinger had nothing to do with this. Lead time on chips is significantly longer than a year pal. He is just the guy in the corner office currently. In another year, this whole thing blow over. Shit happens from time to time.
i9 14900k user here. I have no problems (yet) with my CPU, but I have newest BIOS, I cut down powers to 200W and clock boost to 5.4. And I will never buy Intel again. Not because they f.up two generations of CPU, but because they are playing dumb, and waiting in silence until whole thing will die down. It will not.
What if they know the i branding will have all this negative press because they already know the problem and see it coming, so they changed the brandind for the Ultra one so when someone search about it will get less results about this 13 and 14 gen issues?
Didn't Buildzoid from AHO mention it looked like a ring bus issue two weeks ago? Personally, I was convinced it was a layer interconnect issue (which would be compatible with the via oxidation that Intel has already admitted)... and it's starting to look like we were both right.
I am using a 32nm Xeon X5690 machine to type this comment. It is about 15 years old, and is rock solid. It just burns electricity a bit like a 1970s V8 ;)
People love talking all that shit about how infinity fabric is so bad and causing so much latency when the cpu needs to access system RAM yet this ring bus issue is laughable compared to a slight bit of latency. Also when it comes to the microcode, Buildzoid did a video comparing a gigabyte mobo with the first bios available for it, then with the last bios with the 125 microcode that came out in july. It showed that it slightly reduced voltage in some cases while it reduced multicore performance in CB r20 by 7%. he didn't let a single core test run all the way through but it didn't seem like it effected the single core performance at all. If this NEW microcode reduces pcore performance by another 5% then intel is going to get blasted.
How long before Intel have to bite the bullet and accept they are going to have to replace ALL of the Raptors? People can't be certain that they don't have a chip with a design defect. And business customers aren't going to put up with unreliability, once they realise the issue. If they don't do it before Arrow Lake is launched, they are going to have such poor PR for their launch they won't survive it.
@@andersjjensen Intel is not going out of business. AMD was in a much worse situation in the past with many billions of debt, and yet they are still here.
Why replace all the 13th gen? Most CPUs sold are i5 and i7. My i5-13600K Vcore never goes over 1.200v. I would be very surprised if this chip failed. Clearly, all i9 are in danger. The gray zone is the i7 series. Maybe most of them can be salvaged if the next microcode update can lower the voltage. At the end, Intel may have to replace, I don't know, between 10% to 20% of the whole 13th batch?
Improved longevity is only enough if they issue a 5 year warranty for desktop parts from the point in time the patch is available. For laptops that warranty needs to cover the full machine.
This is why I tune into this channel, Tom delivering the goods with excellent leaks. As an intel consumer it's great to have this info to help inform my future decisions.
13:20 ever since Meltdown and Spectre both AMD and Intel microcode is shipped by Microsoft via Windows update because they realized they can't trust mobo vendors. Windows then loads it during the boot process (it's temporary and has to be done every boot). You can find out which microcode revision windows loaded via powershell. The same can also be done on Linux, most distros ship microcode packages you just have to install. It wouldn't really be a major problem to roll out the fix for everyone that way, unless it requires additional changes on the BIOS side..
I absolutely believe it's the ring and this is what the OC community has been discussing for a while now. Recently buildzoid broke down why that's likely the case in a video. They really screwed up and if it's not something that can be fixed easily they'll have no choice but to limit clock speeds.
@@devins7 I saw their coverage of it. I was mostly referring to future chips. If they really are affected and can't be reworked, they'd have no choice but limit voltage which means reducing their max clock speeds. As for existing chips, I'm not sure what you're referring to by "make it worse". Unless a chip is already significantly damaged, limiting voltage (and thus clocks) should help for the time being. Nothing will fix degraded/damaged parts but the first hot patch Intel released has worsened performance by nerfing boost behavior in various loads. If they still put out another update next month instead of a recall, I suspect it'll just get worse. There's only so much they can do and I think they're just buying time.
I don't want my 14700K's lifeline to be extended by an update; I want to be able to use 100% of the advertised capacity of the nearly $650 CPU I paid for --- or get my *money back in full.*
Should the fact that a single one of their wafers costs as much as a Model X be a concern ? How are they even going to compete with IFS if the wafer pricing is so high
