Whats crazy is that even at this severely nerfed 188w power limit is still higher than the max power draw of a 7950X3D

So as consumers were we lied with false advertising ?

The old bait and switch! Too hot to handle! Hot time in the old town tonight! Some like it hot! LOL!

Here are the tests with 'Baseline' profile by Quasarzone. Differences are clearly visible. It looks like that MT rendering and content creation workloads are affected more.
https://quasarzone.com/bbs/...
https://uploads.disquscdn.c...

Wow, 10% on 14900k is quite a bit. So now on average 15% slower in gaming than 7800x3d?
What is crazy the 14900, 14700 and 14500 non k lose extremely much performance!!

Gaming is not affected. Hardware Unboxed tested it yesterday. Have a watch.

MT workloads are affected. Again, HUB will have an update on this soon, as Steve is re-testing 7950X3D against 14900K in all possible MT workloads.

HUB tested with the wrong PL though on his latest video about it, Intel is enforcing 188W which is the 'Intel Default Settings' he tested with 253W which is the "Baseline Profile" from certain motherboard manufacturers, the performance hit is going to be even greater.

188W is Baseline profile and 253W is Default, from what I can see in Intel's annoucements.

Intel benefited from benchmark proxies.

What I would like to see is a news outlet go out and buy a stable motherboard like a supermicro X13SAE-F and compare them to all the gaming motherboards.. Intel used to sell their own motherboards, but no one bought them and they had weak bios update history. Gaming motherboards were always THIRD PARTY motherboards.

Even my Z590 when I got it, the default setting was "AI Optimized". Which isnt the default Intel settings.

Should be set to default to begin with....

LoL "AI" optimized...

That's ASUS' fault. They have been doing this for over 20 years. Back in the day there was "Load Setup Defaults."

When I saw the temps and voltages the i7 and i9 were working at, I knew I would get almost the same performance from my 13600K while having much superior efficiency, too, the biggest winner this generation for sure!

Yep

Since when was there a PL4? When does that apply?

PL4 is for transient spikes.

Has anybody created their own custom profiles to ensure stability?

If so, what values did you choose?

I didn't need to, but I undervolted with load line calibration set to minimize vdroop and it does a better job of keeping high clocks during benches. (I use LLC 6 on Asus which is second from minimum vdroop)
Otherwise my chip would thermal throttle. If you undervolt, you have to keep the volts still high enough to keep the chip from crashing. The most likely place for this is when vdroop can drop it 10s to 100s mv under loads in the 300w+ range. Motherboards weren't ready for that much power consumption. Also my chip needs increasingly more volts, relative to the volt/frequency curve set by Intel/Asus, the higher the p-cores get over 5.4 ghz. This efficient limit frequency number likely changes from chip to chip. Personally I have a ramp of added millivolts on my per point voltage offset for frequencies above 5.4GHz. Also, with my undervolting I found adding 20-30mv to my L2 cache voltage helped, but I don't think this fits with the general narrative and might just a quirk with my particular 13900kf.

If you combine the two characteristics: 1. Motherboard's power delivery allowing too much vdroop at "default" settings at the (until 13th gen) extreme power draw of 300w+, and 2. CPUs that are designed to run past their efficiency limits and need more volts vs their volt/frequency curve the higher their clocks are past their efficiency limits, you have a compounding scenario pushing affected CPUs into a situation that the cores are not getting enough volts for the clocks they are running and crash.

But by addressing these issues one may find their CPU getting light single core volts over 1.5v which I think is too much for daily use. The volts can be capped at 1.5v if your chip needs that much for your desired top clocks, and if your chip needs more then maybe you should weigh potential issues vs the rewards of keeping the single clocks that high.

If you look at the common fixes noted so far they seem to be different ways of addressing this. Power limiting limits vdroop. The Intel failsafe SVID behavior floods the CPU with volts. Jufes has a recommended fix of limiting your few core boost to all core boost if you have problems which avoids the out of efficiency range clocks as much as possible.

Good question.

I was thinking of a new custom spec:

I'll call it "High Performance Spec"
- PL1 = 125 Watts
- PL2 = 250 Watts
- PL4 = 375 Watts
- _______ iPL2 = 200 A
- IccMax.App = 300 A
- _____IccMax = 400 A
- Tau = 65.500 Seconds
- Turn On = TVB (Thermal Velocity Boost)
- Turn On = Voltage Reduction Initiated TVB
- Turn On = CEP (Current Excursion Protection); Applies to BOTH (IA & GT) variants
- AC/DC Load Line = Tune for minimal V-Droop.
- Max Temp Limit = 100°C

Do you think this kind of spec makes sense assuming you have the Cooling to back it along with the MoBo & PSU to power it?
It's supposed to be a more sensible compromise between Intel's (Performance & Extreme) Profiles.

why do you want to tune LLC for minmal v-droop? vcore is supposed to droop under load. going from idle to load with such an aggressive setting that your vcore on average isnt drooping anymore almost always means that on load change the VRM overshoots the voltage target and hammers the cpu with voltage spikes.

let it droop with a medium setting and compensate for low voltage in with an offset elsewhere - raising LLC should be the last step when you need just a tiny bit more voltage for load when you sit at a medium setting to begin with.

if your vrm is beefy enough you can raise switching frequency for reduced ripple, which also helps with ripple induced load switching instability.

id also reduce the 100°C temp limit. high temperature is one key ingredient to degraded silicon, with high voltage being the other.

So Medium LLC?
What about going with AMD's 95°C Maximum Temp Limit then?
If 95°C Maximum Temp Limit is good enough for Zen 4 Silicon, it should be good enough for Intel as a form of Safety Margin.

check buildzoids youtube, he released a video talking about settings he tested and found working well.

I already watched all of it, but BuildZoid has different settings then what Intel recommends, but he also mentioned that he doesn't have the reliability data that Intel has.
For however little that Intel Reliability data is worth given that Intel got themselves into this mess on their own accord, I wanted to see what others have done on their own.

i'd keep single thread load vcore below 1,45v and core temperature below 80°C and adjust clocks accordingly, and to be sure not some weird workload sends the juice flying id set a power or amplimit as safety net slightly above the target regular power draw.

Good. 253 W is obscene power consumption for a consumer CPU anyway. It will also put pressure on Intel to deliver as much performance as possible within the ~180 W power envelope for future CPUs.

Finally. Would be even better just force limit somehow all theese at 175w max pl2 and no pl3/4, and thats it. It was absurd anyway how much power theese was sipping…

No PL4? Are you sure?

Lower 14900K performance, launch 285K with better results comparatively... profit? 🫲😗🫱

Core Ultra 9 285K = PL2 188W? 🤔

Wait, even if these are the new profiles, they're just that, profiles. These are all profiles put out by Intel, and they are therefore stock spec or allowed by Intel. Therefore not an overclock and covered by warranty as they still fall within their profiles. So you can use baseline, performance, or extreme, to your liking. Actually this makes it easier for the end user to select a profile, or a profile that works with their silicon. Awesome news! Great for the console regulars, but enthusiasts will still go in an tune for max performance.

Good luck with Extreme profile and warranty.
*edit: clarified below - some users may still return CPUs and ask for exchange after running extreme profile, as prelimianary tests published earlier this week show that not all CPUs run stable at 253W.
https://wccftech.com/only-5...

No luck needed. Intel advertised it as a stock option so there's no discussion about this. It's in their own datasheet and there is no asterix. It's what Intel themselves used in testing.

I should have been more clear in expression. My apologies. I meant that extreme profile is more likely to degrade some CPUs, as recently reported by tech media, and users need to use warranty to return/exchange it.

Please cite proof that Intel's recommended and new profile of 253w voids warranty or degrades the CPU. Otherwise you are just talking out of your A.

- I did not say that warranty is voided. Read more carefully
- preliminary tests of hundreds of CPUs have been already published and re-posted, including 253W. We are waiting for more tests from around the world.
https://wccftech.com/only-5...
- and, temper your vulgar language

The extreme profile isn't used at all in practice. None of the profiles are. Almost every mobo runs with no limit at all. I haven't seen any reports at all about the extreme profile so I'm curious what you're talking about here?

All the reports I've seen is about instability at no power limit but Intel doesn't test their chips for that which is why it's stupid to have mobos run specs like that. It wasn't a problem before for 12th gen and earlier gens but it's one now.

From May 31st new profiles will have to be used officially. Intel has published the three profiles within warranty, so that mobo vendors cannot add any secret juice anymore and dump CPU replacement cost on Intel if something goes wrong. It's biting them both financially and in reputation, so they had to act. Intel had silently allowed mobo vendors to do Wild West with auto and unlocked settings 'within the spec', which has resulted, as we can see in recent weeks, in growing number of degraded CPUs and returns, damaging their reputation for system stability. There is no doubt about it.

Here is another report that was not published here. Large sample of CPUs was tested in various profiles:
https://wccftech.com/only-5...

This is GOOD! 253w is more than enough for the CPU. Going beyond that is just stupid. And now they're providing actual profiles for users to select in the BIOS and use based on their own cooling and needs. This whole situation is actually a good thing, and finally Intel moving forward will ensure motherboard vendors use profiles they approve and then let users go in there and tune beyond those profiles if they want. But doing so and crashing or ruining the CPU is on them.

I said in the other post that Intel is trying to get their ducks in order, finally.
Hardware Unboxed is now testing all MT workloads with new profiles. That will be interesting to see, as there is expected hit in those worklaods. Gaming performance is not affected.

Good news is that were at the EOL for LGA 1700 as many like to point out in their talking points and reviews. Moving forward in a few months with Arrow Lake, we will have power profiles out of box, you can bet on it.

They'd better get their act together and sort out this mess once and for all.

How can a MB change CPU default profile and dont let the costumer knowing... 90% of the people dont know how to use bios...
And this is intel fault to... Because this happen a long time a go...
I have 13900k without limits... But to put the cpu safe from more than 2 years i change Vcore voltage etc etc... We need to do little tunning to making in safe! But to 90% of the people they dont know...

I mean it worked for a long time. It's only come back to bite them with the 900k skus. Seems like the 700k's are doing fine. Probably because they added a bunch of extra E cores or something.

Gamers Nexus pointed this out years ago, that certain mobo manufacturers were doing this. Particularly Gigabyte and ASUS were the worst for it.

Intel very much knew about these issues for a long time now but it was convenient to stay the way it was to have better Reviews, they're merely going to change for the simple fact that Arrow Lake is a new uarch (with much newer nodes) this time around and not a rebrand of the 12th like 13/14th was, this will also make Arrow Lake look much better against the previous generations in the near future, this is a win-win for Intel (they blamed motherboard manufacturers too, of course) and consumers get shafted in the face yet again, thanks Intel, true Snake Oil indeed.

This could very well be the case.

Let's see if this is true.

Curios if these Temp issues are effecting those that use that de-Lid their CPUS or used the Aftermarket CPU Bracket. My 13700k dropped 16 degrees with the aftermarket bracket.

Intel loyalists have been shafted big style.

LMAO, get your reduced performance for free by Intel, LOL

Buying the "Intel just works out of the box" experience isn't really what its shaped out to be, especially with a 'flagship' product.