The patch won't come until August because AMD is presenting their new CPUs on 31st July, and they want the benchmarks to look good. We can extrapolate from that that the patch is gonna drop performance severely.
@@chrismay2298 As long as 14900K hits 3.2GHz on P-cores, 2.4GHz on e-cores, then they've hit the marketed base frequencies, everything above that is listed as a turbo frequency which Intel would argue is a "maximum target" that they would try to hit assuming power and thermal ability to do so, but they offer no guarantee of doing so. Unfortunately, I think Intel has the money and legal position to likely get away with standing their ground, eating the cost of any dead CPU's and then weathering the storm on the patched CPU's that hadn't died yet. While the right thing to do is to replace all of the affected CPU's, there are vendors (I believe it is Dell or HPE) claiming at least 8M units affected. If that were somehow only 10M total 13th and 14th gen K series impacted, Intel being willing to replace even the non-dead chips with whatever the next K-series is I'd guess would like cost them somewhere in the $3B to $5B range. They'll do the calculus, maybe replacing everything dead plus patching everything else costs them $1B up front and a few billion in future sales due to lost good-will, money people would argue that a guaranteed loss of $1000 today is more painful that a "potential" loss of $2000 tomorrow, I'm doubting they'll do the right thing here outside of ensuring their larger OEM's are given free stock to resolve RMA issues, while everyone else including those out of warranty will be left high and dry with a patch and a "sorry guys"..
@@KarolusTemplareVnot sure how many will/can take that effort to update reviews... since it will be after the AMD review, maybe best case they exclude the i9
@@Tarrasq-Eredar I would suspect that channels like Hardware Unboxed and Gamers Nexus will retest following the patch. Thing is Zen 5 X3D is speculated to arrive in September and if this patch really does take a significant hit to performance its going to be a awful look for Intel when those X3D reviews come out.
Reminds me of the initial Intel Meltdown/Spectre BIOS updates: “You may experience more unplanned sudden system reboots than expected.” - What?! I expect 0 unplanned sudden system reboots.
4790k overclocked, though slightly for 8 years. No issues, updated to a 13400 recently, hope that model isn't in the danger zone, so far not according to all sources. Midrange bang for the buck buyers win again
Its not BS, the issues with 13-14th gen intel CPU's are well proven with data to back it up by now. This has been done by basically all the well known reviewers and tech channels. And now intel has officially stated there are issues with them and says a "fix" via microcode will come. If it actually fixes it we will have to wait and see. Fix or not , the damage has been done and their brand will take another hit from it.
@@lost4468yt Nope, unless you're a millionare. Fees, laywers and few years of fight will drain you dry, that's most corpos' strategy, and it works damn well. Welcome to capitalism.
Forget it, nobody will buy used 13th and 14th gen. Alone the "oh yeah, we fixed the 13th gen oxidation problem last year. We knew it was our fault, yet we still denied warrenty." With that they killed the used market for their chips
@@seeibe AMD is paying review channels to shill for their products and make them seem like more powerful in a first place so dont worry intel is still better in gaming and everything else
they will drop them all the way down to 5.4 max. If that was my issue id adjust the operation temperature to voltage ratio. They know they have oxidation issues and temperature is playing a huge role if not the most important role
I don't think Intel has enough physical CPUs to handle the RMAs, so they're going to run over the motherboard makers with their struggle bus until they figure out something else. Although, I do credit this for introducing me to Frame Chasers, the UserBenchmark of CZcams.
Intel storing the current as a pseudo float is a semi-standard practice, when you want the certainty of ints but need more precision. Bit stuffing can make the maximum value a bit odd, but you see this a lot in older games. Starcraft 2, for example, stores hitboxes as a 5 bit value where the bits are quarters, halves, ones, twos, and fours, for a maximum of 7.75. Any value that's set higher in the unit editor will default to 7.75 under the hood. Intel is, apparently, using either an 11 bit unsigned quarter or a 12 bit signed quarter. You don't see this sort of thing too often anymore, because bit stuffing isn't a thing that's done too often anymore, but the design constraints of a motherboard's BIOS are tighter than most programs.
The design constraint isn't anything to do with data size. It's that VID is the format used to communicate between the PMU and the VRM controller, neither of which have an FPU or any reason to have an FPU.
Warranties are the law. So many people spending hours adjusting voltages and clocks on a broken CPU and getting nowhere. If it's faulty then get it fixed under warranty
Well, well, well, what do we have here? Last year, I wrote Gigabyte about an issue with some unusual high voltages being measured by HWinfo on my Z690 Aorus Pro. Gigabyte's response was, please send a screenshot of that measurement but I couldn't catch that second that showed the 1.588 Volts being put on my 13700K. And after sending the screenshot, they said "we don't see an issue here...and we don't know nothing". With Raptorlake being still new, I doubted myself and thought, I didn't observe that new architecture long enough to make an educated guess, that that wasn't supposed to be there. Since there wasn't any mention at the time anywhere, I thought, it was a fluke. Self-doubt can put a real dampener on your life! However, in my 28 years of system-building, I may make some observations over the years and this one struck me ass very odd, only to read about it a few weeks ago. It is almost like I told them so..
Ryzen 3700X had this problem on launch, never hit boost clocks, pbo would hard reset. You didn't have the crashing at stock, but it was false advertising.
While you are mining copper from a mine in Minecraft running on a distant Minecraft server, the electrons in the Minecraft Server's i9-14900K CPU die are doing some copper mining of their own. This happy symbiotic arrangement goes on until the CPU copper mine runs dry or there is no more room in the previously working silicon die for any more copper "nuggets" that those squirrely electrons have managed to bury and forget about in the surrounding silicon.
I mean a g3258 would still use ddr3, so memory limited If u could get a Z790 Mobo in a blade, u can do a crazy thing like 7600 96GB DDR5, disabled E cores and HT and overclocked 14600K
Run two myself. Not a bad middle tier cpu. But just like the “exploding” x3d issue back in the day, this to shall pass. It’s the two cores boosting to heaven at 1,5 causing this. Lock your cores and be done if you own a 13-14900k. Or rma your cpu if you have issues then lock em. They wanted to be on top of stupid benchmarks and it’s coming back on em.
...middle tier @ half the wattage and faster than 14900k in gaming. Half of these Intel CPU's will fail sooner or later according to everything we've seen so far. The "exploding" AMD CPU issue was a motherboard problem with mainly ASUS and only affected very few people. That Intel stock is going to go way down.
@Coliflower185 well, yeah. The 7800X3D isn't the best CPU they make, you know. The Threadrippers and EPYCs of the world are way better for this use case, at the price of being much more expensive. For these types of servers, Ryzen/Core IS midrange.
I’ve recently updated my motherboard bios and find all the update does is seriously neuter power to my 13900k. My chip barely ever hits 5ghz on multicore loads now usually floating around 4.7ghz and I am seriously pissed. This isn’t what I payed for, and I’ll be having a few choice words for their rma department.
This is the first example of this I've ever encountered, the updates were released early July so its not something that's just dropped. Might wanna check your thermals.
Technically this is what you paid for and well above it, even. The 13900k has a (guaranteed) base frequency of 3.00GHz for the P-cores, everything above that is considered "turbo" and is not guaranteed at all. Look up the spec sheet at Intel Ark.
Insane Levels of customer screwing. They push the fix beyond AMDs release to look better at the charts. After that their CPUs will get decreased performance. It also pushes the issue beyond warranty. So the degradation is slowed down and the people who are affected then can stick their CPUs into their rears, because Intel tells them it's out of said warranty now. Well played Intel, well played. Deceiving, playing on time... that is perfect corporation shenaningans at it's best right in front of every customer. And they will go away with it with just a slight slap on their hands. I bet they even knew about the issue way before but sold their stuff anyways because the profits sould out the costs afterwards in their calculations. Pure insanity if you really think about such behavior.
That doesn't make any sense. RMAs cost Intel money, that's profit lost. If they knew about the issue they'd have fixed the microcode bug. If they knew some of the processors were going to degrade and need RMA they would have scrapped the processors instead of selling them in the first place. People forget that this doesn't affect *all* their processors. There's a reason it's almost exclusively the i9's. If I had to guess it seems reasonable to be only those already running the highest voltages from the factory ( barely making the cut ) that are the most affected because Intel's buggy microcode is pushing even more voltage into them and causing the initial instability and eventual "threshold shift" in required operating voltage for maximum boost clocks.
@@zcomputerwizIntel has to RMA them anyways. Most of the Issues ppl do have now until in a Month will not spike from 10% now to suddenly 70%. The Damage on most CPUs now are done. It’s better for them to sit there with a „better“ Review against new Ryzen and 3% more RMAs, vs 3% less RMAs but a worse Review. Right now even GN just said „he look… we don’t and can’t recommend those Intel CPUs but we will test them as released“. If Intel would have fixed it Allready it would be like „they fixed it, it’s not a big deal (or it is who knows) but we would not recommend Intel right now“. They are fked regardless imo.
1.5v is what CPUs used 20y ago, when the process was 130nm! Even 32nm Sandy Bridge overclock was 1. 4v pushing it. These are far, very far from safe for a 10nm chip! Even the E cores are at that voltage. Why!? No wonder people are reporting E core failures. Even more, it is almost impossible the temp reading is reporting hotspots. This much voltage and clocks, 50W on a single thread, the core itself is now so small it should be close to Tj Max, not 65C. (or Intel has used a lot of dark silicon...)
Agree. For the latest 5nm process node, 0.8v should be the normal operating voltage. I bet current crop of processors won't live to 5 year mark. Such a waste of resources all for consumerisum to keep going on and on and on.
@@bozilla9975 My AMD chips are set to throttle at 65C, which is the universal way I found it doesn't really pump much than 1.2v Zen chips dont seem to be far from top clocks but Zen 3, being a smaller process, generate many hotspots. Zen 3 also seems to push more voltage than Zen+, so it becomes worse. A slow 5500 was pushing 1.35v, so single and dual core loads heat up much more than multi, where TDP limits rule. It was a while back that people were reporting issues with 3600s but seems that was limited.
@@zcomputerwiz AMD runs 5nm chiplets at 1.5v max boost voltage, which is kind of insane. I don't know what their next generation will run at but I suspect it's going to be a lot higher than 1.2v.
AIUI the actual voltage being fed to the cores is not 1.5V; modern Intel CPUs have onboard polyphase buck regulators (FIVR) that provides the actual operating voltage.
Buying a secondhand albeit unused 13900KS is the worst thing I've done in a while. I was like "CPUs rarely die, it's relatively low risk". Now I've got a degraded CPU with no warranty. At least I've got it stable at 5.6 although I had to underclock and crank the llc to get it stable. Weirdly at stock 1.4v at default llc it bsods at idle every few days. At 1.35v with the llc bumped up from 4 to 6 it's stable.
I am on the same situation but with an 2nd hand 14900K, back in november. At start all worked like a charm, now is unstable. Had to low voltage to 0.85v @ 5.5 GHz. Seems pretty rock solid now, my nerfed down 13900T now at 550 bucks with no invoice at all.
@@peterpain6625 its fine with amd. their cpus literally blowing up were actually the fault of motherboard manufacturers (asus) feeding the cpu too much power.
Maybe Intel accidentally activated their next planned phase to make more money: The CPUs are only good for a certain amount of computation then you have to change ‘em - just like car tires!
To Intel AMD and Nvidia . Can we have chips that are not pushed to the limit. Can we just have one generation where the fab improvements are all used on efficiency . K thanks .
I have a Ryzen 5800X, which I have run as stock, overclocked, and undervolted/overclocked with PBO. I have used this CPU to run a technical minecraft server which occasionally experiences INSANE loads while also running an instance of minecraft at the same time AND streaming it all to youtube. This server has been online pretty much 24/7 since early 2021, and I have never observed even the slightest degradation in performance or stability. This is completely unacceptable from Intel's part.
My brothers and I have run about the same settings and have had no stability issues so far, we're all running PBO with a negative 30mV all-core offset on 2 5800x's and an 7800X3D. I have my 5800X running a 200Mhz boost override and the highest single core voltage I've seen is about 1.419 at 5050Mhz IIRC. Not sure if we won silicon lottery or AMDs cpus just have a ton of headroom from the factory. Either way, I'm impressed. My little brother sure is glad he got the 7800X3D instead of a 14th gen lol.
server space. regular people who just play games on the computer don't care what happens with server hardware, even if it is the exact same hardware, they would just brush it off as "oh but it's servers, my pc doesn't run this hard" or something either servers guys didn't make it public because nobody asked or they made it public but people who heard of it just didn't care so it didn't gain any traction and got buried under all the other news till cpus started failing in regular desktop computers
They have not. I've been seeing more and more issues for 2 years since 13 series launched lol. Just people said they were FUD because there are so many intel share holder shills on forums.
Turn down the CPU current limit (IccMax). Mainboards often set it way too high to 512 A even on Auto settings. Even Intel strongly recommends to never exceed 400 A.
Saw a 12900k die in the way you are speaking of a few months ago, the host was fully shocked, first crash seemed out of the blue, then after second one it wouldn't pass validation, think its been in the box since 12 was new.
@@SteelBlueVision @ 12900k is extremely similar to 13700k.and just because Intel has not yet acknowledged that is not affected doesn't mean it might not have an issue.
so a single 14900K is for about a year of minecrafting, 8 cores, one dies, disable and force next core etc. jokes aside, if they are clocking it down to 5.7ghz, then isn't it the same what ryzen could perform?
Welcome to the club. Well actually tbf I'm on an Intel right now but it has degraded over the years, and I really thought AMD was a powerhouse as a teen until all my friends convinced me Intel was better. Now after having Intel for years, on many different PCs, id rather have an AMD again. I think Nvidia gpus are a lot more stable, but I still love AMD gpus. AMD ftw
@dementedpeep I think intel is definitely better just works with everything and I think intel will eventually get this corrected I just should have stayed on my 12900k as far as Nvidia yeah they are a powerhouse I'm rocking a 4090 it's a beast but Amd is definitely probably better value
I have an idea: Manually set the voltage to 1.45v with a moderate LLC - disable all but 1 P-core - leave all E-cores enabled. See how fast it degrades with a looped stress test. I have a feeling that the degradation that's occurring stems from the cache. Similar to how X3D chips will degrade the fragile cache when overclocked.
8:07 About PL4, from the Intel 12th gen documentation: The package power control settings of PL1, PL2, PL3, PL4, and Tau allow the designer to configure Intel® Turbo Boost Technology 2.0 to match the platform power delivery and package thermal solution limitations. Power Limit 1 (PL1): A threshold for average power that will not exceed - recommend to set to equal Processor Base Power (a.k.a TDP). PL1 should not be set higher than thermal solution cooling limits. Power Limit 2 (PL2): A threshold that if exceeded, the PL2 rapid power limiting algorithms will attempt to limit the spike above PL2. Power Limit 3 (PL3): A threshold that if exceeded, the PL3 rapid power limiting algorithms will attempt to limit the duty cycle of spikes above PL3 by reactively limiting frequency. This is an optional setting. Power Limit 4 (PL4): A limit that will not be exceeded, the PL4 power limiting algorithms will preemptively limit frequency to prevent spikes above PL4. Turbo Time Parameter (Tau): An averaging constant used for PL1 exponential weighted moving average (EWMA) power calculation. Notes: Implementation of Intel® Turbo Boost Technology 2.0 only requires configuring PL1, Tau and PL2. PL3 is disabled by default. So this is a value that can't be exceeded because the frequencies are limited "preemptively".
The fact we're even trying to figure out if a slightly out of spec power setting is causing a 1/4 of these chips to fail within a few months is just wild.
They will claim it's normal wear and tear .....bet or say it will * as long as Temps are below 40c ....another option is make it show 6ghz using null cycles hurts performance but looks good on paper
The only numbers the CPU's would need to reach to avoid losing a lawsuit are the base clocks for the P-cores and E-cores, which I think top out at 3.2GHz and 2.4GHz respectively, every boost is an "up-to", and all of the turbos and boosts specifically call out other conditions that must be met and the maximum numbers are always "up-to". If you bought a 13900K 2 years ago and it never hit 5.8GHz, or more specifically, it never even hit 5.4GHz, even before this issue became a thing, Intel could tell you to pound sand as everything above 3GHz clocks are dynamic and best-case given conditions. They would not RMA a chip that wouldn't hit max thermal velocity boost clocks, or max Turbo 3.0 clocks, or any other turbo /boost clocks unless it was crashing at totally stock settings.
@@racerex340 My understanding would be from within the EU. Up to isn't allowed to deviate too much. Internet providers have been slapped because internet connections wouldn't come close to the advertised 'up to'. I think it's a 10% deviation that's allowed for some time. I imagine that can be applied to CPUs, too.
@@maskharat well, if it were 10% with thermal velocity boost, the Intel would need to maintain a single core boost of 5.4GHz for the boost window, and for any others like turbo boost 3.0 I think it'd have to be about 4.9GHz. I don't know how much the issue they're having is impact single/dual core boost performance, I thought it was mostly under multicore workloads, but if it is due to resistance, then hitting even 90% those single and dual core clock boosts might not be sustainable.
The "13th Generation Intel® Core™ and Intel® Core™ 14th Generation Processors Datasheet, Volume 1 of 2" PDF is no longer available. Volume 2 is still up. Are there alternative downloads anywhere?
I think the only solution for Intel is to release a Socket 1700 "Minecraft Xeon" that doesn't fail, does decent clocks and has like 8 performance cores. Then offer that one as a replacement for the cpus that fail. For now they're only trying to push the failure out of the warranty window.
Everyone is talking about power limits and stuff but I think it's the voltage. My 14900K when set to Auto voltages was hitting almost 1.7v. I manually set it to 1.32 and it's been perfectly fine. I'm wondering what the actual vcore voltage is when running.
Electronics engineer & PCB board designer here. About AC DC LL values: if you set it to 1mohm it is not safer! The CPU will ask for more voltage under the assumption of 1mohm*current voltage droop. If the motherboard power delivery impedance is lower (hence better) than 1mohm you will fry your cpu with higher than expected voltage
Exactly, came here to say this. If the whole LL calibration works as I understand it does, and given buildzoid says that "setting LL slope to a lower value than the board capability allows will feed less voltage to the CPU than it expects", then the inverse is obviously true as well. I.e., setting LL slope _higher_ than the actual capabilities of the board will thus feed _more_ voltage to the CPU than it expects - which sounds exactly like the recipe for a disaster?
Hello, You seems very knowledgeable on the subject, is there's any ressources you would recommend to learn the propers methods for tweaking a cpu?i am looking to undervolt a 7800x3d. Cordially.
Even if the die is perfectly symmetrical, the power plane in the MB and substrate is not, so maybe the shorter-path side is getting enough extra voltage to degrade the odd P-cores fastest.
SIde note: I love it when people are not aware how CPU-hungry Minecraft can get, especially on mods. That game is no joke - it is, after all, a single-threaded simulation game on the server side. To add insult to injury, the game runs a simulation 20 times a second - the load is not constant, and the resource usage fluctuations will definitely cause some heavy power level switching and adjustments twenty times per second. (!)
Spec-wise those Intel cpus look like godsends for the workload. Pity they fail that badly. I built a minecraft server for the kid of a friend from "old parts" and the coils in that thing a whining like crazy ;)
For some time, I was running a mod on MC that splits the simulation so each dimension gets a thread. Not perfect, but at least Twilight Forest doesn't lag when someone was doing overworld shenanigans. I am considering adding it back and removing AE2, for that mod conflicts with Dimensional Threading Reforked
I appreciate the insight. There are a lot of creators talking about this that have no clue what is actually going on. At least you have something to add to the conversation even if it still needs a giant asterisk. 😂
Could you create a video about undervolting 13th and 14th generation Intel processors? Specifically, which value should we modify: Actual VRM Vcore or Global core SVID? And should we undervolt in offset or adaptive mode?
How does the 200 mV VID offset get triggered? What if it accidentally got enabled for a fraction of a second by the ucode? That would occasionally zap the cpu with say 1.6V instead of 1.4V
This is exactly how my 11900k I just RMAd was acting. It was less than a year old. New one tried giving me 1.7v @ 5.2. Have to keep it under 4.9 to stay under 1.4v
The fix won’t fix an already degraded CPU. Factory OC bites manufacturer in A55. Now thousands of reviews and benchmarks are invalidated, but almost nobody will bother updating them for those that will still be buying these CPUs over the coming years.
If these CPUs are getting degraded and damaged by excessive voltage spikes, then the 'fix' that lowers the voltage back to safe levels is going to come with a steep frequency degredation or it's going to destabilize a lot of CPUs in the field.
The fix specifically means disabling the 1-2 core thermal boost which is what is pushing the 1.5+ volts through the CPU. Without that the CPU works fine. But yes, its disabling a feature that the CPU was specifically sold for.
If The CPU is ran at stock, it will automatically adjusts its frequency based on the voltage it's fed. They will not get less stable, but boost clock will decrease to accommodate for the lower boost voltage.
Weirdly, I had a 9700k that went bad, randomly one day was unstable at stock, even like -700mhz on all cores, it was able to run stress tests a bit longer but eventually BSOD. The cpu was also hosting a lot of modded minecraft at that time. Was told it was a rare 9th gen bug and intel ended up offering me a refund. Always found it weird because cpus are meant to be super rare to fail - and i never played with unsafe voltages and temps would always be below 80c.
They wanted to be on top of stupid benchmarks. Let two cores boost to the moon at 1,5 and here we are. Rma your faulty chip, get a new one and lock the cores and be happy forever. Cheers.
Class action is the only way now. Intel couldn’t compete effectively (as in power efficiently) with zen 4 so it ran raptor lake far too close to the sun. And anyone that has tinkered with Intel K CPUs from at least skylake knows that some motherboards pump FAR too much voltage on “auto” settings. My msi z170 board was notorious for this so I learned to “undervolt”. The same has been true for every subsequent Intel CPU I’ve owned. Comet lake. Alder lake. And now raptor lake. It’s not only msi. Gigabyte and ASUS are notorious for high voltages as well. The auto and optimized default settings can be too aggressive. Especially for a CPU boosting to 5.8 GHz or 6 GHz on one or two cores. So it’s about time that Intel got its sh*t together and clamp down on the auto/default settings. It’s like they make the CPUs but don’t actually test them in consumer motherboards. Also these bugs in microcode in the eTVB and now voltage should’ve been caught in QC and QA. But if you look at the amount of bugs in Maple ridge thunderbolt 4 firmware (early maple ridge NVM versions did not support hot plugging certain alpine ridge TB3 devices), it’s clear that Intel does not have a coherent and quality QA/QC process. And that is the root cause of the problem. Bugs happen. But robust validation via a robust QA/QC should catch show stoppers like ‘instability/degradation due to higher voltages than spec.’ If they had actually used RoG motherboards to test raptor lake in the factory before release, they would’ve seen the high voltages. My mother always told me from a little boy: an ounce of prevention is worth more than a pound of cure. Intel’s reputation is suffering. A robust QA/QC is the prevention that Intel needs so as to build and maintain customer loyalty and support. At this point I’m considering not ever buying another Intel cpu ever again.
Gamers Nexus' video has some really strong arguments to suggest this is a manufacturing defect that no microcode or patch can fix. Intel just screwed the pooch on a manufacturing node they should have plenty of experience on and are trying to cover their tracks.
I grew up in a time when the gates were so big, there wasn't even measurable variance that would make binning a thing. Why not happening to AMD? Because chiplets. This is the ultimate end of monithic chips that can turn into either a i9 or an n100 depending on how it comes out of the oven.
@@kekoraaaa they don't have too. $#!+ chiplets go into the $#!+ bin, very simple way to maximize yields of good silicon. Intel needs to save whole monolithic dies, which is why they bin them so aggressively and have a HUGE inventory of n100 class chips. Each could have been an i9 but only four efficiency cores and the integrated graphics and quicksync (etc) came out of the oven. Ever wonder why you can get mini PCs for as low as $25ish per core? The tail end of Intel's monolithic process will mean dirt cheap low end minis for years.
Someone need to CONNECT THE DOTS. 1- Intel had problems in production, specifically in the coating of anti-oxidizer 2- Oxidation adds RESISTANCE. 3- CPU wants power, 4- finds resistance, 5- ask for more power.... 6- and ded! Intel says about Microcode, yeah sure.
Modern Intel CPUs (since 10th gen or so) applying some hidden 1.5 ratio on specified AC_LL so in order to get requested VIDs close to fused VIDs you need to set AC_LL / 1.5 from DC_LL/VRM_LL. Or if you follow "standard" AC_LL=DC_LL=VRM_LL rule then your CPU will be constantly 1.5x overvolted depending from current it draws
I love this topic ahhhhh, only one question do you think I should upgrade to the patch if I’m on a stable bios since December? 59 p 45 e 50 ring LLC7 undervolt svid best case scenario ever exceeds 1.35 and never have had a stability issue
@@PAIN-ot4cj from all these years overclocking learned that cpus face hard wall overclocking them. voltage wall* For example. Your cpu runs at 1.2v (load voltages*) 5ghz . then at 1.25 5.3ghz .and then needs 1.35 to run 5.4ghz. you don't run 5.4ghz if you like your cpu! If your next step in frequency needs 4 times the voltage it took the previous step. you have a wall infront of you. That wall needs exotic cooling to overcome and not harm your silicon. The end
Thanks fellas I thought so, the voltage algorithm only affected boosting behavior on stock settings and degrading them with super high voltage spikes. Since I locked my cores and voltage I saw no problems with stability ever. Won’t be upgrading till I run into some windows or other troubleshooting problems due to security or devices with my system
I think 1.35V is probably safe long term. Though Chances are if you back it of to 5.8 and 4.8 ring you won't really notice a perf drop but you should able to significantly reduce the voltage.
It actually would be a good match for a Minecraft server, just need to find a host provider that uses blades/servers based off it. Supermicro does sell them and guessing they will be selling a lot more AMD based systems in the next few months.
Wendell/Level1 Tech indicated that there was a significantly higher failure rate for servers running 4 DIMMs of mem. compared to those using only 2 DIMMs. Prob. just inviting inevitable doom faster. Also, Intel denies oxidation on 4th gen. How long before this is retracted?
They're already retracting by not making a general recall. They'll push a microcode update limiting max voltage requests, but beyond that they said RMA. HOW ABOUT THOSE WE BOUGHT 2ND HAND AND DOESN'T HAVE ANY INVOICE OR PROOF OF BUY?
Bought my 14900KF 08.05.2024 so 2 months ago and I am in RMA process with Intel waiting for intels response of how to stop this happening to replacement chip. (rendering workflows and mostly gaming including minecraft) CPU never passed R15 even as new with default settings now it no longer passes R23 that it passed before.
Would be interesting to know if it's just "normal" degradation of the ring (where a lower clock would temporarily fix the issue), or if it's actually a hardware failure
I have a 13900k and I realised that if you simply disable c states so it doesn't boost the p cores or the ring, it stopped the crashing in games (Also I don't think I have ever seen it boost to 5.8 on the p cores or 5 on the ring when gaming because the task scheduler generally tends to spread out a load to more cores. so games run 4 or more at the least)
@@TheHighborn I bought mine in nov 2022 and i didnt want to dable with dual ccd problems. But in hindsight, 7800x3d is way better for gaming. Don't get me wrong I think every platform/cpu has some problems it's just that some have easy workarounds and some don't. And apart for some IMC/memory compatibility shenanigans initially (Thank you gigabyte for the hilariously bad qvl list) I haven't had any other problems
lol 1.5v vcore. My 13700k runs at 1.22v during Cinebench at 5.6ghz 1.34v idle. It is super stable for me. Running 24/7, but 12h it is just idle, with plenty of decompression since 11.22. I would instantly shut down my PC if I saw anything like this in a daily system. But it is just terrible for anybody not checking their system when it is new. And that's totally on Intel. A new CPU should be always far below 1.4 if you want to use it 4-5 years. What a terrible product the 900k is. Anyways will pick AMD next time. Once they support my 7200 DDR5. No wonders, to go over 5.6 the vcore required ramps up like crazy for 100mhz.
Same.... well I'm running 1.43v on vcore at idle. It droops to 1.36v under load and has been completely stable and running for ~20 months now. Rule of thumb is not to go over a 1.4v or in my case not much more. The cpu's that keep degrading are running stock insane voltages that are not sustainable. People who know how to OC/undervolt these chips aren't running into issues.
So continuous 1.5v looks to be a problem and 1.4v is probably fine. Thanks for the info. Glad my 13900kf doesn't need 1.4v for 6GHz single, even at idle. Good to get numbers out there.
Intel hoped for the best with uber binned chips with 13th and 14th gen to compete with AMD now they're reaping and will probably have the gall to still blame consumers for running their chips at their advertised specs
I had my first chip fail, but I blamed myself because I limited to 300 watts, in fact I think Buildzoid led me here to 300 watt limit. LOL, I don't blame you, bro BUT, on 2nd chip I did not make that mistake and limited to 262 as it was a sweet spot where cores won't clock down, and also did a proper undervolt, like my first CPU So far so good. I smell that Law Action Lawsuit is coming against Intel, and I speculate people will get a $50 paycheck in mail similar to when I owned AMD FX-8320, and their 8-Core dispute lawsuit and $50 showed up at my door 3 years later. I am not celebrating this situation for Intel, I rather have them fix the issue on next CPUs and continue honoring RMAs the way they did with my CPU. AMD has been guilty of similar things, but the issue is that Intel is not addressing it properly, which will likely result in lawsuit, because this is....AMERICA, Harold, AMERICA
How about the rest of the world who are fucked also? And who the fuck cares about 50 bucks after 3 years when you needed your right new chip to work as intended for years since day one?
I disabled turbo boost on 14700K since its in my unRAID NAS. It runs a pretty high load, encoding, vms, dockers going 24/7. Hopefully the short periods of time I ran this CPU before disabling turbo didn't cause too much issue. Hopefully without turbo the voltages are way lower - I'm assuming they are since the 14700K loaded in my unRAID system usually maxes out around 53C.
what is your recommended LLC setting on a gigabyte aorus elite X z790 then? low or medium? my 14900KF has been running on unlocked power limits and its still rock solid for 7 months straight.
I've managed a few hundred servers and blades for the last 10 years or so. I've had maybe 2 CPUs die in that time span. I say maybe, because both the board and CPU were swapped after failures. CPUs dying is extremely rare, to the point of it not being considered during troubleshooting.
Gonna be hilarious if intels 15th gen runs on a diffrent socket (for no reason yet again) resulting in anyone who bought and trashed their 13/14th gen cpus just being shit out of luck. Buy a new motherboard, cooler, cpu, everything, nerd.
Hi, nice video I was wondering about phase counting. Are they always real phases ? 12 phases produced by a 6 phases driver chip with phase multipliers is better on not than 6 phases chip with better mosfets ? Do they delay correctly the phases to have a better output with those 12 phases ?
They''re not "gone" the 15th gen will have E-core and non E-core editions which will definitely eat a lot more power and be more expensive, they arent getting rid of the mixed model.
@@tmsphere if you dont understand the comment you should watch the video further or watch some older material. This comment is about current cpu's losing e-cores due to failure.
Can you provide a link to the " Patch now " ? I have read the stuff from Intel and they said the microcode to fix these is coming in August and I know there have been a lot of BIOS updates from MB vendors but I can not find anything about a patch that is available right now. Thanks BZ!
@@tmsphere Built my wife a new system with the 14700k and set all P cores to 4.7ghz and E cores to 3.4ghz. No more than 1.12 VID and a total of 144w of power at 100%. I will not touch anything more until the new microcode is in place but going back to BZ's comment, I still do not see a patch now available since most MB venders have beta bios up
Would be interesting to understand more about how AC & DC_LL works. What are the benefits of lower or higher values and why? I assume closer to 0.01 AC_LL that the voltage would match the VID table of the cpu if the MB LLC was set to no vdroop? Just trying to understand as Intel recommends that AC/DC LLmatch LLC? I’m confused lol
TBF, this happening on a pretty much a reference board is absolutely horrible. But how can you even get a reference board from a major manufacturer anymore? All of them set the limits to ridiculous heights to just one up one another. Are there even motherboards that run in spec out of the box?
From what Intel is saying the issue at this point is with their microcode, not necessarily the boards themselves. The CPU is requesting more voltage than is intended and destroying itself at full single core boost clocks. If you want to be safe, limit max boost and max voltage.
In a 14nm white paper from yonks ago I saw a MTBF vs voltage curve. Basically it showed MTBF got worse by a factor of 10 for every 100mV higher Vcore. I wonder if theres a similar thing for their 10nm, ahem, "Intel 7". If I owned a RL I'd be limiting it to 1.35V or something and just seeing what clocks it can sustain. As it is this sheetbox Acer laptop has undervolting blocked and the best I can do on this 13900HX is to limit turbo multipliers. I actually did this from early on as it was yoyoing off thermal and power limits (100W long power) so bad that letting it go up to its stock 5.4ghz and clock back down, was slower than limiting to where it throttles to, around 4.3 P / 2.8 E. Now I check VIDs, they're 100mV lower at idle/low load and 150-200 mV lower under part load, interestingly enough basically no difference when it throttles down under full load (just slower...)
The most recent Bios on my Asus Maximus Formula Z790 was "Updated with microcode 0x125 to ensure eTVB operates within Intel specifications". Fortunately i set my P-Core boost ratios to max 5.7Ghz so i would never have experienced that behaviour anyway. But i have noticed that the previous Bios " Still Intel Baseline btw" i was seeing a consistent 5.7Ghz in games, now it fluctuates from 5.4Ghz upto 5.7Ghz. Now they're talking about another Bios to fix irregular voltage requests. So am i going to see further nerfing to gaming performance?
Waaait - you do not know what PL4 is? It is the absolute powerlimit. By default NOT enabled it is a preemptive limit the CPU will try everything to stay under. While the CPU is allowed to go above PL2 for up to 10ms, cracking PL4 is a shutdown. As for the voltages - if memory serves me right intel specified something like +-5% or so maximum tolerance. And similar ranges for the memory-controller (rather narrow).
Until someone measures the voltage on that blade then I would assume the reason your contact has a higher failure rate than other companies is because Supermicros blade’s core voltage is higher than the others. Would love it if you measured it. I can’t think of anyone else who would actually connect a scope to see the “before and after” for these upcoming patches. Edit: Apparently 30% is lower than others. Wow.
Well, on my asus strix z790itx board AC/DC loadline varies after every boot. AC between 0.07-0.12 and DC 0.42-0.49 when set to “auto” in bios. When system boots and AC/DC is 0.08/0.049 system is not stable - avx2 stress test freezes the system or reboots.
My i9 14900HX Laptop had an Intel microcode update a couple days ago despite Intel saying these are not affected. Curious why it needed an update if it supposedly wouldn't have cooked itself anyway?
The emperor of mankind requires the sacrifice of 1000 14900K's a day.
The Omnissiah protects. 🙏
@@DanielWW2 The Machine God sets us free. 🙏
Until the Tyranids puts him out of his misery...
PRAISE THE OMNISSIAH
Lmao. Isn't it like 10,000 14900ks a day. All ln2 cooled to 7ghz.
The patch won't come until August because AMD is presenting their new CPUs on 31st July, and they want the benchmarks to look good. We can extrapolate from that that the patch is gonna drop performance severely.
Exactly. They're getting ready to open themselves up to a LOT of false advertising claims.
@@chrismay2298 As long as 14900K hits 3.2GHz on P-cores, 2.4GHz on e-cores, then they've hit the marketed base frequencies, everything above that is listed as a turbo frequency which Intel would argue is a "maximum target" that they would try to hit assuming power and thermal ability to do so, but they offer no guarantee of doing so. Unfortunately, I think Intel has the money and legal position to likely get away with standing their ground, eating the cost of any dead CPU's and then weathering the storm on the patched CPU's that hadn't died yet. While the right thing to do is to replace all of the affected CPU's, there are vendors (I believe it is Dell or HPE) claiming at least 8M units affected. If that were somehow only 10M total 13th and 14th gen K series impacted, Intel being willing to replace even the non-dead chips with whatever the next K-series is I'd guess would like cost them somewhere in the $3B to $5B range. They'll do the calculus, maybe replacing everything dead plus patching everything else costs them $1B up front and a few billion in future sales due to lost good-will, money people would argue that a guaranteed loss of $1000 today is more painful that a "potential" loss of $2000 tomorrow, I'm doubting they'll do the right thing here outside of ensuring their larger OEM's are given free stock to resolve RMA issues, while everyone else including those out of warranty will be left high and dry with a patch and a "sorry guys"..
I hope all the main enthusiast channels redo the Intel benchmarks after the patch.
@@KarolusTemplareVnot sure how many will/can take that effort to update reviews... since it will be after the AMD review, maybe best case they exclude the i9
@@Tarrasq-Eredar I would suspect that channels like Hardware Unboxed and Gamers Nexus will retest following the patch. Thing is Zen 5 X3D is speculated to arrive in September and if this patch really does take a significant hit to performance its going to be a awful look for Intel when those X3D reviews come out.
Reminds me of the initial Intel Meltdown/Spectre BIOS updates: “You may experience more unplanned sudden system reboots than expected.” - What?! I expect 0 unplanned sudden system reboots.
And reduced performance. Oopsies
As someone who overclocked their 2500K from Day 1 and ran it for more than ten years without degradation, this entire situation is BS.
I passed mine on to my parents. Still runs at 4.5 GHz since day 1! I even had it running at 4.7 for about 3 years without issues...
They had to beat AMD in gaming. They failes big time with this bet.
Bruh!! Ole school. Me too. Man, those were the days.
4790k overclocked, though slightly for 8 years.
No issues, updated to a 13400 recently, hope that model isn't in the danger zone, so far not according to all sources.
Midrange bang for the buck buyers win again
Its not BS, the issues with 13-14th gen intel CPU's are well proven with data to back it up by now. This has been done by basically all the well known reviewers and tech channels. And now intel has officially stated there are issues with them and says a "fix" via microcode will come. If it actually fixes it we will have to wait and see. Fix or not , the damage has been done and their brand will take another hit from it.
cannot wait for all the lawsuits coming from these "updates" :)
Can't wait to get my $2 settlement payout after the lawyers and businesses get their payouts
@@pseudonymous1382 two whole dollars? we gonna be eating well when that happens
@@marcogenovesi8570when you have ramen that comes with 3 seasoning packets
@@pseudonymous1382then don't be a member of any class action? And you can always sue yourself?
@@lost4468yt Nope, unless you're a millionare. Fees, laywers and few years of fight will drain you dry, that's most corpos' strategy, and it works damn well. Welcome to capitalism.
14900K used prices are gonna go way down. Better stock up on 7950Xs guys
Forget it, nobody will buy used 13th and 14th gen. Alone the "oh yeah, we fixed the 13th gen oxidation problem last year. We knew it was our fault, yet we still denied warrenty."
With that they killed the used market for their chips
I suspect the August update will prolong the lifespan a bit but not fix the issue.
It will also probably reduce performance, conveniently after the review cycle for the new Zen 5 CPUs.
It'll probably prevent further degradation, but all of the already dead chips have to be returned.
@@seeibe AMD is paying review channels to shill for their products and make them seem like more powerful in a first place so dont worry intel is still better in gaming and everything else
Om Nom please do not eat the CPU.
they will drop them all the way down to 5.4 max.
If that was my issue id adjust the operation temperature to voltage ratio. They know they have oxidation issues and temperature is playing a huge role if not the most important role
Rushed to the market products that were pushed to their limits in order to win some benchmark scores leaded to fail themselves. Who would know?
lol its like the xbox 360 red ring of death all over again.
I don't think Intel has enough physical CPUs to handle the RMAs, so they're going to run over the motherboard makers with their struggle bus until they figure out something else.
Although, I do credit this for introducing me to Frame Chasers, the UserBenchmark of CZcams.
"userbenchmark of youtube" is so accurate it hurts. lmfao
@@dirg3music if userbenchmark outright scammed it's readers by selling them 600 dollar unstable memory overclocks.
Ofc they'll throw Mobo vendors under the bus. They already did that . How is this a question?
@@TheHighborn ...it wasn't?
I specifically implied they were going to keep doing something they already were. No question mark ever appeared.
Bartlett lake is going to be used for the RMAs. The microcode update is just a stopgap until they can produce enough
Intel storing the current as a pseudo float is a semi-standard practice, when you want the certainty of ints but need more precision. Bit stuffing can make the maximum value a bit odd, but you see this a lot in older games. Starcraft 2, for example, stores hitboxes as a 5 bit value where the bits are quarters, halves, ones, twos, and fours, for a maximum of 7.75. Any value that's set higher in the unit editor will default to 7.75 under the hood. Intel is, apparently, using either an 11 bit unsigned quarter or a 12 bit signed quarter.
You don't see this sort of thing too often anymore, because bit stuffing isn't a thing that's done too often anymore, but the design constraints of a motherboard's BIOS are tighter than most programs.
It's called fixed point and it is an art.
If you actually analysed a bios, There is soo much padding waste... Just better structural design and MBs are freed
The design constraint isn't anything to do with data size. It's that VID is the format used to communicate between the PMU and the VRM controller, neither of which have an FPU or any reason to have an FPU.
Warranties are the law. So many people spending hours adjusting voltages and clocks on a broken CPU and getting nowhere. If it's faulty then get it fixed under warranty
Well, well, well, what do we have here? Last year, I wrote Gigabyte about an issue with some unusual high voltages being measured by HWinfo on my Z690 Aorus Pro.
Gigabyte's response was, please send a screenshot of that measurement but I couldn't catch that second that showed the 1.588 Volts being put on my 13700K. And after sending the screenshot, they said "we don't see an issue here...and we don't know nothing".
With Raptorlake being still new, I doubted myself and thought, I didn't observe that new architecture long enough to make an educated guess, that that wasn't supposed to be there. Since there wasn't any mention at the time anywhere, I thought, it was a fluke.
Self-doubt can put a real dampener on your life!
However, in my 28 years of system-building, I may make some observations over the years and this one struck me ass very odd, only to read about it a few weeks ago. It is almost like I told them so..
After the update, no processor will reach the turbo speed of the advertisement!
Agreed and they are waiting for mid August to get reviewed as is vs new amd
New patches just slows down even more and turns core off each time lol..
Like what Apple kept doing in their iOS updates for older devices..
they will but clock stretching
Ryzen 3700X had this problem on launch, never hit boost clocks, pbo would hard reset. You didn't have the crashing at stock, but it was false advertising.
@@JohnDoe-ip3oq that was fixed in 2-3 months. This has been an issue since at least 4th quarter 2023 and still no fix.
Out of all things Minecraft is eating 14900K(S) like potato chips - sometimes the universe just writes pure gold XD
Clearly Mojang did a bad job at removing Herobrine /j
We went from "Can it run Crysis" to "Can it run Minecraft server" LOL this timeline
While you are mining copper from a mine in Minecraft running on a distant Minecraft server, the electrons in the Minecraft Server's i9-14900K CPU die are doing some copper mining of their own. This happy symbiotic arrangement goes on until the CPU copper mine runs dry or there is no more room in the previously working silicon die for any more copper "nuggets" that those squirrely electrons have managed to bury and forget about in the surrounding silicon.
shoulda ran the server on the ole trusty G3258
I mean a g3258 would still use ddr3, so memory limited
If u could get a Z790 Mobo in a blade, u can do a crazy thing like 7600 96GB DDR5, disabled E cores and HT and overclocked 14600K
@@krazybonnie5523 not a bad idea but you are throwing out reliability and power const are going to be pretty high. Just not worth it
I had a g3258. Upgraded to 4790k with liquid metal for $200 at that time
Still have my g3258 sitting on my shelf. Good times overclocking it.
@@krazybonnie5523 and STILL have the same problems the 14900ks also have, super fast degradation.
just get a super sleeper Ryzen 7950x system lmao
Damn it feels good to be a 7800X3D owner
Run two myself. Not a bad middle tier cpu. But just like the “exploding” x3d issue back in the day, this to shall pass. It’s the two cores boosting to heaven at 1,5 causing this. Lock your cores and be done if you own a 13-14900k. Or rma your cpu if you have issues then lock em. They wanted to be on top of stupid benchmarks and it’s coming back on em.
...middle tier @ half the wattage and faster than 14900k in gaming. Half of these Intel CPU's will fail sooner or later according to everything we've seen so far. The "exploding" AMD CPU issue was a motherboard problem with mainly ASUS and only affected very few people. That Intel stock is going to go way down.
@@96kylar WHAT HZ SHOULD I LOCK THEM AT?
@@96kylar ah yes, a middle tier CPU, just like my favorite middle tier GPU, the RTX 4090
@Coliflower185 well, yeah. The 7800X3D isn't the best CPU they make, you know. The Threadrippers and EPYCs of the world are way better for this use case, at the price of being much more expensive.
For these types of servers, Ryzen/Core IS midrange.
I’ve recently updated my motherboard bios and find all the update does is seriously neuter power to my 13900k. My chip barely ever hits 5ghz on multicore loads now usually floating around 4.7ghz and I am seriously pissed. This isn’t what I payed for, and I’ll be having a few choice words for their rma department.
This is the first example of this I've ever encountered, the updates were released early July so its not something that's just dropped. Might wanna check your thermals.
@@tmsphere there's a beta bios available too
Technically this is what you paid for and well above it, even. The 13900k has a (guaranteed) base frequency of 3.00GHz for the P-cores, everything above that is considered "turbo" and is not guaranteed at all. Look up the spec sheet at Intel Ark.
The new ASUS Bios cut multicore performance by 50%, it's insane they expect people to be okay with that.
Im glad I switched to AMD couple years ago.
Insane Levels of customer screwing. They push the fix beyond AMDs release to look better at the charts. After that their CPUs will get decreased performance. It also pushes the issue beyond warranty. So the degradation is slowed down and the people who are affected then can stick their CPUs into their rears, because Intel tells them it's out of said warranty now.
Well played Intel, well played. Deceiving, playing on time... that is perfect corporation shenaningans at it's best right in front of every customer. And they will go away with it with just a slight slap on their hands.
I bet they even knew about the issue way before but sold their stuff anyways because the profits sould out the costs afterwards in their calculations.
Pure insanity if you really think about such behavior.
This is the most Intel thing Intel has ever Intelled
Reviewers should include a crash frequency / failure rate benchmark
They are accepting RMAs, WTF are you talking about?
That doesn't make any sense. RMAs cost Intel money, that's profit lost. If they knew about the issue they'd have fixed the microcode bug. If they knew some of the processors were going to degrade and need RMA they would have scrapped the processors instead of selling them in the first place.
People forget that this doesn't affect *all* their processors.
There's a reason it's almost exclusively the i9's.
If I had to guess it seems reasonable to be only those already running the highest voltages from the factory ( barely making the cut ) that are the most affected because Intel's buggy microcode is pushing even more voltage into them and causing the initial instability and eventual "threshold shift" in required operating voltage for maximum boost clocks.
@@zcomputerwizIntel has to RMA them anyways. Most of the Issues ppl do have now until in a Month will not spike from 10% now to suddenly 70%. The Damage on most CPUs now are done. It’s better for them to sit there with a „better“ Review against new Ryzen and 3% more RMAs, vs 3% less RMAs but a worse Review.
Right now even GN just said „he look… we don’t and can’t recommend those Intel CPUs but we will test them as released“.
If Intel would have fixed it Allready it would be like „they fixed it, it’s not a big deal (or it is who knows) but we would not recommend Intel right now“.
They are fked regardless imo.
1.5v is what CPUs used 20y ago, when the process was 130nm!
Even 32nm Sandy Bridge overclock was 1. 4v pushing it. These are far, very far from safe for a 10nm chip!
Even the E cores are at that voltage. Why!? No wonder people are reporting E core failures.
Even more, it is almost impossible the temp reading is reporting hotspots. This much voltage and clocks, 50W on a single thread, the core itself is now so small it should be close to Tj Max, not 65C. (or Intel has used a lot of dark silicon...)
Agree. For the latest 5nm process node, 0.8v should be the normal operating voltage. I bet current crop of processors won't live to 5 year mark. Such a waste of resources all for consumerisum to keep going on and on and on.
@@bozilla9975 My AMD chips are set to throttle at 65C, which is the universal way I found it doesn't really pump much than 1.2v
Zen chips dont seem to be far from top clocks but Zen 3, being a smaller process, generate many hotspots. Zen 3 also seems to push more voltage than Zen+, so it becomes worse.
A slow 5500 was pushing 1.35v, so single and dual core loads heat up much more than multi, where TDP limits rule.
It was a while back that people were reporting issues with 3600s but seems that was limited.
@@bozilla9975 pretty sure most of the TSMC 3nm stuff has a max safe voltage of 1.2v.
AMD and Nvidia run them around 1.1v or less.
@@zcomputerwiz AMD runs 5nm chiplets at 1.5v max boost voltage, which is kind of insane. I don't know what their next generation will run at but I suspect it's going to be a lot higher than 1.2v.
AIUI the actual voltage being fed to the cores is not 1.5V; modern Intel CPUs have onboard polyphase buck regulators (FIVR) that provides the actual operating voltage.
Buying a secondhand albeit unused 13900KS is the worst thing I've done in a while. I was like "CPUs rarely die, it's relatively low risk". Now I've got a degraded CPU with no warranty. At least I've got it stable at 5.6 although I had to underclock and crank the llc to get it stable. Weirdly at stock 1.4v at default llc it bsods at idle every few days. At 1.35v with the llc bumped up from 4 to 6 it's stable.
Yes. Those times are over it seems :(
Don't buy Intel garbage 😂😂😂😂😂
I am on the same situation but with an 2nd hand 14900K, back in november.
At start all worked like a charm, now is unstable. Had to low voltage to 0.85v @ 5.5 GHz. Seems pretty rock solid now, my nerfed down 13900T now at 550 bucks with no invoice at all.
There will be a recall eventually, submit a case to the FTC
@@peterpain6625 its fine with amd.
their cpus literally blowing up were actually the fault of motherboard manufacturers (asus) feeding the cpu too much power.
Maybe Intel accidentally activated their next planned phase to make more money: The CPUs are only good for a certain amount of computation then you have to change ‘em - just like car tires!
Considering how AMD is competing so hard lately, that's not a good idea
Yes, it's the new Rent-a-Ring bus.
To Intel AMD and Nvidia . Can we have chips that are not pushed to the limit. Can we just have one generation where the fab improvements are all used on efficiency . K thanks .
It would be nice if one of the CZcamsrs that are freaking out about this had an actual example 13900k or 14900k that are degraded to test it.
I can sell you mine, and you tell us the truth.
I have a Ryzen 5800X, which I have run as stock, overclocked, and undervolted/overclocked with PBO. I have used this CPU to run a technical minecraft server which occasionally experiences INSANE loads while also running an instance of minecraft at the same time AND streaming it all to youtube. This server has been online pretty much 24/7 since early 2021, and I have never observed even the slightest degradation in performance or stability. This is completely unacceptable from Intel's part.
My brothers and I have run about the same settings and have had no stability issues so far, we're all running PBO with a negative 30mV all-core offset on 2 5800x's and an 7800X3D. I have my 5800X running a 200Mhz boost override and the highest single core voltage I've seen is about 1.419 at 5050Mhz IIRC. Not sure if we won silicon lottery or AMDs cpus just have a ton of headroom from the factory. Either way, I'm impressed. My little brother sure is glad he got the 7800X3D instead of a 14th gen lol.
cores for days!
With in 2-3 months? How tf have they kept this quiet this long.
server space. regular people who just play games on the computer don't care what happens with server hardware, even if it is the exact same hardware, they would just brush it off as "oh but it's servers, my pc doesn't run this hard" or something
either servers guys didn't make it public because nobody asked or they made it public but people who heard of it just didn't care so it didn't gain any traction and got buried under all the other news till cpus started failing in regular desktop computers
Servers go thru runhours much faster than consumers. Expect consumer problems in like another 3~5 months.
They have not. I've been seeing more and more issues for 2 years since 13 series launched lol. Just people said they were FUD because there are so many intel share holder shills on forums.
@@N4CRI mean, you also don't want to think that you paid 3-900(? IDK the latest Gen prices) for something that may fail in less than a year
@@ThylineTheGay Absolutely not, always assume everyone is a paid shill out to get you. Trust noone, much less yourself.
Turn down the CPU current limit (IccMax). Mainboards often set it way too high to 512 A even on Auto settings. Even Intel strongly recommends to never exceed 400 A.
Saw a 12900k die in the way you are speaking of a few months ago, the host was fully shocked, first crash seemed out of the blue, then after second one it wouldn't pass validation, think its been in the box since 12 was new.
Except that the 12900K chips are not affected by this particular defect, only 13/14 gen.
@@SteelBlueVision @
12900k is extremely similar to 13700k.and just because Intel has not yet acknowledged that is not affected doesn't mean it might not have an issue.
so a single 14900K is for about a year of minecrafting, 8 cores, one dies, disable and force next core etc.
jokes aside, if they are clocking it down to 5.7ghz, then isn't it the same what ryzen could perform?
buildzoid rant, fuck yes
Noice
@@weetjewatikwil1 DCS
incel warning lol
@@Jeff.55649 thanks, but we don't need a warning that you're entering the chat
I guess im gonna buy my first AMD cpu
Welcome to the club.
Well actually tbf I'm on an Intel right now but it has degraded over the years, and I really thought AMD was a powerhouse as a teen until all my friends convinced me Intel was better. Now after having Intel for years, on many different PCs, id rather have an AMD again.
I think Nvidia gpus are a lot more stable, but I still love AMD gpus. AMD ftw
@dementedpeep I think intel is definitely better just works with everything and I think intel will eventually get this corrected I just should have stayed on my 12900k as far as Nvidia yeah they are a powerhouse I'm rocking a 4090 it's a beast but Amd is definitely probably better value
@@timtoomuch Is not better
@lenoirx yeah I know I probably won't lol but while I was waiting on my new 13900k from intel I put my 12900k back in and ran perfectly
I have an idea: Manually set the voltage to 1.45v with a moderate LLC - disable all but 1 P-core - leave all E-cores enabled. See how fast it degrades with a looped stress test. I have a feeling that the degradation that's occurring stems from the cache. Similar to how X3D chips will degrade the fragile cache when overclocked.
outwendelled the wendell
8:07 About PL4, from the Intel 12th gen documentation:
The package power control settings of PL1, PL2, PL3, PL4, and Tau allow the designer to configure Intel® Turbo Boost Technology 2.0 to match the platform power delivery and package thermal solution limitations.
Power Limit 1 (PL1): A threshold for average power that will not exceed - recommend to set to equal Processor Base Power (a.k.a TDP). PL1 should not be set higher than thermal solution cooling limits.
Power Limit 2 (PL2): A threshold that if exceeded, the PL2 rapid power limiting algorithms will attempt to limit the spike above PL2.
Power Limit 3 (PL3): A threshold that if exceeded, the PL3 rapid power limiting algorithms will attempt to limit the duty cycle of spikes above PL3 by reactively limiting frequency. This is an optional setting.
Power Limit 4 (PL4): A limit that will not be exceeded, the PL4 power limiting algorithms will preemptively limit frequency to prevent spikes above PL4.
Turbo Time Parameter (Tau): An averaging constant used for PL1 exponential weighted moving average (EWMA) power calculation.
Notes: Implementation of Intel® Turbo Boost Technology 2.0 only requires configuring PL1, Tau and PL2. PL3 is disabled by default.
So this is a value that can't be exceeded because the frequencies are limited "preemptively".
In short: PL1/PL2 is a thermal limits, PL3 is for battery protection (not used on desktops), PL4 - electrical limit for VRM decoupling
vanilla runs fine, but large modpacks like atm9 and rlcraft completly randomly crash at 400A icc, but at 307a everything runs fine at about 5,4ghz
A? Amps?
@@ThylineTheGay yes. iccmax
Modpacks tend to run much higher server load than vanilla...
The fact we're even trying to figure out if a slightly out of spec power setting is causing a 1/4 of these chips to fail within a few months is just wild.
If after the update, none of the cpus will reach the advertised 6.0 boost, wouldn’t that make them liable for a class action lawsuit? 🤔
They will claim it's normal wear and tear .....bet or say it will * as long as Temps are below 40c ....another option is make it show 6ghz using null cycles hurts performance but looks good on paper
I'm no lawyer, but I think that if it says "up to" on the box, then nobody can sue for that because any number between 0 and 6 is "up to 6".
The only numbers the CPU's would need to reach to avoid losing a lawsuit are the base clocks for the P-cores and E-cores, which I think top out at 3.2GHz and 2.4GHz respectively, every boost is an "up-to", and all of the turbos and boosts specifically call out other conditions that must be met and the maximum numbers are always "up-to".
If you bought a 13900K 2 years ago and it never hit 5.8GHz, or more specifically, it never even hit 5.4GHz, even before this issue became a thing, Intel could tell you to pound sand as everything above 3GHz clocks are dynamic and best-case given conditions. They would not RMA a chip that wouldn't hit max thermal velocity boost clocks, or max Turbo 3.0 clocks, or any other turbo /boost clocks unless it was crashing at totally stock settings.
@@racerex340 My understanding would be from within the EU. Up to isn't allowed to deviate too much. Internet providers have been slapped because internet connections wouldn't come close to the advertised 'up to'. I think it's a 10% deviation that's allowed for some time. I imagine that can be applied to CPUs, too.
@@maskharat well, if it were 10% with thermal velocity boost, the Intel would need to maintain a single core boost of 5.4GHz for the boost window, and for any others like turbo boost 3.0 I think it'd have to be about 4.9GHz. I don't know how much the issue they're having is impact single/dual core boost performance, I thought it was mostly under multicore workloads, but if it is due to resistance, then hitting even 90% those single and dual core clock boosts might not be sustainable.
Peak content. Not even joking
Intel should offer everyone with a 13900k(f)(s) and 14900k(f)(s) models a complete refund or a free cpu of the next gen.
I rather take refund and change platform to AMD (currently in RMA process with intel to get my CPU replaced)
The "13th Generation Intel® Core™ and Intel® Core™ 14th Generation Processors Datasheet, Volume 1 of 2" PDF is no longer available. Volume 2 is still up. Are there alternative downloads anywhere?
I think the only solution for Intel is to release a Socket 1700 "Minecraft Xeon" that doesn't fail, does decent clocks and has like 8 performance cores. Then offer that one as a replacement for the cpus that fail. For now they're only trying to push the failure out of the warranty window.
You are describing Intel Bartlett Lake, which is slated for 2025.
@@Vhalikuporamee447 Bit late to the party ;)
Everyone is talking about power limits and stuff but I think it's the voltage. My 14900K when set to Auto voltages was hitting almost 1.7v. I manually set it to 1.32 and it's been perfectly fine. I'm wondering what the actual vcore voltage is when running.
1.7!!! 🤯
@@The_Man_In_Redmust be an asus mb, mine went to 1.78 and the 13900k died FAST.
Electronics engineer & PCB board designer here.
About AC DC LL values: if you set it to 1mohm it is not safer!
The CPU will ask for more voltage under the assumption of 1mohm*current voltage droop. If the motherboard power delivery impedance is lower (hence better) than 1mohm you will fry your cpu with higher than expected voltage
Easier to blame Intel than Minecraft admins I think, narrator says "nothing wrong" with chip destroying settings because it's an overclocking channel
Exactly, came here to say this. If the whole LL calibration works as I understand it does, and given buildzoid says that "setting LL slope to a lower value than the board capability allows will feed less voltage to the CPU than it expects", then the inverse is obviously true as well. I.e., setting LL slope _higher_ than the actual capabilities of the board will thus feed _more_ voltage to the CPU than it expects - which sounds exactly like the recipe for a disaster?
Hello,
You seems very knowledgeable on the subject, is there's any ressources you would recommend to learn the propers methods for tweaking a cpu?i am looking to undervolt a 7800x3d.
Cordially.
Even if the die is perfectly symmetrical, the power plane in the MB and substrate is not, so maybe the shorter-path side is getting enough extra voltage to degrade the odd P-cores fastest.
And this is why you undervolt your parts on day 1 (whether CPU or GPU, there are only gains to be made from this.)
SIde note: I love it when people are not aware how CPU-hungry Minecraft can get, especially on mods. That game is no joke - it is, after all, a single-threaded simulation game on the server side. To add insult to injury, the game runs a simulation 20 times a second - the load is not constant, and the resource usage fluctuations will definitely cause some heavy power level switching and adjustments twenty times per second. (!)
Spec-wise those Intel cpus look like godsends for the workload. Pity they fail that badly. I built a minecraft server for the kid of a friend from "old parts" and the coils in that thing a whining like crazy ;)
For some time, I was running a mod on MC that splits the simulation so each dimension gets a thread. Not perfect, but at least Twilight Forest doesn't lag when someone was doing overworld shenanigans. I am considering adding it back and removing AE2, for that mod conflicts with Dimensional Threading Reforked
Supposedly, Arrow Lake will have a max VID that is even higher.
I get sick of Intel adding "Lake" to everything. They sound like they have no idea what their doing
@@Nunya58294 It's just a naming convention, like AMD uses Italian cities.
Supposedly they give it a higher tjmax as well...
I see absolutely no way this could possibly go wrong
@@tomhsia4354Literally "this is fine" meme lmao
The ring failure makes sense indeed. I didn't think this way before.
This is the analysis we need on this topic
I appreciate the insight. There are a lot of creators talking about this that have no clue what is actually going on. At least you have something to add to the conversation even if it still needs a giant asterisk. 😂
Can you illuminate us with that great asterisk?
Could you create a video about undervolting 13th and 14th generation Intel processors? Specifically, which value should we modify: Actual VRM Vcore or Global core SVID? And should we undervolt in offset or adaptive mode?
How does the 200 mV VID offset get triggered? What if it accidentally got enabled for a fraction of a second by the ucode? That would occasionally zap the cpu with say 1.6V instead of 1.4V
This is exactly how my 11900k I just RMAd was acting. It was less than a year old. New one tried giving me 1.7v @ 5.2. Have to keep it under 4.9 to stay under 1.4v
The fix won’t fix an already degraded CPU. Factory OC bites manufacturer in A55. Now thousands of reviews and benchmarks are invalidated, but almost nobody will bother updating them for those that will still be buying these CPUs over the coming years.
Of course a software/firmware patch won't fix degraded CPUs.
@@ActuallyHardcoreOverclocking If my CPU is currently passing Prime95, will the new microcode protect it from going bad?
@@ActuallyHardcoreOverclocking Let’s hope Intel find a way to look for degraded CPUs so that customers can get a refund, or a replacement.
@@ActuallyHardcoreOverclocking They try to push the failure out of the warranty window. Nothing else.
This was really interesting, this is a type of load that can be used as a controlled test.
If these CPUs are getting degraded and damaged by excessive voltage spikes, then the 'fix' that lowers the voltage back to safe levels is going to come with a steep frequency degredation or it's going to destabilize a lot of CPUs in the field.
The fix specifically means disabling the 1-2 core thermal boost which is what is pushing the 1.5+ volts through the CPU. Without that the CPU works fine. But yes, its disabling a feature that the CPU was specifically sold for.
So tell me you've never undervolted without telling me you never undervolted a cpu.
If The CPU is ran at stock, it will automatically adjusts its frequency based on the voltage it's fed. They will not get less stable, but boost clock will decrease to accommodate for the lower boost voltage.
It's never the CPU. The CPU never fails...until it does.
Weirdly, I had a 9700k that went bad, randomly one day was unstable at stock, even like -700mhz on all cores, it was able to run stress tests a bit longer but eventually BSOD.
The cpu was also hosting a lot of modded minecraft at that time. Was told it was a rare 9th gen bug and intel ended up offering me a refund. Always found it weird because cpus are meant to be super rare to fail - and i never played with unsafe voltages and temps would always be below 80c.
They wanted to be on top of stupid benchmarks. Let two cores boost to the moon at 1,5 and here we are. Rma your faulty chip, get a new one and lock the cores and be happy forever. Cheers.
Intel's done
Class action is the only way now. Intel couldn’t compete effectively (as in power efficiently) with zen 4 so it ran raptor lake far too close to the sun. And anyone that has tinkered with Intel K CPUs from at least skylake knows that some motherboards pump FAR too much voltage on “auto” settings. My msi z170 board was notorious for this so I learned to “undervolt”. The same has been true for every subsequent Intel CPU I’ve owned. Comet lake. Alder lake. And now raptor lake. It’s not only msi. Gigabyte and ASUS are notorious for high voltages as well.
The auto and optimized default settings can be too aggressive. Especially for a CPU boosting to 5.8 GHz or 6 GHz on one or two cores.
So it’s about time that Intel got its sh*t together and clamp down on the auto/default settings. It’s like they make the CPUs but don’t actually test them in consumer motherboards. Also these bugs in microcode in the eTVB and now voltage should’ve been caught in QC and QA. But if you look at the amount of bugs in Maple ridge thunderbolt 4 firmware (early maple ridge NVM versions did not support hot plugging certain alpine ridge TB3 devices), it’s clear that Intel does not have a coherent and quality QA/QC process. And that is the root cause of the problem. Bugs happen. But robust validation via a robust QA/QC should catch show stoppers like ‘instability/degradation due to higher voltages than spec.’ If they had actually used RoG motherboards to test raptor lake in the factory before release, they would’ve seen the high voltages.
My mother always told me from a little boy: an ounce of prevention is worth more than a pound of cure. Intel’s reputation is suffering. A robust QA/QC is the prevention that Intel needs so as to build and maintain customer loyalty and support. At this point I’m considering not ever buying another Intel cpu ever again.
Gamers Nexus' video has some really strong arguments to suggest this is a manufacturing defect that no microcode or patch can fix. Intel just screwed the pooch on a manufacturing node they should have plenty of experience on and are trying to cover their tracks.
"oh no it's 1h long"
Me: MOOOOOOAAAR
Yes, despite of the recent fashion of "shorter is better" on CZcams, we can listen to someone 1 hour long.
I grew up in a time when the gates were so big, there wasn't even measurable variance that would make binning a thing. Why not happening to AMD? Because chiplets. This is the ultimate end of monithic chips that can turn into either a i9 or an n100 depending on how it comes out of the oven.
Pretty sure it's not happening to AMD because they didn't overvolt the living hell out of their processors to be able to keep up with their competitor
@@kekoraaaa they don't have too. $#!+ chiplets go into the $#!+ bin, very simple way to maximize yields of good silicon. Intel needs to save whole monolithic dies, which is why they bin them so aggressively and have a HUGE inventory of n100 class chips. Each could have been an i9 but only four efficiency cores and the integrated graphics and quicksync (etc) came out of the oven. Ever wonder why you can get mini PCs for as low as $25ish per core? The tail end of Intel's monolithic process will mean dirt cheap low end minis for years.
@@kekoraaaaidk if AMD, but am5 mobo are sticking 1.5v on idle right oob
according to the transscript you said the word "like" 516 times. impressive!
she's a valley girl
Someone need to CONNECT THE DOTS.
1- Intel had problems in production, specifically in the coating of anti-oxidizer
2- Oxidation adds RESISTANCE.
3- CPU wants power,
4- finds resistance,
5- ask for more power....
6- and ded!
Intel says about Microcode, yeah sure.
That was a problem already dealt with in 2013, was a fault in the production process that's been amended.
@@tmsphere so the current problem still persists??
NO refunds
Intel stated it was only 13th gen effected in 2023, so late production runs(?)
Explain why 14th gen has this issue then?
@@tmsphere 2023*
Modern Intel CPUs (since 10th gen or so) applying some hidden 1.5 ratio on specified AC_LL so in order to get requested VIDs close to fused VIDs you need to set AC_LL / 1.5 from DC_LL/VRM_LL.
Or if you follow "standard" AC_LL=DC_LL=VRM_LL rule then your CPU will be constantly 1.5x overvolted depending from current it draws
I love this topic ahhhhh, only one question do you think I should upgrade to the patch if I’m on a stable bios since December? 59 p 45 e 50 ring LLC7 undervolt svid best case scenario ever exceeds 1.35 and never have had a stability issue
Not all CPU’s were affected, but a good portion of them are and it happens relatively early on if you have an issue, so far.
@@PAIN-ot4cj from all these years overclocking learned that cpus face hard wall overclocking them. voltage wall*
For example. Your cpu runs at 1.2v (load voltages*) 5ghz . then at 1.25 5.3ghz .and then needs 1.35 to run 5.4ghz. you don't run 5.4ghz if you like your cpu!
If your next step in frequency needs 4 times the voltage it took the previous step. you have a wall infront of you.
That wall needs exotic cooling to overcome and not harm your silicon.
The end
@@ChrisGR93_TxS yeah this was 6ghz for me all core core that’s why I stood on 59 it was very easy to run
Thanks fellas I thought so, the voltage algorithm only affected boosting behavior on stock settings and degrading them with super high voltage spikes. Since I locked my cores and voltage I saw no problems with stability ever. Won’t be upgrading till I run into some windows or other troubleshooting problems due to security or devices with my system
I think 1.35V is probably safe long term. Though Chances are if you back it of to 5.8 and 4.8 ring you won't really notice a perf drop but you should able to significantly reduce the voltage.
If it's so memory intensive, then minecraft server perhaps should run on R9 X7950X3D? Same thread count and clocks, four times the cache?
Or at the very least, 7800x3d for the extra cache and reasonable clocks at reasonable cost.
It actually would be a good match for a Minecraft server, just need to find a host provider that uses blades/servers based off it. Supermicro does sell them and guessing they will be selling a lot more AMD based systems in the next few months.
@@philipa1246 then probably a self made setup like miners did, but this time mIITX AM5 boards stacked on some rack 😂
Wendell/Level1 Tech indicated that there was a significantly higher failure rate for servers running 4 DIMMs of mem. compared to those using only 2 DIMMs. Prob. just inviting inevitable doom faster. Also, Intel denies oxidation on 4th gen. How long before this is retracted?
They're already retracting by not making a general recall. They'll push a microcode update limiting max voltage requests, but beyond that they said RMA.
HOW ABOUT THOSE WE BOUGHT 2ND HAND AND DOESN'T HAVE ANY INVOICE OR PROOF OF BUY?
Didn't Intel have the same issue in some of their lower power Atom chips? I vaguely remember this because this killed a lot of sylonogy NAS systems.
"Oh no it's an hour long"
it do be like that sometimes.
Can’t wait to receive a $10 check in the mail in a couple of years for this mess
What happens to the ring & LLC frequencies and voltages when you're boosting a core? Are they entirely separate clock & power domains?
Bought my 14900KF 08.05.2024 so 2 months ago and I am in RMA process with Intel waiting for intels response of how to stop this happening to replacement chip. (rendering workflows and mostly gaming including minecraft)
CPU never passed R15 even as new with default settings now it no longer passes R23 that it passed before.
Would be interesting to know if it's just "normal" degradation of the ring (where a lower clock would temporarily fix the issue), or if it's actually a hardware failure
I have a 13900k and I realised that if you simply disable c states so it doesn't boost the p cores or the ring, it stopped the crashing in games (Also I don't think I have ever seen it boost to 5.8 on the p cores or 5 on the ring when gaming because the task scheduler generally tends to spread out a load to more cores. so games run 4 or more at the least)
Yeah but at that point would the math have been different enough for you to buy AMD at the time of purchase?
@@TheHighborn the 7800 X3D was not out when I bought my 13900K. I bought it about 1 month after launch in Nov 22.
Yea this is what I ended up doing with mine ages ago which is probably why its still running fine.
@@TheHighborn I bought mine in nov 2022 and i didnt want to dable with dual ccd problems. But in hindsight, 7800x3d is way better for gaming. Don't get me wrong I think every platform/cpu has some problems it's just that some have easy workarounds and some don't. And apart for some IMC/memory compatibility shenanigans initially (Thank you gigabyte for the hilariously bad qvl list) I haven't had any other problems
Rma that pos
lol 1.5v vcore. My 13700k runs at 1.22v during Cinebench at 5.6ghz 1.34v idle. It is super stable for me. Running 24/7, but 12h it is just idle, with plenty of decompression since 11.22.
I would instantly shut down my PC if I saw anything like this in a daily system.
But it is just terrible for anybody not checking their system when it is new. And that's totally on Intel. A new CPU should be always far below 1.4 if you want to use it 4-5 years.
What a terrible product the 900k is. Anyways will pick AMD next time. Once they support my 7200 DDR5. No wonders, to go over 5.6 the vcore required ramps up like crazy for 100mhz.
Same.... well I'm running 1.43v on vcore at idle. It droops to 1.36v under load and has been completely stable and running for ~20 months now. Rule of thumb is not to go over a 1.4v or in my case not much more. The cpu's that keep degrading are running stock insane voltages that are not sustainable. People who know how to OC/undervolt these chips aren't running into issues.
When Asus looking stellar compared to how you handled the situation, you fucked up.
So continuous 1.5v looks to be a problem and 1.4v is probably fine. Thanks for the info. Glad my 13900kf doesn't need 1.4v for 6GHz single, even at idle. Good to get numbers out there.
Intel hoped for the best with uber binned chips with 13th and 14th gen to compete with AMD now they're reaping and will probably have the gall to still blame consumers for running their chips at their advertised specs
So 1.72v is the vid implementation limit. It is nothing about the safety limit. Board vendors may misinterpret this limit. It is bad on Intel side.
I had my first chip fail, but I blamed myself because I limited to 300 watts, in fact I think Buildzoid led me here to 300 watt limit. LOL, I don't blame you, bro
BUT, on 2nd chip I did not make that mistake and limited to 262 as it was a sweet spot where cores won't clock down, and also did a proper undervolt, like my first CPU
So far so good.
I smell that Law Action Lawsuit is coming against Intel, and I speculate people will get a $50 paycheck in mail similar to when I owned AMD FX-8320, and their 8-Core dispute lawsuit and $50 showed up at my door 3 years later.
I am not celebrating this situation for Intel, I rather have them fix the issue on next CPUs and continue honoring RMAs the way they did with my CPU.
AMD has been guilty of similar things, but the issue is that Intel is not addressing it properly, which will likely result in lawsuit, because this is....AMERICA, Harold, AMERICA
How about the rest of the world who are fucked also? And who the fuck cares about 50 bucks after 3 years when you needed your right new chip to work as intended for years since day one?
I disabled turbo boost on 14700K since its in my unRAID NAS. It runs a pretty high load, encoding, vms, dockers going 24/7. Hopefully the short periods of time I ran this CPU before disabling turbo didn't cause too much issue. Hopefully without turbo the voltages are way lower - I'm assuming they are since the 14700K loaded in my unRAID system usually maxes out around 53C.
Hope that helped. If not there will be a lot of cheap 7000 Ryzen out there when 9000x3d arrives ;)
@@peterpain6625 Heh, me too! Good point, 7900's will probably drop a bit more on the used market
what is your recommended LLC setting on a gigabyte aorus elite X z790 then? low or medium? my 14900KF has been running on unlocked power limits and its still rock solid for 7 months straight.
I've managed a few hundred servers and blades for the last 10 years or so. I've had maybe 2 CPUs die in that time span. I say maybe, because both the board and CPU were swapped after failures. CPUs dying is extremely rare, to the point of it not being considered during troubleshooting.
Gonna be hilarious if intels 15th gen runs on a diffrent socket (for no reason yet again) resulting in anyone who bought and trashed their 13/14th gen cpus just being shit out of luck. Buy a new motherboard, cooler, cpu, everything, nerd.
If someone has to do that, it'll go AMD. FOR NEVER COMING BACK.
Question: if it using intel defaults why its fried? Vid exceed 1.5? Does it have Intel.default switch ?
Chips failing even when limited 5.7GHz is wild
6 months lifetime for a 700 doller CPU. That sounds pretty resonable to me. And i'm not just saying that because of my Intel stock holdings. No siree!
Intel employee spotted. Sell them now while you still can, or you will crash, like your CPUs do.
Hi, nice video
I was wondering about phase counting. Are they always real phases ?
12 phases produced by a 6 phases driver chip with phase multipliers is better on not than 6 phases chip with better mosfets ? Do they delay correctly the phases to have a better output with those 12 phases ?
E-cores going away might be related to the oxidation issue as its more of a sudden once and done thing
They''re not "gone" the 15th gen will have E-core and non E-core editions which will definitely eat a lot more power and be more expensive, they arent getting rid of the mixed model.
@@tmsphere if you dont understand the comment you should watch the video further or watch some older material. This comment is about current cpu's losing e-cores due to failure.
Can you provide a link to the " Patch now " ? I have read the stuff from Intel and they said the microcode to fix these is coming in August and I know there have been a lot of BIOS updates from MB vendors but I can not find anything about a patch that is available right now. Thanks BZ!
Yeah wait until August for now enforce all intel limits as per your CPU.
You sound like the type to go flash a BIOS not meant for the motherboard, then go to the reddits to go "HELP I NO WORK". idiots
@@tmsphere Built my wife a new system with the 14700k and set all P cores to 4.7ghz and E cores to 3.4ghz. No more than 1.12 VID and a total of 144w of power at 100%. I will not touch anything more until the new microcode is in place but going back to BZ's comment, I still do not see a patch now available since most MB venders have beta bios up
Would be interesting to understand more about how AC & DC_LL works. What are the benefits of lower or higher values and why? I assume closer to 0.01 AC_LL that the voltage would match the VID table of the cpu if the MB LLC was set to no vdroop? Just trying to understand as Intel recommends that AC/DC LLmatch LLC? I’m confused lol
TBF, this happening on a pretty much a reference board is absolutely horrible. But how can you even get a reference board from a major manufacturer anymore? All of them set the limits to ridiculous heights to just one up one another. Are there even motherboards that run in spec out of the box?
From what Intel is saying the issue at this point is with their microcode, not necessarily the boards themselves. The CPU is requesting more voltage than is intended and destroying itself at full single core boost clocks.
If you want to be safe, limit max boost and max voltage.
@@zcomputerwiz I know, I watched the video. Just a general observation
In a 14nm white paper from yonks ago I saw a MTBF vs voltage curve. Basically it showed MTBF got worse by a factor of 10 for every 100mV higher Vcore. I wonder if theres a similar thing for their 10nm, ahem, "Intel 7". If I owned a RL I'd be limiting it to 1.35V or something and just seeing what clocks it can sustain.
As it is this sheetbox Acer laptop has undervolting blocked and the best I can do on this 13900HX is to limit turbo multipliers. I actually did this from early on as it was yoyoing off thermal and power limits (100W long power) so bad that letting it go up to its stock 5.4ghz and clock back down, was slower than limiting to where it throttles to, around 4.3 P / 2.8 E. Now I check VIDs, they're 100mV lower at idle/low load and 150-200 mV lower under part load, interestingly enough basically no difference when it throttles down under full load (just slower...)
The most recent Bios on my Asus Maximus Formula Z790 was "Updated with microcode 0x125 to ensure eTVB operates within Intel specifications". Fortunately i set my P-Core boost ratios to max 5.7Ghz so i would never have experienced that behaviour anyway. But i have noticed that the previous Bios " Still Intel Baseline btw" i was seeing a consistent 5.7Ghz in games, now it fluctuates from 5.4Ghz upto 5.7Ghz.
Now they're talking about another Bios to fix irregular voltage requests. So am i going to see further nerfing to gaming performance?
Waaait - you do not know what PL4 is?
It is the absolute powerlimit. By default NOT enabled it is a preemptive limit the CPU will try everything to stay under. While the CPU is allowed to go above PL2 for up to 10ms, cracking PL4 is a shutdown.
As for the voltages - if memory serves me right intel specified something like +-5% or so maximum tolerance. And similar ranges for the memory-controller (rather narrow).
Until someone measures the voltage on that blade then I would assume the reason your contact has a higher failure rate than other companies is because Supermicros blade’s core voltage is higher than the others.
Would love it if you measured it. I can’t think of anyone else who would actually connect a scope to see the “before and after” for these upcoming patches.
Edit: Apparently 30% is lower than others. Wow.
Well, on my asus strix z790itx board AC/DC loadline varies after every boot. AC between 0.07-0.12 and DC 0.42-0.49 when set to “auto” in bios. When system boots and AC/DC is 0.08/0.049 system is not stable - avx2 stress test freezes the system or reboots.
At this point these data servers are going to need to design a magazine to unload the dead 14900k and reload a good one 😆
My i9 14900HX Laptop had an Intel microcode update a couple days ago despite Intel saying these are not affected. Curious why it needed an update if it supposedly wouldn't have cooked itself anyway?