Intel says "Buy an overclockable motherboard that disables Current Excursion Protection, Set PL1 to 4000 amps, and your i9-14900KS may burn out"

[Kisai]

16 hours ago, sounds said:

While Intel's statement is written to hide the facts, I think the facts are simple:

1. Buy overclockable motherboard
2. Buy i9-14900KS
3. Overclockable enthusiast motherboard's BIOS default PL1 is "beyond Intel recommended limits"
4. Intel blames motherboard, but CPU is already toast

False advertising maybe?

I bought a new i7-14700k, the out-of-the-box motherboard configuration turns on optimizations of "ROG STRIX Z790-E GAMING WIFI II"

Notice what's default. First thing I did was "Enforce All limits" and then turn XMP off so I could update the BIOS. Cause it literately would not update the BIOS with XMP on.

 

Right beneath that:

 

Note what BIOS version introduces the Intel baseline profile:

 

Anyhow, I'm disappointed that manufacturers are still "cheating the benchmarks" after all these years. The out-of-the-box configuration should be the CPU manufacturer's settings. I don't know how these MB's pass QA checks being able to burn out the CPU.

 

 

[RejZoR]

I always found reviews of motherboards funny. Out of the box they all performed within 1% and even when all this crap was enabled, it was maybe 2%. I've not bought a single motherboard ever with premise of it giving me any extra performance. I always looked at that from perspective of what overclocking feature it offered to me.

[Mark Kaine]

16 hours ago, PDifolco said:

but when the BIOS default is too aggressive fault is on the board manufacturer

yes, but not when the cpu manf didn't specify exactly  - as has been mentioned there are no "safe margins" only "specs" and these defaults are within those...

 

maybe the manf knew, who knows, but point is this is within specs, unlike (probably) some ryzen boards where manf just ignored those margins. 

 

This is on intel for not providing proper safe margins  - which is why everyone's up in arms now for them blaming board makers. 

[porina]

1 hour ago, RejZoR said:

I've not bought a single motherboard ever with premise of it giving me any extra performance.

If one manufacturer does something to get ahead, everyone else has to do it to keep up. So they end up all about the same anyway.

[Stahlmann]

It always looks bad when they blame the customer when the problem was caused by DEFAULT BEHAVIOR. I have always found MCE to be stupid because it drives power consumption through the roof while giving single digit performance improvements in multi-threaded loads. In most games, which are still mostly limited by single thread performance, it makes no real difference. But you still get the higher voltage and heat that comes with it.

 

AMD has a much better handle on their AIBs. Intended behavior by default, overclocking as an option.

[starsmine]

4 hours ago, RejZoR said:

I always found reviews of motherboards funny. Out of the box they all performed within 1% and even when all this crap was enabled, it was maybe 2%. I've not bought a single motherboard ever with premise of it giving me any extra performance. I always looked at that from perspective of what overclocking feature it offered to me.

I was thinking that for quite some time until I saw the trash fire that is mid range AMD mother boards now. (100-160USD) Like I would not ask of those boards to be amazing overclockers but that half of the boards in that price range fail to even run the higher end chips at stock behavior is abhorrent. 

[RejZoR]

2 hours ago, Stahlmann said:

It always looks bad when they blame the customer when the problem was caused by DEFAULT BEHAVIOR. I have always found MCE to be stupid because it drives power consumption through the roof while giving single digit performance improvements in multi-threaded loads. In most games, which are still mostly limited by single thread performance, it makes no real difference. But you still get the higher voltage and heat that comes with it.

 

AMD has a much better handle on their AIBs. Intended behavior by default, overclocking as an option.

MCE used to be dope because the multithreading clocks were pretty lax. Like, my i7 920 or i7 5820K were boosting really poorly. MCE just forces all cores to run at boost clocks and not just 2 out of 8. Those chips did it no problem coz they weren't that stupid high power out of the box and they had quite some unused reserve. Even when overclocked to 4GHz and beyond, they didn't consume stupid high power.

 

It's different with 13900K or 14900K because they already run high power as is without doing anything. MCE just puts it so far out of the sweet spot zone it's just crazy.

[AnonymousGuy]

Funny enough I smoked a 14900k with it just *existing* with all of the limit settings turned off.  My guess is that the default state sends way too much voltage at idle (like 1.55V). It eventually got to the point where SVID had to be set to "Intel Fail Safe" which dumps in even more voltage to get the thing stable.  And then it really easily hits TJmax.

 

I just said fuckit and put a non-K in.   Seems to have a way more sane VF curve and less fuckery of clock settings than you get with a K even when you're leaving everything on Auto.  There's probably a combination of Intel pushing too far and mobo vendors not bothering to define stable, validated default settings.

[Beerzerker]

12 hours ago, dogwitch said:

their a.i oc by  manf. aka boost clock and manual oc.

I understand that and that's why I said:

14 hours ago, Beerzerker said:

For most, you won't see big performance gains because they aren't setup to get them like the XOC guys are.
Mainly air and watercooling for the average user so you won't see alot gained as mentioned.

I know the difference because I am one of the XOC guys to know the difference.
For most here, XOC use doesn't apply - Instead it's done along the lines you've pointed out, if it's done at all.

[Godlygamer23]

On 4/29/2024 at 12:31 PM, BiotechBen said:

If Yamaha makes an engine at rates max RPM at 6000, and then it goes into a car and the redline is set at 9000rpm, and the bozo pulls up to a car meet and pins it at the red line for 20 minutes straight, is it yamaha's fault for it blowing up?

Who put the engine into the car? Yamaha, or some other manufacturer(including the 'bozo' mentioned in this quote)? Yamaha only makes bikes, so as it stands, Yamaha would be off the hook on the surface. However, if Yamaha had some internal documentation that said "officially it can reach 6000RPMs max, but many of the engines will survive 9000 for x amount of time", they may hold some of the blame, especially if someone made a car with that specific engine and contacted Yamaha about the true RPM limit of the engine. 

 

To the customers, the immediate manufacturer or individual would be at fault. However, that's only one layer of the onion that would be unraveled. 

[Sauron]

On 4/29/2024 at 5:54 PM, sounds said:

While Intel's statement is written to hide the facts, I think the facts are simple:

1. Buy overclockable motherboard
2. Buy i9-14900KS
3. Overclockable enthusiast motherboard's BIOS default PL1 is "beyond Intel recommended limits"
4. Intel blames motherboard, but CPU is already toast

False advertising maybe?

Overclocking has never been a guarantee. After all, if all or even most chips could reliably reach 3-400 mhz above the default they'd just release them with those values, or take the opportunity for a more expensive SKU.

 

As consumers looking to overclock, it's up to us to inform ourselves on what these chips can realistically tolerate, and (always) risk burning ours for the sake of a relatively small improvement.

[BiotechBen]

42 minutes ago, Godlygamer23 said:

Who put the engine into the car? Yamaha, or some other manufacturer(including the 'bozo' mentioned in this quote)? Yamaha only makes bikes, so as it stands, Yamaha would be off the hook on the surface. However, if Yamaha had some internal documentation that said "officially it can reach 6000RPMs max, but many of the engines will survive 9000 for x amount of time", they may hold some of the blame, especially if someone made a car with that specific engine and contacted Yamaha about the true RPM limit of the engine. 

 

To the customers, the immediate manufacturer or individual would be at fault. However, that's only one layer of the onion that would be unraveled. 

They used to be an engine OEM, may not be anymore? The engine in the XC90, which then went into the Gumpert Apollo was manufactured by Yamaha, and they and Toyota are teaming on hydrogen ICE powertrains.

 

And oh, I agree that there is multiple blame, in the engine analysis I would say maybe a better one would be Subaru with oil starvation on the EJ20? where even normal operation caused engine damage, and driving exuberantly only killed it faster?  Or maybe the THETA II GDI with Hyundai/Kia where they knew it was a flawed engine, and had manufacturing issues, and "normal" operation would lead to unexpected premature failure (ours was a great engine, and didn't have issues beyond the standard Hyundai oil guzzling).

[IkeaGnome]

Intel is now sending out guidance to motherboard manufacturers to enforce power limits more closely.

Intel wants Default Settings with PL1/PL2 at 125W/188W to be implemented by motherboard vendors by the end of May - VideoCardz.com

About freaking time!

[AnonymousGuy]

I still feel like the power limits are being used to indirectly clamp the frequency down...so it  lowers the voltage.  Reliable voltage spec on these things is 1.425V (like Nvidia, Intel calls it Vrel) but they needed more than that to hit the multi core turbo frequencies (AKA turbo boost) that marketing asked for. (yes marketing asks for specs and there's then negotiation with engineering whether it's possible or not and then there's more discussion about whether the wafer yields can support the demand at the various binning levels)

[RejZoR]

On 5/7/2024 at 2:28 AM, AnonymousGuy said:

I still feel like the power limits are being used to indirectly clamp the frequency down...so it  lowers the voltage.  Reliable voltage spec on these things is 1.425V (like Nvidia, Intel calls it Vrel) but they needed more than that to hit the multi core turbo frequencies (AKA turbo boost) that marketing asked for. (yes marketing asks for specs and there's then negotiation with engineering whether it's possible or not and then there's more discussion about whether the wafer yields can support the demand at the various binning levels)

I think the main issue are power spikes. Giving CPU basically unlimited power envelope can mean CPU can spike to +500W before it reacts to other limiting factors. All mobo manufacturers are just relying on temperature and amps being the sole limiting factors.