Regarding Intel i9-13900K and 14900K stability/crashing issues
HTML-код
- Опубликовано: 7 сен 2024
- going over potential fixes for these high end chips, mostly applies to asus board owners
NOTICE: I AM IN NO WAY SHAPE OR FORM A PROFESSIONAL IN THIS FIELD, I WILL NOT BE HELD LIABLE FOR POTENTIAL DAMAGES OR DEAD CHIPS, FOLLOW AT YOUR OWN CAUTION
pc specs:
Intel Core i9-14900K
ARCTIC Liquid Freezer III (420)
Asus ROG MAXIMUS Z790 DARK HERO ATX WIFI 7
G.Skill Trident Z5 Neo RGB 64 GB (DDR5-6000 CL30)
WD Black SN850X 4 TB
Asus TUF OC GeForce RTX 4090
Lian Li O11 Dynamic EVO XL (White)
be quiet! Dark Power Pro 13 (1600W)
Thermalright TL-C12C (120mm)
Thermalright TL-C14C (140mm)
Just by activating Synch All Cores you have fixed the problem.
What cooks these CPUs is when a single core, even for a background task, asks to boost to 5.8Ghz. Then, to reach that freq. it is given 1.55, or even 1.6v, depending on the individual CPU. THAT'S what's killing it.
Even though the chip has power limits, when only one core wants to boost, all of the power is offered. This is a feature just so Intel can get a good CineBench score by reviewers. It's totally stupid.
Lock those cores together and you're good.
u dropped this
👑
Exactly , i Feel like We Are Heroes By Doing Better Than Intel Itself . Lock That Shit And Live With Peace ! Cheers
Just to add to the statement, higher voltage allows for more current to run through the CPU cores to overcome resistance in the copper traces in the silicon. Better binned chips with higher SP numbers typically have more pure copper traces with lower resistance and thus require lower voltage to maintain stability. Resistance creates the heat as current (Amps) has a less straight path to flow through copper, voltage can over come resistance to provide more current flow if required.
It's the current flow and heat that kills the chip, the voltage is like a mechanical lever that supplies more or less current (Amps) for a given resistance level based on silicon quality.
I would avoid anything over 1.45V for Intel 10nm on single core boost, years ago I killed a 22nm Xeon 8 core running 1.48V OC after 6 months.
Also the CPU input Voltage (PLL), powers the voltage regulators inside the silicon that feed your CPU, the motherboard feeds the vCore rail socket pads on 1.8V which is then stepped down to the vCore configured. Going below 1.8V increases temperature marginally as lower voltage means that more current flows into the CPU from the backside socket/pads. I run 1.88V on the PLL, reduced temps by 2C under full load for the 13900KS, and undervolted the vCore leaving it at stock core frequencies to reduce temps and power consumption further.
WRONG. Eventually, it will continue to burn out.
How? If it's fixed?
Because people updated their BIOS MONTHS AGO and after the small band aid, it surely began to crash and burn again.
It's imperative to get off the Intel platform as fast as possible!
wow that's a great tldr
0:10 this dude finger is faster than his SSD🤣🤣🤣
Hey, you don’t need any of these settings, they are all wrong. You have a good CPU, just cap the best two cores to 57x and set SVID to Trained. And use your VF Table, you should have max VF below 1.42xV, so you don’t have to worry with this Intel issue. I have an SP106 and I run 5.7x, LLC4, you’ll have an excessive overshoot set to 1. If you have a flat line set to a high LLC and try with Vcore Voltage, but again, you don’t need this. You also don't have any problem if you increase the CPU LLC Level to 2 and for example set COU input voltage to 1.85V, for example if you wanted to stabilize YC at 8400/8533/8600, which is not possible on a Hero Use MCE Enable remote all limits to 90, set PL1 and 2 to 253W, ICCMAX to 320A/340, IVR at 1.5 and enjoy your CPU. This configuration that are you showing in this video is worse than a 13700k.
THE **FIX** PLUS EXPLANATION OF THE INTEL CPU PROBLEM:
This is not a mystery or drama anymore. From J2C's BSODs, Actually Hardcore Overclocking, + many more, we've now observed multiple examples of EXACTLY what the issue is:
The crashing and damage is caused when a single core goes into boost behavior. It can be two cores also, but it's the same problem. The CPU has a max voltage. But if you make ALL of that voltage available to one or two cores that want to boost to 5.8GHz, then they will get 1.55 or 1.6 volts. Whatever voltage is required to reach that frequency. That damages your CPU silicon. And that is an intentional design choice made by both Intel and AMD. It's not a manufacturing problem.
Why doesn't this issue affect Xeon chips? Because most Xeon CPUs don't have single core boost clocks. When all cores boost together, the max voltage enforced by the chip's microcode is distributed across all cores, and no single core can get crazy high voltage. Anything over 1.5v has the potential to damage that core. Or possibly the cache package strapped to that core.
The "FIX" is to lock/synch all the cores together, eliminate single core boost. This stops a single core from getting 1.6v. It's simple.
If you already have damage, you may have to reduce your all core max boost frequency multiplier, because the damaged cores can no longer reach those freq. and will crash your PC.
First, lock the cores together, no single core boost. I would recommend also putting a cap on the max voltage. Maybe 1.4v max, but because there's overshoot, personally I'd go lower. Not to mention the challenge of temps if all core turbo is too high.
Validate what works with your CPU, what is a reasonable match between freq., voltage and temp. That's it. You're done. And no more damage. No more shouting into the void.
The point of single core boost is stupid anyway. It's literally to get on the top of some reviewer's chart. But since both Intel and AMD allow their chips to do it, and we're at the limits of what the substrate can provide, we will keep seeing this.
Lock/Synch your cores together, so when Win Update suddenly runs, a single stupid core isn't provided 1.6v so it can reach 6GHz.
Problem solved.
The fix is to rma your degraded CPU and use upgraded BIOS which doesn't have this behaviour on the new one and everything should be fine moving forward
No issues here with my SP106 13900K. Just undervolt and lower clock speed. Do not use the motherboard built in AiOC, that is what is ruining CPU's. I'm running 5.5GHz at 1.34V and an offset of .05V. AiOC will run your CPU @ 1.45V to 1.5V+ which is crazy.
^THIS, INSANELY FAX
Funny you bought something and you downclocking and basically telling people not get what they paid for
Lol😂
YT channel "Frame Chasers" have good explanation about solution to prevent/stop degradation...
@@mrrexy4151JUFES IS KING!
altho the wild support is nice, the nolife copium comments is kinda...... depressing ngl 😅
If you don't have a LGA bracket i'd suggest that, got my temps down about 10 degrees under full stress
^this (its a must)
If your system is already crashing with Intel 13th/14th Gen you need it replaced. Im a hardware engineer and see many red flags with this behavior on these chips. Intel needs to replace all of them because large server companies are losing money & switching to AMD Server chips quickly. If I had one of the chips I would RMA it ASAP. Only bad things can happen to Intel from this point forward without fixing the problem. Thanks
Nothing can fix it, it's a hardware issue not a software
Can i get intel 15 gen for future...?
Will they fix issue for 15gen....? Please answer 🙏🏻
Bro you got an SP rating of 110! I’m over here with 96 sp, bloody Intel was using my chip as a door stop before it got to me lmao FML
so true!!! bintel moment ⛄️
I'm at 71 (when I last checked)
92 here, SP rating means fairly little from what I've seen so far unless you're direct die all-core ocing. What really matters is the IMC for everyone else
Shit my 14900k got an sp of 89…. I heard avg is 85-91 so you both got very good samples…
That being said my chip is 100% stable since I manually tune bios every upgrade and never use mb auto settings. I wouldn’t let vcore go above 1.35 (1.2 is default voltage) just something I have learned building pc’s for 20 years is you shouldn’t ever got .2 v past default voltage unless you want to degrade your cpu long term.
Currently running a fixed 1.25vcore LLC7 @ x57 5.7pcore/x44 4.4ecore/x45 4.5 ring bus. Capped at a max P1/P2 of 253w. Maxes out at 87c in C23 with 38k score.
Highest score in C23 is at 1.25vcore and 320w p1/p2 with 41k in C23. Daily use max P1/P2 is 253w for longevity.
My opinion: dont lock your cores. You have a really good cpu, sp 110 for 14900k is the best you can get. What is killing these cpu is the ring, not handling the voltage for p cores, e cores, and the memory controller. By your sp score, your vid table has likely very low vids for the p core frequency. I would let the bios on intel default, remove-enable limits and work on LLC, Dc and Ac LL. You will have low VID and will never go above 1.4v if not 1.385v based on your SP. You have a beautiful chip to lock at 5.0ghz.
amen, god bless 💞
although me personally 1.3 is already pretty high and anything nearing 1.4 will make me faint
therefore im trying to do some degeneracy undervolting (as low as 1.1-1.2x) as i got inspiration from a 13900ks clip from a year ago with it sustaining load at ONLY 1.15x, insane…
it’s not the i9 it’s all 13th and 14th gen chips.
My buddies kid just built his first pc with a 14700 non k. This was his birthday present and I got sick when I heard. I don't want to even tell them to flash the bios yet as I can imagine their power will go out mid flash. I have not seen many non k failure reports. What makes you say all 13 and 14 are affected?
@@charlesg5085 I have a fix. Get a 7950X3D or 7800X3D, its cheap and fixes any stability issues.
for people that are afraid to not degrade the chip i always advise them to do a manual vcore cpu overclock, step by step and see how their frequency steps scale with voltage until they find out where the cpu needs a lot more vcore for the next step in frequency. Thats the only way to see if you're falling off the efficiency curve of your silicon quality and also you know that holding back a bit from that you can't harm the cpu (at reasonable voltages and temperatures)
For example. (prime95 small ffts stability test)
you need 1.2volt for 5.5ghz all core
Then for 5.6 all core 1.250v
for 5.7 1.3v
for 5.8 1.35v
for 5.9 1.42v
for 6ghz 1.52v
Your efficiency curve starts falling off at 5.8ghz and if you want to not degrade the chip you have to either run the cpu at easier workloads or and a lot lower temperatures. or never go over the efficiency curve.*
I can't imagine a silicon to not degrade at 1.5+ over 50c
maybe a rly old 45+nm silicon from 2000s (or fx amd😂)
this is what i did to my 13700k immediately when i bought my 13700k: disable hyperthreadding, set max core voltages to 1.35v lock down clocks to 5.5ghzP 4.4ghzE. thing runs like a charm and wont be peaking 1.4xx+v during idle.
thats the dream! 💞💝❤️🩹
intel be like: 1.5 is fine!
the reality: anything close to even 1.4 is mild panic zone
what settings?
any dx12 game would instantly crash for me but the recent bios update for my asus z790 e gaming wifi has seemed to fix the crashing issues. has run flawlessly for around a month now and i no longer have to force dx11 or undervolt the shit out of my 13900k
I have a i9 13900KF that I got right when it came out. The only thing I have done to it was set a offset of -0.060. I have never had one problem. No crashing, vcore never above 1.38. It's baffling to me, almost as if the more recent batches have an issue.
perchance, 10nm has such a huge gap for errors its crazy
you add 3 more fans for a push/pull on that 420 artic AIO get about 9-12C lower temps recommended if not.
Good job mate, thanks for the video.
Nice chip, good bin 👍
6:21 with 110sp chip it should use best case scenario for svid behaviour.
Unless the table is stupidly low or your motherboard has rly bad voltage regulation.
Intel beats AMD, good useful video thank you 👋
You have a golden chip there. If it were me I'd be running it no lower than 5.8ghz P core and 4.6ghz e core at 1.35-1.4v vcore. You should be very safe under 1,4v vcore and under 1.3v on sa. The reason some of these chips are running into degradation issues is because they are overvolting the chip. I've seen as high as 1.6v on vcore. Rule of thumb for the longest time with Intel is don't go over 1.4v. Also, no one should be getting an i9 if they don't have sufficient cooling.
amen 😇🙏
I'm not a pro too, anyway i copy some settings like Overclock TVB, voltage load line lvl 3, svid to typical behavior and other settings that avoided Overclock boosts, i set the chip to asus profile and from 37.7 on cinebench23 i get now 38.3, also single core boosted from 2337 to 2349, on voltages max i get for now was 1.43, temperature package cpu 83C
thanks for the support nonetheless
To me the moment Intel stated that there are oxidation issues with their chips...it was game over, you can not recover from that.
True but only on the 13th gen. so if ur on 14th gen atleast u dont have that issue
microtechnologies all have that problem wym, at the micrometer level those types of defects are to be expected lol
That oxidation story is completely useless as it would affect only a small number of 13th gen cpus. The problem is with i7 and i9 of 13th and 14th gen, some say even i5. Besides intel would not give a official statement using reddit.
Dude burned his chip so bad that it’s only stable at 50 ….
LOL? cope cope cope i been freq limiting way before the drama
haha xoxoxo-42 is jealous he can’t afford a good processor!!
Not sure why im here but i have been an intel user for years and just decently switched to my first amd just because i wanted to try something new and different and what do you know a few months after i build my pc i hear about all these intel problems. Wow i got some really good luck somehow lol.
In 2017 I bought a 7th gen intel motherboard with 7th gen Intel Celeron (2c/4t), the goal was to upgrade a few years later to i7 7700k when it was cheaper. But in 2020-2021 it wasn't cheaper, it was still more expensive to buy a 7700k than a Ryzen 3700X - 5600X ... That made me switch to AMD (5600X) and I don't regret it.
This isn’t about luck really, amd isn’t competent compared to intel, specially when it comes to single core
my I7-13700KF be pulling 300Watts under full load. too damn much wattage
Compared to the i9 13 and 14th gens, the i7's REALLLLLYYY got a boost gen by gen. 12th is 12 cores, 13th is 16 cores, and the 14700k has 20 cores. Wth that's insane power from an i7. If I went 14th gen I'd go with i7 14700. Cheaper and almost just as many cores as the 9. Probably more stable and manageable also.
@@beardeddragon9255 nah I’d go 13700k. The 2 extra cores on the 14700 don’t really make a difference. Plus my 13700K pulls 300watts at full load somehow. Waiting for that microcode update
@@cademckenna863 Can't you change that in IntelXTU? I have i9 laptops, one being a 13980hx and I modify it's behaviour via IntelXTU. Voltages, wattage, etc.
@@beardeddragon9255 I can I think I just have to update to the last MSI bios update they put out like 2 days ago. But heard that only drops it down 30W
@cademckenna863 recently I changed to 13700k. With asus tuf. Have not had 0 crashes yet with xmp ram 6400. And default at 5.1 ghz.
Benchmark on passmark cpu point was 46455. Overall top 98%
You can just HOLD the del key....
USB keeps sending check requests. No need to spam.
noted 📝
This not a fix in anyway. You gimped a i9 to run as an i5. Don't DO THIS to your CPU.
i dont understand these copium comments
The theory was halfway there.
What should have been done is set a manual core voltage of say 1.35v or whatever he's comfortable with, then up the all core ratio up till it's unstable, then back it off one.
Power limit and ratio didn't need to be so low.
The best fix is switching to AMD.
@@iLegionaire3755 None of them are looking too good right now.
4:10 lol i have like 45 - 52 °C Idle Temperature with a 360 AIO and undervolting
id be banging my head against the wall, I WOULD NOT sit happy with those temps idle 😔
@@sparkyexclusives but i have much better temperatures while playing Cyberpunk than before. Its very weard. Just like 72 °C average in 2K with a 4090 what means the 13900K have a lot to do. Before undervolting and a 240 AIO the temperatures were like 20 °C hotter in Cyberpunk but the idle temperatures were much lower. Ununderstandable :D Maybe because im using thermal-grizzly pads now. And i have like 33 °C in my room. I dont know
@@DeadCry1000 the problem with intel is that yndervolting isn’t going to help and it’s been known to make things worse
Wow why pay i9 prices for an i5 level of performance, if it does not work return it simple intel users need to punish intel for a defective product period. Imagine buying chocolate ice cream and getting vanilla 😂
ppl misunderstanding pretty hard like its literally a enthusiast chip but when u run into it blind (400a, unlim pl1 pl2, free single core boost, inadequate cooler) and complain abt it is simply laughable
@@sparkyexclusives Dude. You have just shown that you are the clueless one.
Even if you do ALL the right things, you will still get a burnt out CPU. Point in fact, you have server farms with 50%+ failures of these CPUs. You going to tell them they have no clue how to run their core business too?
The issue, confirmed by Intel, is a faulty algorithm in their CPU's microcode, resulting in incorrect voltage requests to the processor - ie causing excessive voltages within the CPU. And if your CPU has started to have instability issues, then it has started on the accelerated path to total failure. And any damage sustained by this over volting is irreversible - getting the microcode update is too late for that CPU. So yeah, it is not caused by the end user - it is fully on Intel.
@@sparkyexclusives cope
@@gbnq2513you also don’t know how important taking care of products is 🙄
@@user-wo3mq7is5t What an asinine statement. I am not even sure what you are implying?
You think you can mitigate what was an unknown problem at the time ( the faulty algorithm within the CPU microcode causing over voltage requests to the CPU)) by taking care of the 'product' ? GTFO How you do that then?
VID table issues, ringbus issue, singlecore issue, you'r screwed, 5ghz you just could have bought i5 lol whats the point then in spending more?
getting mad over someone preserving his sp 110 chip? understandable LMAO
If you just xtu and lock the cores to 5.0 instead through the bios and i dissable ecores and idle volts is 1.2ish
my bin is still better 😈
@@sparkyexclusives dang silicon lottery forgot to mention its a 13900k
I had to watch this at 1.5 speed
me too 😌
Dude you should press f5 then load base default you are all mess up there
bro think he cookin :pepeskull: LMFAOOAOAOOAOAOO
Ahh Intel, the reason Intel wanted to boost the processors power is to stay on the top of the leader board in tasking and gaming no matter what it takes to be in #1 spot even if it heave to sacrifice the stability and life time of the processors, they will do it.... 😂
The problem is that Intel didn't came up with new architecture at all that should be power efficient and good performance but they just used last generation and boosted the power limit and increased the voltages to squeeze as much as possible from the cpu to be called a new generation.... 😂😂😂
amd has been doing single core boosting algo way before them, therefore the copycat wars
about 14th gen, theyre just highly binned 13th gens, everyone knows
maybe arrow lake will provide a new architecture? possibly no more monolithic
And AMD, NVIDIA and Apple are at the top of the Silicon leaderboards in 2024, NOT Intel for this exact unresolved issue by Intel. The patch isn't a fix, its a 30% decrease in the performance you paid for from Intel.
Spicy
110 sp is insane
SP 110 IS CRAZY
This new bios bricked my 13700K haha
PL 2 253 w intel spec , u sure u know how to use google? i dont believe u can hold 5 ghz on full load with 125 w ,at this point we wasting money buying a i9
just for temps, not clock 🥶
Even an i9 that’s held back is still better than an i5 or i7 of the same generation LOL
a non held back i9 in not much better then a i7 in gaming
8 Raptor Cove performance cores in i7 and i9
i9 got better bin so u can push the clocks higher take that away is worse than i7 and e cores useless for gaming u got 8 p cores and are the same in gaming workloads is at parity most of time. PS i take any day 12 p cores over 24 hybrid bullshit
in tweakers paradise that you said NOPE .you can actually go and set max vcore limits so your motherboard doesn't auto vcore shoot to the sky for no reason. your settings are rly bad optimised for performance, like a LOT . your llc at 1 have massive vdroop and terrible transient response.( 3to 5 depending the workload has the best transient response) also the settings for vrm efficiency has nothing to do with how much voltage the cpu receives but for how clean , fast and how stable will come. Your settings are for your motherboard's lifespan that actually never have issues if you got a decent motherboard. mosfets have temperature min limit to 115 if you check it😂 you can't set it lower!
learning every day 🤧🙏
@@sparkyexclusives watch buildzoid
This has to be a troll video. I understand you want to protect your cpu but you literally just chopped its legs off and crippled it with those settings.
who wouldnt wanna protect & pamper their god bin? 56 is already a great enough clock, although voltage might be high for some ppl, preferences ig
You don’t lose too much performance plus for a little bit more stability, it’s worth it 😂
Load the bios update for Asus microcode 0x125 to ensure eTVB operates within intel spec dropped my vcore to 1.341v in bios. Also sets other preameters 253 wats 400 amps etc
should be 307a shrug
@@sparkyexclusives 400A is correct for Extreme profile on KS chips. In practise you'll never even get close to that value in applications (or the lower 'application' value) as the voltage requested via the VID table will always make you power rather than current limited. Asus Z790 Hero running 2402 with 0x125 microcode with a 14900KS at LLC6 and 7200MTs RAM on a MoRa 420.
can you show us the VF offset page in extreme/AI tweaker. i wanna see your 5.6/5.8 and 6.0 GHZ VIDs in bios. thats a crazy cpu at 110 sp
Just sell and buy ryzen.
im gettin tired of these bait+copium comments baha 🥶🤧💯
Dont comment if you can’t afford intel 😒