Intel Core i9-13900K and i7-13700K Gaming Stability Issues Linked to Power Limit Unlocks

[chrcoluk]

Article is funny to me, the media has kind of acted in a way that if you keep to the stock clocks and remove the power limits, you still in a safe configuration, but its still out of spec. Likewise of course if you tinker with tjmax.

So in short its only really news if the CPUs are crashing when you at spec'd power limits 'and' spec'd clocks 'and' spec'd voltages.

To the wccftech dude many of us already undervolt, you late to the game dude.

That's been said many times by many people, but I haven't seen a single motherboard that has unlocked power limits by default, so I have no idea what reviewers are talking about.

I'm not saying that reviewers are wrong, just that there's a gaping contrast between two realities here.

Not sure about unlocked to 4096w, but I am pretty sure my ASRock does a pl1=pl2 by default, is that spec on these new chips?

This page doesnt say e.g.

Intel® Core™ i7-13700K Processor (30M Cache, up to 5.40 GHz) - Product Specifications | Intel

Intel® Core™ i7-13700K Processor (30M Cache, up to 5.40 GHz) quick reference with specifications, features, and technologies.

www.intel.com

 

[dir_d]

My Cousin has this same exact issue and games would crash almost immediately if they were DX12 but everything else was stable. We lowered his P cores from 55x on his 13900k to 53x and that solved the problem. Even under volting at 55x would not solve the issue completely. He still had random crashes (less frequent) when I set his computer to PL125. He does not mind running the CPU 53x but even on the latest Asus Prime Z790-p wifi bios the stability is not there.

 

[MarsM4N]

These limits in question are set by mobo manufacturers, not Intel

Not only 13900K or 13700K, even the 14900K has stability issues even with STOCK settings.

Post from Reddit 1
Post 2
Post 3

Here we go again. ASUS, ASUS, ASUS, ... just another "Groundhog Day".

Tbh. the customers who buy those pointless "Super Duper Overclocking" boards don't deserve any better. There is like nothing to gain in gaming performance nowadays with overclocking, you just burn money & power for nothing. While companies like ASUS you throw money at are pushing it over the safe limit & laughing all the way to the bank.

Spoiler: Push it to the Limit

 

[as1234884]

Guys, I checked the Intel Spec, and there is no specification regarding the maximum value for PL (Power Limit). This means that the setting of PL by motherboard manufacturers cannot be considered their fault, as there were no restrictions originally.

By the way, Intel "Recommended" that PL1 is 125 and PL2 is 188.

 

[chrcoluk]

Guys, I checked the Intel Spec, and there is no specification regarding the maximum value for PL (Power Limit). This means that the setting of PL by motherboard manufacturers cannot be considered their fault, as there were no restrictions originally.

By the way, Intel "Recommended" that PL1 is 125 and PL2 is 188.
View attachment 335966

The page I linked to implies a max of 253w, but it isnt made clear if thats only a temporary boost limit or if can be run sustained.

Just checked again I think Intel now support pl1=pl2, judging by this quote, but its deffo max 253w. The quote and 253w, is from the link in my post a few posts up.

The maximum sustained (>1s) power dissipation of the processor as limited by current and/or temperature controls. Instantaneous power may exceed Maximum Turbo Power for short durations (<=10ms). Note: Maximum Turbo Power is configurable by system vendor and can be system specific.

The same link also states 253w is only permissible in turbo clocks, when not in turbo clocks the spec is 125w.

Your link with the recommendations is interesting and I think no bios vendors bothered with it. But check my link above it does show a 'max' of 253w.

 

[as1234884]

The page I linked to implies a max of 253w, but it isnt made clear if thats only a temporary boost limit or if can be run sustained.

Just checked again I think Intel now support pl1=pl2, judging by this quote, but its deffo max 253w. The quote and 253w, is from the link in my post a few posts up.



The same link also states 253w is only permissible in turbo clocks, when not in turbo clocks the spec is 125w.

Your link with the recommendations is interesting and I think no bios vendors bothered with it. But check my link above it does show a 'max' of 253w.

Yes, I have seen that the Turbo Power Limit is specified as 253 on the official website's exhibit page. However, it is uncertain whether Intel's specifications for motherboard manufacturers are based on this number or on the Spec figure. From what we can tell at the moment, it seems that Intel has not set specific regulations for PL for manufacturers.

 

[chrcoluk]

Here we go again. ASUS, ASUS, ASUS, ... just another "Groundhog Day".

Tbh. the customers who buy those pointless "Super Duper Overclocking" boards don't deserve any better. There is like nothing to gain in gaming performance nowadays with overclocking, you just burn money & power for nothing. While companies like ASUS you throw money at are pushing it over the safe limit & laughing all the way to the bank.

Spoiler: Push it to the Limit

He basically admitted their review process was inadequate there, these bios bugs mentioned in that video are the sort of things I want reviewers to find. They shouldnt really be making it to public stable version of bios's.

Probably the worst 2 bugs being gigabyte auto/bios optimised defaults failing to reset SOC voltage, and the 116C instead of 106C safety shutdown of X3D chips, really nasty bugs.

 

[Nucleoprotein]

Rad Game Tools is aware:
Intel Processor Instability Causing Oodle Decompression Failures (radgametools.com)

 

[dgianstefani]

Rad Game Tools is aware:
Intel Processor Instability Causing Oodle Decompression Failures (radgametools.com)

As far as we can tell, there is not any software bug in Oodle or Unreal that is causing this. Due to what seem to be overly optimistic BIOS settings, some small percentage of processors go out of their functional range of clock rate and power draw under high load, and execute instructions incorrectly. This is being seen disproportionately in Oodle Data decompression because unlike most gameplay, simulation, audio or rendering code, decompression needs to perform extra integrity checks to handle accidentally or maliciously corrupted data, and is thus likely to spot inconsistencies very soon after they occur. These decode failures then typically result in an error message.

As expected then.

 

[tpa-pr]

So if I'm understanding the issue correctly: this is mainboard manufacturers pushing past Intel's recommendations to (most likely) get the upper hand in benchmarks? Similar to what they did on the AMD side to cook the AM5 chips with the SOC voltages? Because if so, this is getting ridiculous. Why these mainboard manufacturers are allowed to get away with it is beyond me, at least with the settings being default.

 

[AusWolf]

So if I'm understanding the issue correctly: this is mainboard manufacturers pushing past Intel's recommendations to (most likely) get the upper hand in benchmarks? Similar to what they did on the AMD side to cook the AM5 chips with the SOC voltages? Because if so, this is getting ridiculous. Why these mainboard manufacturers are allowed to get away with it is beyond me, at least with the settings being default.

They get away with whatever they want as long as Intel/AMD don't have a standard set in stone for operating voltage like they did 15-20 years ago.

Not sure about unlocked to 4096w, but I am pretty sure my ASRock does a pl1=pl2 by default, is that spec on these new chips?

As far as I know, yes.

My 11th gen Asus TUF board (that was also criticised for unlocked power limits if I remember right), has Intel spec power limits by default, and unlocks them by enabling "Multi Core Enhancement". But when you hover over the function, the information panel tells you that it does exactly that, so 1. I don't understand the problem, and 2. You have to go into the BIOS to change this setting, so it's not default, right?

Every single Intel board that I've seen so far works in a similar way, including the Asus ROG Strix Z690-something that I used to build a rig for someone not long ago. So yeah, I don't see unlocked limits by default.

 

[chrcoluk]

So if I'm understanding the issue correctly: this is mainboard manufacturers pushing past Intel's recommendations to (most likely) get the upper hand in benchmarks? Similar to what they did on the AMD side to cook the AM5 chips with the SOC voltages? Because if so, this is getting ridiculous. Why these mainboard manufacturers are allowed to get away with it is beyond me, at least with the settings being default.

Motherboard reviews should stop doing benchmarks (kills that motivation) and concentrate on bios functioning/quality with the time saved. Problem is it requires a mind shift from the review industry as they seem just used to running benches and calling that a review.

 

[InVasMani]

Review sites could just start holding MB makers more accountable on their handling of bios defaults. That of course requires a lot more backbone than many or most often are showing given the nature of free review samples from these companies.

I mean it's a breeding ground for this kind of behavior when they risk not receiving review samples due to reviews of higher integrity and honesty that puts consumers concerns at the forefront rather than the companies hawking merchandise with questionable practices. Anyway it's a challenging dilemma overall to go around for both reviewers and consumers that in part rely upon them for insight.

In the end I'd say it's in part up to consumers being more vocal about their own concerns with these companies practices and holding them more accountable for them. We have the most leverage to make them take notice in the end, but just like the reviewers it involve as a certain amount of spine and integrity in part to stand up for what's right versus what is misguided.

The problem isn't that MB makers offer the flexibility to push the upper limits a bit nearly as much as it is about them actively doing just that at default and unaware consumers having a negative experience as a direct result of it. Many people aren't particularly tech savvy and those are the ones most vulnerable to these sorts of issues that crop up.

 

[efikkan]

My Cousin has this same exact issue and games would crash almost immediately if they were DX12 but everything else was stable. We lowered his P cores from 55x on his 13900k to 53x and that solved the problem. Even under volting at 55x would not solve the issue completely. He still had random crashes (less frequent) when I set his computer to PL125. He does not mind running the CPU 53x but even on the latest Asus Prime Z790-p wifi bios the stability is not there.

If you're absolutely sure that the BIOS hasn't set a power limit beyond the spec, then that's a defective chip and should be returned.
No one should settle for a "sort of stable" PC.
Any desktop PC should handle load for months without crashing, whether it's games, rendering, encoding or other loads, without any stability issue, otherwise it's defective. (of course assuming the system has a solid PSU, adequate cooling, etc.) There should be no need to undervolt the CPU, nor is it wise beyond a certain point. I would not tolerate new hardware working like described, and neither should you or him.

While I suspect most complaints to be user error (or bad default settings in BIOSes, if true), it is important that defective hardware is returned, as this serves as useful feedback to detect widespread defects.

Rad Game Tools is aware:
Intel Processor Instability Causing Oodle Decompression Failures (radgametools.com)

It doesn't surprise me.
Many have the misconception that if just some heavy workloads passes without any crash, then a system is perfectly stable over time, when the fact is that you can have lots of undefined behavior before it causes something critical enough to crash a program, driver or the system. Different applications can stress different parts of the CPU and cause different edge cases.

I've seen such behavior on overclocked systems; even when passing stresstests like Prime95, 3DMark and many games, running a fairly "minor" workload can still expose CPU or memory instability. Like running a medium sized code compilation will suddenly throw unexplained errors (just once every few hundred tries), and of course lots of random file corruption.

 

[MaMoo]

At such high power draw the control systems for these things probably can't respond fast enough, hence why this instability occurs.

People think CPUs have temperature/clock/power limits for safety reasons but that's not the only thing, they have to regulate the temperature because the properties of the silicon change and these things are designed to operate correctly only within certain parameters. Processors start to spit out incorrect results long before they're actually "damaged" by the heat.

Temperature is also a proxy for power draw, if two pieces of silicon reach the same temperature limit in the same time but one consumes twice as much power that means the control system for the one that consumes more power has to respond faster and is probably more prone to failure.

Thank you for this excellent insight. Makes sense to me.

 

[dgianstefani]

If you're absolutely sure that the BIOS hasn't set a power limit beyond the spec, then that's a defective chip and should be returned.
No one should settle for a "sort of stable" PC.
Any desktop PC should handle load for months without crashing, whether it's games, rendering, encoding or other loads, without any stability issue, otherwise it's defective. (of course assuming the system has a solid PSU, adequate cooling, etc.) There should be no need to undervolt the CPU, nor is it wise beyond a certain point. I would not tolerate new hardware working like described, and neither should you or him.

While I suspect most complaints to be user error (or bad default settings in BIOSes, if true), it is important that defective hardware is returned, as this serves as useful feedback to detect widespread defects.


It doesn't surprise me.
Many have the misconception that if just some heavy workloads passes without any crash, then a system is perfectly stable over time, when the fact is that you can have lots of undefined behavior before it causes something critical enough to crash a program, driver or the system. Different applications can stress different parts of the CPU and cause different edge cases.

I've seen such behavior on overclocked systems; even when passing stresstests like Prime95, 3DMark and many games, running a fairly "minor" workload can still expose CPU or memory instability. Like running a medium sized code compilation will suddenly throw unexplained errors (just once every few hundred tries), and of course lots of random file corruption.

Karhu for memory.
OCCT for CPU/System stability.

Overnight.

Motherboard power delivery, cooling, PSU (especially) all factors.

I suspect many prebuilt PCs that generally skimp on PSU and have crap BIOS settings, plus the population of amateur tuners, is the real background here (besides the known factor of MB makers simply having poor "stock" or "optimized" settings).

 

[P4-630]

Only issues with MSI and Asus boards?

 

[efikkan]

Karhu for memory.
OCCT for CPU/System stability.
Overnight.

It might still not be enough to ensure stability (if overclocked), as proving a negative is hard and some errors may only occur fairly rarely or under very specific workloads. Especially overclocked memory can be tricky to validate, while hardware defects in the memory itself are usually showing up quickly, timing issues with the memory controller are much harder to rule out (and is also something which degrades significantly over time). The best approach then is to disable the OC and see if the problem disappears, assuming the user can find something reproducible. If so, then you'll know there is a problem there. But overclocked memory is by far the larges culprit when it comes to stability issues in custom builds, and is included when I talked about "user error".

When Intel (and AMD) designs their CPUs, they've estimated safety margins to account for sample variation and degradation (under spec conditions) throughout the rated lifetime. That's why I'm saying if everything is set to stock (and there isn't any user error), and the system is clearly not stable, then it's defective.
Defective components happens, I've seen many cases;
Back when I built my server I used a server motherboard from Asus (P9D-C/ 4L). The first one was DoA, second one DoA, then third one has been working fine for 10 years.

Motherboard power delivery, cooling, PSU (especially) all factors.

I suspect many prebuilt PCs that generally skimp on PSU and have crap BIOS settings, plus the population of amateur tuners, is the real background here (besides the known factor of MB makers simply having poor "stock" or "optimized" settings).

For sure, most prebuilt "home PCs", "gaming PCs" and even baseline "office PCs" have cut down PSUs and cooling to the bone.

I've had a few "mishaps" with poor PSUs in systems I've built too, even though I've seen enough to know better, like my old Haswell system having a bad PSU shorting out so bad that plugging the cable immediately triggered the fuse. That system was never stable again, even with a new PSU…
Since then I've gone "overkill" with PSUs, not in terms of wattage, but only using top quality stuff.

 

[dgianstefani]

It might still not be enough to ensure stability (if overclocked), as proving a negative is hard and some errors may only occur fairly rarely or under very specific workloads. Especially overclocked memory can be tricky to validate, while hardware defects in the memory itself are usually showing up quickly, timing issues with the memory controller are much harder to rule out (and is also something which degrades significantly over time). The best approach then is to disable the OC and see if the problem disappears, assuming the user can find something reproducible. If so, then you'll know there is a problem there. But overclocked memory is by far the larges culprit when it comes to stability issues in custom builds, and is included when I talked about "user error".

When Intel (and AMD) designs their CPUs, they've estimated safety margins to account for sample variation and degradation (under spec conditions) throughout the rated lifetime. That's why I'm saying if everything is set to stock (and there isn't any user error), and the system is clearly not stable, then it's defective.
Defective components happens, I've seen many cases;
Back when I built my server I used a server motherboard from Asus (P9D-C/ 4L). The first one was DoA, second one DoA, then third one has been working fine for 10 years.


For sure, most prebuilt "home PCs", "gaming PCs" and even baseline "office PCs" have cut down PSUs and cooling to the bone.

I've had a few "mishaps" with poor PSUs in systems I've built too, even though I've seen enough to know better, like my old Haswell system having a bad PSU shorting out so bad that plugging the cable immediately triggered the fuse. That system was never stable again, even with a new PSU…
Since then I've gone "overkill" with PSUs, not in terms of wattage, but only using top quality stuff.

While this is pretty true, I can assure you if you have a problem with your memory tune, Karhu will error eventually. Usually within a couple of minutes. A 24 hrs stability test error free, practically guarantees your tune is stable.

This doesn't rule out issues from the CPU side of things however, or other components.

 

[below ambient]

Even on Direct Die ive had to undervolt my 14900k... Itll boost to 1.7v with unlocked power limits... 6ghz all core static overclock so im happy at 1.4v

 

[Block10]

Lowering your CPU voltage can be a great way to reduce heat and power consumption, especially if you have a CPU with as much power as the 13900k. Searching for information from various sources, but especially one that I will leave at the end of this post, I have come to the conclusion that it is the one that has finally worked for me and I want to leave my experience with the ASUS. Strix Z790 motherboard. F. Make all changes in the BIOS.

For undervolting. The ONLY setting I messed with is the ACTUAL VRM core voltage. It will be necessary to put it in COMPENSATION MODE. The COMPENSATION MODE SIGNAL must be negative (-). Then enter a value, in 0.08000. I leave the rest of the changes in the following graphs. I want to thank the user (weidemanTV), from the Asus ROG forum, for his detailed guide. https://rog-forum.asus.com/t5/overc...rvolting-13900k-with-strix-z790-e/td-p/941755

 

[bug]

Lowering your CPU voltage can be a great way to reduce heat and power consumption, especially if you have a CPU with as much power as the 13900k. Searching for information from various sources, but especially one that I will leave at the end of this post, I have come to the conclusion that it is the one that has finally worked for me and I want to leave my experience with the ASUS. Strix Z790 motherboard. F. Make all changes in the BIOS.

For undervolting. The ONLY setting I messed with is the ACTUAL VRM core voltage. It will be necessary to put it in COMPENSATION MODE. The COMPENSATION MODE SIGNAL must be negative (-). Then enter a value, in 0.08000. I leave the rest of the changes in the following graphs. I want to thank the user (weidemanTV), from the Asus ROG forum, for his detailed guide. https://rog-forum.asus.com/t5/overc...rvolting-13900k-with-strix-z790-e/td-p/941755

Just be careful and test for stability when you do that.
Base voltage is there for a reason. It's just that motherboard tend to feed more voltage to the CPU so that it will be stable at higher frequencies and make the motherboard look ahead of the curve in benchmarks. That overvoltage you can easily do without. Going lower than that is usually about luck.

 

[Block10]

Just be careful and test for stability when you do that.
Base voltage is there for a reason. It's just that motherboard tend to feed more voltage to the CPU so that it will be stable at higher frequencies and make the motherboard look ahead of the curve in benchmarks. That overvoltage you can easily do without. Going lower than that is usually about luck.

I understand what you are saying and I thank you for what you point out, here is the result of a stress test in Cinebench R23, for 10 minutes with the configuration that I was able to do.

 

[dgianstefani]

I understand what you are saying and I thank you for what you point out, here is the result of a stress test in Cinebench R23, for 10 minutes with the configuration that I was able to do.

The issue is not at stress test clocks.

You can crash at idle voltages with an undervolt.