Question DEGRADING Raptor lake CPUs

[Kocicak]

I noticed some reports about degrading i9 13900K and KF processors.

I experienced this problem myself, when I ran it at 6 GHz, light load (3 threads of Cinebench), at acceptable temperature and non extreme voltage. After only few minutes it crashed, and then it could not run even at stock setting without bumping the voltage a bit.

I was thinking about the cause for this and I believe the problem is, that people do not appreciate, how high these frequencies are and that the real comfortable frequency limit of these CPUs is probably at something like 5500 or 5600 MHz. These CPUs are made on a same process (possibly improved somehow) on which Alder lake CPUs were made. See the frequencies 12900KS runs at. The frequency improvement of the new process tweak may not be so high as some people presume.

Those 13900K CPUs are probably highly binned to be able to find those which contain some cores which can reliably run at 5800 MHz. Some of the 13900K probably have little/no OC reserve left and pushing them will cause them to degrade/break.

The conclusion for me is that the best you can do to your 13900K or 13900KF is to disable the 5800 MHz peak, which will allow you to offset the voltage lower, and then set all core maximal frequency to some comfortable level, I guess the maximum level could be 5600 MHz. With lowered voltage this frequency should be gentler to the processor than running it at original 5500 MHz at higher voltage. You can also run it at lower frequencies, allowing for even higher voltage drop, but then the CPU is slowly loosing its sense (unless you want some high efficiency CPU intended for heavy multithread loads).

Running it with some power consumption limit dependent on your cooling solution to keep the CPU at sensible temperature will help too for sure.

 

[Markfw]

True. But Intel guarantees nothing.

You may find this interesting:

His P-cores are all running at 6200 MHz but clock down to 6100 when the load increases. For lightly threaded loads like game threads, this is working. But he is also using a crazy water block. You, on the other hand, insist on pushing the cores to 100% above 5800 MHz. What you are doing is more torturous. I hope I do not need to point out that even a seasoned overclocker isn't doing what you are trying to do.

Also, he is not even specifying how many cores, but is gaming. If he were to use all 32 threads on something that stresses the cores, they would probably clock way down, even with this custom water. Gaming only uses 50-150 watts. I did not catch the max threads used if it was said, I am deaf, and its hard to read the captions as fast as he talks.

 

[igor_kavinski]

I did not catch the max threads used if it was said, I am deaf, and its hard to read the captions as fast as he talks.

Pausing the video at different moments running that game, it seems up to 15 threads are being used. Not sure if all of them are being used by the game engine. But even then, the total CPU utilization stays at or below 20%. With such low usage, the CPU isn't getting enough work to feel the heat. But the minimum fps do increase considerably so the higher frequency seems to be paying off and finishing critical tasks quickly enough to improve the game performance.

 

[Markfw]

Pausing the video at different moments running that game, it seems up to 15 threads are being used. Not sure if all of them are being used by the game engine. But even then, the total CPU utilization stays at or below 20%. With such low usage, the CPU isn't getting enough work to feel the heat. But the minimum fps do increase considerably so the higher frequency seems to be paying off and finishing critical tasks quickly enough to improve the game performance.

That is what I suspected. But unless you are gaming, even with water and a workload you can easily overpower the cooling at those speeds.

 

[igor_kavinski]

That is what I suspected. But unless you are gaming, even with water and a workload you can easily overpower the cooling at those speeds.

Yeah. Maybe if the CPU die were larger and the P-cores weren't packed so closely together, the heat dissipation could have been better. It just gets too hot too quickly with all cores at 100%. Something needs to suck the heat from those cores, like a very expensive peltier cooling solution. Too bad Intel didn't ask Cooler Master to release an LGA1700 version of their SUBZERO AIO. It could have helped here. But it would also have eaten 250W of its own to cool the CPU.

 

[Wolverine2349]

I believe that 5.5 GHz limit is there mainly to limit power consumption in certain load intensity. If they got the limit to e.g. 5.7, the CPU may draw too much power during gaming for example, which would be tested in reviews and Intel would look even worse than they look now (power draw wise).

Those two 5.8 cores are probably selected somehow, but many of the other cores are capable of the same frequency or even more.

Unfortunatelly nobody knows (except Intel), what which core is capable of and what it can be safely subjected to during testing.

I guess most of the 5.5 cores can run at 5.7 GHz without much trouble.

I sort of got all 8 cores stable at 5.7GHz with 1.38 VCORE. It passed Y Cruncher OCCT Variable Large Dataset, Prime95 Blend and Large non AVX and CInebench and AIDA64 System stability test.

Excerpt wait it was not truly stable as during CInebench it sometimes spit out a WHEA even though it did not crash.

This was of course with e-cores off.

I had to back down to 5.6GHz all core and could take VCORE down to 1.325V LLC6/8 and fully stable and much better temps too on air. Temps were borderline bad at 1.38 and throttled in CInebench hitting 100C at times even though somewhat stable. At current settings none of that.

 

[Hulk]

My bad, I didnt read it. I just assumed it was like my Broadwell Xeon E5 2696 v4 lol which I've been really studying turbo core behavior on --- but in hindsight the chips couldn't be much more different.

guess I should have done some homework first. So all core turbo is 5.5, right? So if he goes to 6GHz, thats 500MHz OC.

I would like to get a HWINFO64 read out for a 13900K to study just how much variation in clock speed

But back to the main point, I do not believe that he degraded the processor in that short a timeframe.

I agree.
AFA Intel's menagerie of overclocking features it's kind of confusing.

Base clocks are the clocks at defined power ratings.
TB2 is the all core boost at higher power ratings.

TB3 will boost one or two favored cores if there is thermal/power headroom an undisclosed frequency amount.

All Core Thermal Velocity Boost (TVM) will boost all cores if power and thermal headroom is available. Or from Intel "Opportunistically increases clock frequency by up to 100MHz; available if the CPU’s within temperature limits (70°C for desktops, 65°C for mobile) and turbo power budget is available."

Single core TVB will boost the faster of the two favored cores past TB3 if headroom is available.

Adaptive Boost will boost all cores further than TB2 and TVM as long as temps are under 100C.

So for all cores I think it goes like this.
Stock speeds>TB2 speeds>TVB (+100MHz if under 70C)>ABT (+? frequency) if under 100C.

According to Intel's graphs it looks like ABT can push all cores to favored cores TB2 frequency (5.5GHz) if power and headroom is available.

For single cores it looks like.
Stock speeds>TB2 speeds>TB3 increase>TVB

Not sure how you get TB3 without TVB on the 13900K because I have only witnessed either 5.5GHz or 5.8GHz meaning either you get both or neither. To get TVB to work you need to keep temps under 70C (desktop) or 65C (mobile), which is a seemingly not a high benchmark until you realize you have to nuclear fired hotspots in your CPU that need to be cooled. So it's not the amount of thermal energy that needs to be moved that is the problem but rather getting enough heat transfer from such a small area of the die.

Intel needs to simplify this so something like the following.

Base clocks and then Turbo clocks.
All Core Turbo meaning the CPU will at stock settings max out all cores at 5500GHz assuming power is available and thermals are in check.
Favored 2 Core Turbo meaning max these two favored cores will max out at.

So the 13900K for example would be rated as follows for P cores:
Base clock 3GHz
All Core Turbo - up to 5.5GHz
Favored 2 Core Turbo - up to 5.8GHz

If they don't want to list that at full tilt it would consume 300+ watts then they could just list
253W TDP with P's at 5.4 and E's at 4.3

All of the crazy boost modes are to obfuscate power draw and increase advertising materials.

I think they would do better with a little more honestly actually. There is nothing wrong with saying hey our CPU can do 5.2/4.1 frequency at 200W providing this much performance and not requiring extreme cooling but if you want to go for it then provided you have the power supply and cooling it'll boost up to 5.8/4.3.

 

[Storm-Chaser]

I agree.
AFA Intel's menagerie of overclocking features it's kind of confusing.

Base clocks are the clocks at defined power ratings.
TB2 is the all core boost at higher power ratings.

TB3 will boost one or two favored cores if there is thermal/power headroom an undisclosed frequency amount.

All Core Thermal Velocity Boost (TVM) will boost all cores if power and thermal headroom is available. Or from Intel "Opportunistically increases clock frequency by up to 100MHz; available if the CPU’s within temperature limits (70°C for desktops, 65°C for mobile) and turbo power budget is available."

Single core TVB will boost the faster of the two favored cores past TB3 if headroom is available.

Adaptive Boost will boost all cores further than TB2 and TVM as long as temps are under 100C.

So for all cores I think it goes like this.
Stock speeds>TB2 speeds>TVB (+100MHz if under 70C)>ABT (+? frequency) if under 100C.

According to Intel's graphs it looks like ABT can push all cores to favored cores TB2 frequency (5.5GHz) if power and headroom is available.

For single cores it looks like.
Stock speeds>TB2 speeds>TB3 increase>TVB

Not sure how you get TB3 without TVB on the 13900K because I have only witnessed either 5.5GHz or 5.8GHz meaning either you get both or neither. To get TVB to work you need to keep temps under 70C (desktop) or 65C (mobile), which is a seemingly not a high benchmark until you realize you have to nuclear fired hotspots in your CPU that need to be cooled. So it's not the amount of thermal energy that needs to be moved that is the problem but rather getting enough heat transfer from such a small area of the die.

Intel needs to simplify this so something like the following.

Base clocks and then Turbo clocks.
All Core Turbo meaning the CPU will at stock settings max out all cores at 5500GHz assuming power is available and thermals are in check.
Favored 2 Core Turbo meaning max these two favored cores will max out at.

So the 13900K for example would be rated as follows for P cores:
Base clock 3GHz
All Core Turbo - up to 5.5GHz
Favored 2 Core Turbo - up to 5.8GHz

If they don't want to list that at full tilt it would consume 300+ watts then they could just list
253W TDP with P's at 5.4 and E's at 4.3

All of the crazy boost modes are to obfuscate power draw and increase advertising materials.

I think they would do better with a little more honestly actually. There is nothing wrong with saying hey our CPU can do 5.2/4.1 frequency at 200W providing this much performance and not requiring extreme cooling but if you want to go for it then provided you have the power supply and cooling it'll boost up to 5.8/4.3.

Thanks for the detailed response. I will get back to you on this topic at some point.

 

[Hulk]

Thanks for the detailed response. I will get back to you on this topic at some point.

No worries. Just trying to make sense of it.

 

[Kocicak]

I just found out that the newest BIOS for my motherboard offers "Instant 6 GHz", I quickly flashed this bios and turned this profile on to see, what voltage Gigabyte believes is safe for running 1-2 best cores at 6 GHz, at load the voltage reached 1.43V.

Interesting...

 

[igor_kavinski]

at load the voltage reached 1.43V.

Rock solid stable or any issues?

Also, on reboot, is it always cores 2 and 3 that run at 6 GHz or does the BIOS select them randomly?

When you clicked "Instant 6 GHz" option, did it run some test for some time to find the best cores?

 

[Kocicak]

The best 2 cores are selected in manufacture and they are always the same.

I have no idea, if it was stable, I just turned it on for a while to see the voltage.

Now I am at 5700 MHz all P cores maximal frequency, back again on air cooler limited by 160W for change, auto voltage with no change in offset. It runs this high frequency at 1.35 V in load. 160W limit allows just 6 threads of Cinebench to run at 5700 MHz. All 32 threads run at 4500/3600 MHz and 62°C, score is 34K.

 

[Storm-Chaser]

Does the gigabyte 6.0GHz / 1.43v core thing change your mindset at all on the degradation "theory"?

 

[Storm-Chaser]

The bottom line is too much voltage doesn't kill/degrade CPU's. Too much current does.

BINGO

 

[Kocicak]

Voltage is a necessary prerequisite for current, I have no idea what is the debate about. With low enough voltage the CPU is not getting damaged with whatever load you throw on it.

the degradation "theory"

I have no idea what is your agenda, but degradation are physical processes happening in every semiconductor, it just matters how quickly in what conditions.

And also - not every chip is manufactured the same and perfectly.

Remember, how AVX workloads stress CPUs hard and now they are disabled in the new Intel consumer CPUs? Can you guess why? I believe that the CPUs simply would not last long running that workload. That is the simplest and most logical explanation. It all goes back to degradation.

The consumer Intel CPUs are made to run at such speeds and voltages, that they are UNABLE to run certain workloads without killing themselves.

I believe that in certain conditions something melts or fails in other ways in the chip, I am not sure this is technically a degradation or failure.

 

[Storm-Chaser]

Remember, how AVX workloads stress CPUs hard and now they are disabled in the new Intel consumer CPUs? Can you guess why? I believe that the CPUs simply would not last long running that workload. That is the simplest and most logical explanation. It all goes back to degradation.

I totally agree that AVX is no joke. That at wide open throttle and the CPU and generates a lot of heat. But provided you have sufficient cooling, your processor is not going to degrade. So to be clear, if you weren't seeing over 90*C there is really ZERO risk of degradation. But again, you are jumping to conclusions. You are making the assumption that it's disabled due to the damage it does to CPUs, right? And what is your evidence for this conclusion?

We know this because many of us have done 24-48 hour runs of Prime which is arguably just as harsh on the CPU. When I worked out west in Seattle for a IT company, we would run extended torture tests, especially if we were dealing in second hand equipment. We would then support that server out in the wild for many years to come. So I gained a pretty good working knowledge of just how durable some of these servers really are. Including the ones that ran extended torture tests. Hmmmmm...

And where are you getting this idea that a CPU would not last long under that type of work load? Provided you have cooling in place that is a non factor.

 

[coercitiv]

I have no idea what is your agenda, but degradation are physical processes happening in every semiconductor, it just matters how quickly in what conditions.

And also - not every chip is manufactured the same and perfectly.

Is that why you're on your third 13900K? Out of curiosity, what happened to the first two CPUs?

 

[Kocicak]

Out of curiosity, what happened to the first two CPUs?

Is this really your business?

I can assure you that one was fully functional and when I returned the other, which harmed itself with my assistance, the seller was made aware of its weakness.

I also briefly owned one 13600K and one 13700K, both returned fully functional.

Now I own one 12600K, one 13900K and one AMD 5700G. 13900K can still be returned.

After having played with all these CPUs, the one I would like to have is the 13700K with its unused 8 E cores swapped for more powerful integrated graphic card.

Too bad that the 5700G with its CNB ST score of 1500 looks too wimpy compared to 2200 of the new Raptors, and AMD does not make any CPU with usable integrated graphic card for AM5 yet. I really do not need any discrete graphic card, what is in 5700G is plenty for me.

BACK TO THE SERIOUS TOPIC OF DEGRADATION!

 

[Storm-Chaser]

Now I own one 12600K, one 13900K and one AMD 5700G. 13900K can still be returned.

I take it you think all of these processors are now degraded because you OC'd by 500MHz?

 

[igor_kavinski]

AMD does not make any CPU with usable integrated graphic card for AM5 yet.

What application/game are you running that is too hard on Zen 4's iGPU?

 

[Kocicak]

... And where are you getting this idea that a CPU would not last long under that type of work load?

Because that is the simplest explanation to why it is disabled.

It may work without any harm at voltage and frequency so much lower than the stock values, that Intel does not want this necessary huge drop to happen for various reasons, as not looking bad, etc.

 

[Kocicak]

What application/game are you running that is too hard on Zen 4's iGPU?

I have not tested it yet. According to some tests it is even weaker than integrated graphics in Alder and Raptor lakes. AMD themselves speak about that iGPU as an emergency solution for the case a discrete graphic card fails, or I have not seen any other statement which contains any other usage scenario. I believe that the "office machines" on AM5 will use the upcoming CPUs with larger graphic cards.

 

[therealmongo]

Guys, come on. Why waste energy with this malarkey? We know based on past post history that the op is not a reliable source, need peeps forget that this person binned several Ryzen CPUs because they didn't get a "good one" in their opinion because both CCDs could not clock to the same gigahz that they fabricated out of their imagination and now they are abusing Intel CPUs in the same manner, a scourge, hope to see peeps stop feeding this abusive consumer image of self importance

 

[therealmongo]

I take it you think all of these processors are now degraded because you OC'd by 500MHz?

mongoled here, op is not worth your precious time!

 

[Kocicak]

therealmongo: Do you have something factual to say to the topic, or did you come just for a personal attact?

 

[scineram]

Because that is the simplest explanation to why it is disabled.

It may work without any harm at voltage and frequency so much lower than the stock values, that Intel does not want this necessary huge drop to happen for various reasons, as not looking bad, etc.

I am sure it has nothing to do with having a bunch of cores that will crash your program when running those instructions.