Intel’s high-end CPUs Have Degradation Issues?
Vložit
- čas přidán 10. 05. 2024
- Testing various power profiles here: • Checking Intel’s Power...
Article 1: hwbusters.com/cpu/some-though...
Article 2: hwbusters.com/news/motherboar...
Support us through Patreon - / hardwarebusters
🐶 🐱 🐭 🐹 🐰 🦊 You will find the items we review through the following affiliate links 🐻❄️ 🐨 🐯 🦁 🐮 🐷
► Amazon US - amzn.to/3v4D5IG
► Amazon Canada - amzn.to/3vcUleR
► Amazon UK - amzn.to/3HRETMR
► Amazon DE - amzn.to/3BQAQfQ
► Amazon Spain - amzn.to/3BNSFwo
► Amazon Italy - amzn.to/3Wv2QxE
► Amazon France - amzn.to/3jkgUfd
► Banggood - bit.ly/3Jk2fsN
🐥 Our Site 🐥
www.hwbusters.com
🐧 Follow us on Social Media! 🐧
Facebook: / hardwarebusters
Twitter: / hardwarebusters
Instagram: / hardwarebusters
Discord Server: / discord
🐼 SUBSCRIBE to Our Channels 🐼
Greek Channel: bit.ly/3rccnNU
English Channel: bit.ly/3PJGeYf
#Intel #CPU #Default settings
Music track: Chill Vibes by Pufino
Source: freetouse.com/music
Free Music Without Copyright (Safe) - Věda a technologie
If you want to support us, please consider becoming our patrons. www.patreon.com/hardwarebusters
Best ATX v3.0 & PCIe 5.0 Ready PSU Picks 2023: hwbusters.com/best_picks/best-atx-v3-0-pcie-5-0-ready-psus-picks-2023-hardware-busters/
Best Case/Heatsink/Radiator Fan Picks 2023: hwbusters.com/best_picks/best-case-heatsink-radiator-fan-picks-2023-hardware-busters/
For shopping at Amazon please use the following affiliate link to help us do more reviews: amzn.to/3v4D5IG
Join our discord server: discord.gg/pv9rYXqm9w
Our site: www.hwbusters.com
The short answer is: Intel allowed the motherboard manufacturers to push the CPUs because it helps them to get higher benchmark numbers.
If they didn't, they would have a hard time selling CPUs to compete against Ryzen.
YES the unlimited power mode is a problem (and I'm saying this because last year I had a 13900k that idle'd at 80 degrees and definitely degraded)
however some manufacturers are not only doing this, they are also setting the AC LL and DC LL wrong.
on my asus board they had the AC LL at 0.5 and DC LL at 1. after messing around with this for awhile I found that using an AC LL of .6 and a DC LL of .6 is much better, and this is much better than the other setting in the bios (auto) which was AC LL of 1 and DC LL of 1 (the reason that the performance was bad was because when the AC LL is not equal to DC LL its misrepresenting the voltages)
just manually set your AC LL and DC LL guys and it will fix most of your problems.
EDIT: you can't just run one benchmark and think your CPU is stable. my CPU ran with a AC LL and DC LL of .4 without crashing in cinabench but had problems in Unreal engine games until it was set to .6
EDIT X2: Watch the video from "ACTUALY HARDCORE OVERCLOCKING"
Lol Intel does not have 1 official power settings as they have a variety of different power settings. Intel has been directly quoted that a MB with no power limits is still "in spec".
its not the "unlimited power mode" thats the problem, its the mobo manufacturers not setting AC LL and DC LL properly
Intel said that but have blamed mobos and warned against using "unlimited power mode" bit, despite their blind eye turned for years to default mobo out of spec operation.
The fact is they've screwed up and likely face class action lawsuit.
In the past a P3 1.13GHz CPU had to be recalled because it was unstable.
They cannot guarantee performance or extreme power settings will run well, because CPUs may have been damaged.
@@RobBCactive MOBO manufacturerers were setting the AC LL and DC LL wrong, and when they would release new BIOS updates, they would lower the AC LL making more and more CPU's unstable and seem like they have been degraded. MOST ISSUES ARE NOT BECAUSE THE CPU IS DAMAGED (however there are definite cases of cpus being damaged but that's only about 10-20% of stability problems)
For your system that you stress tested at intel factory defaults, did you undervolt it before beginning your stress test to find the minimum voltage it needed to operate at so that over time you could see the rate of degradation by checking how much less it can undervolt, if less, over that time span? If not, then the long term stress test is kind of inconclusive. it would also need to be compared to a system running unlimited iccmax and power draw.
That is a very good question, and most I think only look at how high the clocks can go rather than the lowest operating voltage at a given clock. To be honest, the last time I did such extensive low-voltage tests was on socket 754 for specifically a low-watt battery-backup proxy/NAT server. Everything after the fx8350/7700k kind of made it nearly impossible to specify a specific operating voltage and clock as things have been a bit fuzzy since then. For example, I disable all boosting/TPO and try to lock my R9-3950x to 3.6GHz and the thing wanders between 3.5 and 3.7GHz while the voltage is all over the place despite being set to 1.1v, the best I can do is keep it under 220 watts while doing 3Delight/RenderMan rendering, lol. That socket 754 on the other hand, was locked solid at 1.125v (instead of 1.4V) and 2.000GHz and pulling around 30 watts under full load. I get that some like to overclock to the point just shy of perceivable artifacting while gaming, there are other uses where you do not want "artifacting" like network routers and science calculations (SETI), lol.
I did something else. I use exactly the same settings measuring the Average TRUE power consumption, and frequency with various coolers, since this CPU is installed on my CPU cooler test station. If something was off, I would have spotted ASAP. I also test previous coolers from time to time to check proper operation. So basically I do way more than just unvervolt it, using extra hardware for the job (Powenetics V2).
The problem is not the power limits, its other settings MB set. CEP disabled and AC_LL not = DC_LL.
Except you lose performance doing that as raising ac load line will inevitably increase voltage and therefore power consumption and heat (even if it brings it back to intel spec) and CEP has been shown to have strange behavior hence why some mb manufacturers have it off by default. If you're going to raise ac loadline, might as well undervolt it yourself at that point and manually configure adaptive offset after raising ac_ll. And doesn't raising voltage shorten the lifespan of the cpu faster compared to running a undervolt?
The problem is not the power limits. Fullstop for now. Until further investigation. :-)
@@Noah_Aizenbut Intel bin chips to run at certain volts and use tables for each core. The mobo guys don't validate with large samples and silicon lottery losing CPUs are swiftly culled from the review program.
Running out of spec, may be unstable and if you turn off protections, you might have damaged your CPU particularly if thermal throttling was the constraint. When Intel encouraged board partners to innovate and damage CPUs those afflicted are not happy.
Mobos can lie about settings and measure them inaccurately so Intel have a problem to provide a universal solution.
It's bad QA practices and debacles are expensive and do reputational damage,
@@Noah_Aizen
- Core Clocks in practice are very simple - more Volts, more Clock. for a Clock to be stable, it must be provided enough Voltage (and not be disallowed to pull enough overall Power either, but that tends to be much looser).
- Loadline is a very important part of the balance of the system. it determines the Voltage shifting behavior from idle up to stressful load.
- raising Voltage itself does not necessarily hurt anything, ever, no. damage to the Transistors is a triangle of Voltage, Current, and Temperature. Transistors are damaged when they are drawing a lot of Current, at a relatively high Voltage for the Architecture, while at a relatively high Temperature.
one of these sides can be extreme without any perceptible damage if the other two are heavily compensated for. two of these sides can be significantly elevated if the third is heavily compensated for. think about how LN2 Overclocking doesn't have the Transistors dying within a few Minutes despite generally using pretty insane Voltage and allowing infinite Current - because the Temperature side of that triangle is inversely extreme.
additionally it depends on what the actual Voltage is. if you're not at a very high Voltage to begin with, increasing it slightly is still not very high and so it itself is not going to realistically ever hurt anything.
Great, thank you !!!
"b-b-b-but chips are fine operating at SiC TJ-Max", lol.
The average overall chip may be at around 100c however that don't imply the individual FETs are not spiking well past 200c and I don't think there is a way to measure that with things more complex than an IRF510, lol.
P.S. There is a huge difference between the silicon thermal limits of SOI and SiC, they are not the same thing, lol.
the power limits are not the main issue, its the mobo manufacturers not setting AC LL and DC LL properly
Fets? i hope you're not meaning the VRM on the Motherboard. it's not getting anywhere near that hot on... actually, on any Motherboard at all, not without trying to run a Computer literally inside a sealed box. for any normal configuration... the VRM on the Motherboard struggles to break 100°C on the low end Boards, and on the midrange and high end Boards, the VRM is more in the 50-75°C area.
if you mean the Voltage Regulator Circuitry on the CPU, that's also not getting anywhere near that hot since it's both on the Silicon and so connected to a lot of cooling while sharing radiance with the entire rest of the CPU, and that the onboard Voltage conversion has small losses. plus the VR on the CPU is covering routing misc rails, not vCore.
it's the Cores and interconnects that's generating almost all of the waste Energy...
for OOTB / basic usage scenarios, i think 8like 80-90% of the issue lies in the Current Limit. Boards/Intel/whoever are letting this be really nuts. a crazy Current Limit makes it easier for the system to not bump Voltage when wanting/needing to pull significant power in doing heavy tasks.
not to be confused with that "intel fail safe" Profile that some Vendors pushed to their Boards, nobody is supposed to use that one, it's an emergency Profile. that can do the opposite, and provide insane Voltage.
GHz race going too far?
I've had my 13700k oc since jan of 2023. 13700k oc to 5.6ghz on all 8 p cores, All 8 E core oc to 4.4ghz, Ring oc to 5.0ghz at 1.34v. Even the mem controller is good on my 13700k. 32g.b Corsair rgb vengeance 6400 cl32 hynix A-Die oc to 7200 34-41-41-83 at 1.45v. On a msi z790 tomahawk. It's been a absolutely great cpu.
Didnt notice sth in my 13700K, as well
im rocking a 13900KS @ 5.8ghz all P core / HT Off / 4.6 Ghz All E core. 50 Ring. @ 1.375 LLC6 MSI (Very droopy) on z690 edge with DDR4 bdie 32GB at 4266 CL16 super tight timings. Running standard 360mm BeQuiet Pureloop 2 FX with liquid metal and its been absolutely rock solid and fast AF whilst keeping very cooling. Best CPU i've ever had
Intel should have known better. Until their official based-on-actual events testing, I dont believe anything about it. We will see. Thumbs up for the vid!
I bought a 13700K because I wanted to underclock it 😅
then you will be ok ;)
I bought mine to oc and has been oc since jan 2023. 13700k oc to 5.6ghz on all 8 p cores, All 8 E core oc to 4.4ghz, Ring oc to 5.0ghz at 1.34v. Even the mem controller is good on my 13700k. 32g.b Corsair rgb vengeance 6400 cl32 hynix A-Die oc to 7200 34-41-41-83 at 1.45v. On a msi z790 tomahawk.
I have heard once someone saying they bought i9 CPUs because of their stability and performance. Really aged like milk
I don't regret buying an all AMD system whatsoever
Intel gives manufacturers some freedom as to how they design their systems, and yet everyone is harping on this and begging intel to regulate their partners. That's like you building your house and your neighbors complaining to the government that they don't like your house and begging the government to make your house compliant to their houses. Some countries have revolutions over this kind of thing...
Ι don't regret buying 3 all Intel systems whatsoever either.
@@KiSs0fd3aTh "all" Intel? I have never bought "all Intel" So you bought the Intel GPU's 3 times? Wow.... How do you like those graphic cards? I only use Nvidia gpu's, but to each their own. I have only used Intel for the last 15 years. Currently 14900k/4090
@@Noah_AizenExactly! Intel allowing mobo vendors tons of freedom is the most pro-consumer move possible. Locking things down is awful for consumers, vendors, and performance. It can backfire sometimes but I'll take unlocked products with tons of user features with the risk of a few issues over a heavily locked down and extremely boring product.
If you don't like tweaking your bio's settings just get the non k parts. They are stable and non affected.
Used both Intel and AMD CPU's and I can tell you the experience and stability is better on Intel.
So far, no problems have been reported with 13600k or 14600k.
Yep, nothing is reported for them.
they're capable of encountering instability too, but since their target Clocks are more reserved, the Chips are able to pull less Current, the Customer is less likely to be running fast Memory configurations and/or have very high end GPU's - all of those sorts of things do make it a lot less likely to happen on the midrange Chips, yes.
I'd love to see someone do stability / stress tests across the 14th gen line as I have a feeling they are all going to die earlier as Intel has pushed the clock too hard, its just the 900k SKU is the canary in the coal mine.
I am running tests on my 13900K, which has the same design as the 14900K. Will report my findings into the next week. That said, so far I didn't see any issues in this CPU, although I did push it to the max, since it is used in my CPU cooler testing station.
@@HardwareBusters the power limits are not the main issue, its the mobo manufacturers not setting AC LL and DC LL properly
Chiphell ran extensive stability tests in high end 13/14th gen and found half the samples failed on 1 or more tests.
@@RobBCactive yes because the MOBO manufacturers are setting the AC LL and DC LL wrong
@@AncientDragonMusicWhy were protective CPU management features turned off? It has under-volting and use of thermal throttling as power limit as contributory factors too.
The real story is how Intel wanted it both ways, official specs and tweaking for favourable review benchmarks on selected samples juiced by 3rd parties.
In the past mobos have lied about BCLK frequency to default overclock, or measured values inaccurately, so doing no validation and blaming those forced into creative performance boosting, simply isn't credible.
Only Asus problem 100%.