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.

 

[Thunder 57]

I love how the OP posted no links. Just some other shmuck on a forum claiming the same thing.

Is Intel still disabling VT-d on "K" SKUs?

No. They stopped that awhile ago. Maybe around Skylake? That was some BS market segmentation. That said, overclocked anything is not for professional use. I am a bit surprised Intel allowed Xeons to be overclocked. Same with Threadripper Pro. I'm all for more user control though. No one serious would overclock them though IMHO.

Bold words coming from someone who ordered multiple Raptor Lake SKUs, overclocked all of them and then decided to keep only one. It almost looks like other honest consumers are paying for your overclocking pleasure.

No almost about it.

 

[A///]

usually never want to buy a new cpu platform the first few months because there's creaks in production and those chips go boom sooner or later it seems. Gpus are the same with the initial runs going to reviewers.

 

[DrMrLordX]

Yes, and the opposite is also true. You can safely run the CPU at higher voltage/clock speeds in lower current loads. In modern CPU's the voltage and clock speed varies right along with the current demand of the workload. The bottom line is too much voltage doesn't kill/degrade CPU's. Too much current does.

The key issue here is user control. He didn't set current by hand, but he did tweak voltage. Current varies according to workload. The only way to avoid high-current scenarios is to simply avoid those workloads, and that's not why we buy these things, is it?

 

[Wolverine2349]

The key issue here is user control. He didn't set current by hand, but he did tweak voltage. Current varies according to workload. The only way to avoid high-current scenarios is to simply avoid those workloads, and that's not why we buy these things, is it?

I would think current also varies by clock speed as well. The higher clock speed is set static and fixed at more current at same workload right?

 

[lopri]

I monitored voltage and temperature in HWinfo. The values were acceptable, load was light, I left the room confident that it is just a light 10 minutes test which will probably pass without any problem, and minutes later I found the computer restarted and the CPU unable to run any benchmark with default settings without positive voltage offset (20mV was enough). The CPU must have some voltage reserve for default use, so it degraded by that reserve + 20mV.

The weakness in the CPU may be of the kind it will allow short term 100% functionality, but it will manifest itself fully only after slight stress. I do not believe that CPU could have been long term reliable, because the process at lower voltage would be happening too, just slower. I already wrote that what failed at 1.45V and 6000MHz in minutes would have failed anyway at 1.3V and 5500 MHz in tens or hundreads of hours (or even earlier).

One member of overclock.net describes quick degradation too.

This sounds eerily similar to my 3700X with a degraded memory controller. Crashes and BSOD followed.

 

[alcoholbob]

I had this happen over time with my 12700K. Was able to run a -100mV undervolt for a good 3 months at launch before it started random reboots. Then it I determined the stable undervolt was now at -85mV. This was stable for another 5-6 months. A year later, anything over -60mV just causes spontaneous reboots within a few minutes of POST.

Whereas my 9900K is still running strong 4 years later, never had to let off my undervolt at all.

 

[coercitiv]

The only way to avoid high-current scenarios is to simply avoid those workloads, and that's not why we buy these things, is it?

There are current limits configurable via UEFI, though as you point out lowering ICC max kinda defeats the purpose of overclocking.

I think the better way to go about this is to think in terms of a trifecta: voltage, current, temps. You can have 2 of them with high values, but the third must always be relatively low:

• high current, high temps, low voltage - typical MT workload with proper power limit
• high voltage, high temps, low current - typical ST or lightly threaded workload
• high voltage, high current, low temps - typical overclock with very good cooling

 

[Hulk]

Voltage is just a patsy in this crime. The current is the real murderer.

More soldiers (current) more damage. More heavily armed soldiers (higher voltage), even more damage.

 

[Markfw]

More soldiers (current) more damage. More heavily armed soldiers (higher voltage), even more damage.

Well, you have 2 choices it looks like. Underclock it a lot, or buy a Zen 4. Intel factory overclocked the most recent generation, just to beat AMD. Well they do by a hare, until they meltdown.

Edit: Just to be clear. I have seen several threads with people trying to cool Raptor lakes, now this thread. I have NOT seen any threads about problems with Zen 4 cooling or degradation, so this is just an observation, nothing more.

 

[DrMrLordX]

I would think current also varies by clock speed as well. The higher clock speed is set static and fixed at more current at same workload right?

It can, actually, but that's going to vary by platform. And when I tested it on my present machine, it didn't vary by as much as I thought it would. Yes current can track upwards with voltage and clockspeed on the same workloads, but not by much.

There are current limits configurable via UEFI, though as you point out lowering ICC max kinda defeats the purpose of overclocking.

It does, and throttling current draw could have some really strange side-effects.

 

[Hulk]

Well, you have 2 choices it looks like. Underclock it a lot, or buy a Zen 4. Intel factory overclocked the most recent generation, just to beat AMD. Well they do by a hare, until they meltdown.

Edit: Just to be clear. I have seen several threads with people trying to cool Raptor lakes, now this thread. I have NOT seen any threads about problems with Zen 4 cooling or degradation, so this is just an observation, nothing more.

As I've written before I'm quite happy with my 13900K undervolted and limited to 190W. It's fun to tinker.

That being said since my next upgrade will require a new motherboard regardless if I go with Intel or AMD I'm going to be open to either one. We'll see how they perform.

 

[Markfw]

As I've written before I'm quite happy with my 13900K undervolted and limited to 190W. It's fun to tinker.

That being said since my next upgrade will require a new motherboard regardless if I go with Intel or AMD I'm going to be open to either one. We'll see how they perform.

I am set to 142 watt. Not sure whats downclocked or not, but I don't like wasting power. I am glad someone else sees this. I have no idea how we would compare in performance at these levels, but I would guess closer than stock. But I still use less. At least you see my point.

Edit: And the reason you are not having problems is most likely due to your downclocking. I am using 33% less power though. (on my 7950x)

 

[TheELF]

I am set to 142 watt. Not sure whats downclocked or not, but I don't like wasting power. I am glad someone else sees this. I have no idea how we would compare in performance at these levels, but I would guess closer than stock. But I still use less. At least you see my point.

Edit: And the reason you are not having problems is most likely due to your downclocking. I am using 33% less power though. (on my 7950x)

If this is just an TDP limit and not an PPT limit then you are probably using just as much power as Hulk is, your CPU is just not reporting it and you have to measure it on the 12V rails.

Explaining AMD Ryzen Precision Boost Overdrive (PBO), AutoOC, & Benchmarks | GamersNexus

stub With the launch of the Ryzen 3000 series processors, we’ve noticed a distinct confusion among readers and viewers when it comes to the phrases “Precision Boost 2,” “XFR,” “Precision Boost Overdrive,” which is different from Precision Boost, and “AutoOC.” There is also a lot of confusion...

www.gamersnexus.net

Package Power Tracking (“PPT”): The PPT threshold is the allowed socket power consumption permitted across the voltage rails supplying the socket. Applications with high thread counts, and/or “heavy” threads, can encounter PPT limits that can be alleviated with a raised PPT limit.

1. Default for Socket AM4 is at least 142W on motherboards rated for 105W TDP processors.
2. Default for Socket AM4 is at least 88W on motherboards rated for 65W TDP processors.

 

[DrMrLordX]

If this is just an TDP limit and not an PPT limit

You can't really limit TDP using conventional tools on AMD platforms. You can limit PPT, or you can try limiting EDC or TDC if you're feeling in a weird sort of mood.

 

[Kocicak]

I had this happen over time with my 12700K. Was able to run a -100mV undervolt for a good 3 months at launch before it started random reboots. Then it I determined the stable undervolt was now at -85mV. This was stable for another 5-6 months. A year later, anything over -60mV just causes spontaneous reboots within a few minutes of POST.

Well, this would suggest that in your conditions (voltage, frequency, temperature and use intensity) the CPU degrades by about 50mV per year. This seems like a lot. Even worse when you realise that you are activelly trying to run it with as low voltage as possible! What would be the rate of degradation if you did not undervolt it?

Perhaps Intel 10nm process is not that great even after they "fixed it".

It is surely usable under some conditions, but pushing CPUs made on this process too hard does not seem to be a good idea.

If somebody had a lab and could measure a rate of degradation of these CPUs dependent on how hard they are run (voltage, frequency, temperature), that would be very good. It would be quite expensive project though, you need more than one CPU being subjected to the same conditions to avoid sample variation. 3 CPUs run at 3 different scenarios, that is just 9 systems? Seems doable. It would not be too time consuming, because the rate seems to be pretty high. (And 5 CPUs in 4 different scenarios makes just 20 systems, not that extreme either )

They would get a lot of views on youtube, for sure.

This could determine the safe conditions for these CPUs, so that consumers wishing to overclock them knew what they can safely afford to do and also other consumers wanting to assure long life of their CPUs would know, how to limit the voltage, etc.

 

[Markfw]

If this is just an TDP limit and not an PPT limit then you are probably using just as much power as Hulk is, your CPU is just not reporting it and you have to measure it on the 12V rails.

Explaining AMD Ryzen Precision Boost Overdrive (PBO), AutoOC, & Benchmarks | GamersNexus

stub With the launch of the Ryzen 3000 series processors, we’ve noticed a distinct confusion among readers and viewers when it comes to the phrases “Precision Boost 2,” “XFR,” “Precision Boost Overdrive,” which is different from Precision Boost, and “AutoOC.” There is also a lot of confusion...

www.gamersnexus.net

According to Ryzen master PPT is 142 watt. Its is expressly shown, as well, as cpu power and soc power.

 

[Carfax83]

Edit: Just to be clear. I have seen several threads with people trying to cool Raptor lakes, now this thread. I have NOT seen any threads about problems with Zen 4 cooling or degradation, so this is just an observation, nothing more.

You can easily configure Raptor Lake to run cool. Mine maxes out at 77c at 5.2ghz on air cooling. At stock settings though it is toasty to be sure.

Also, I'm pretty sure what the OP is experiencing is not degradation. As someone suggested, resetting CMOS likely could have fixed it.

I did a search on Google for "Raptor Lake CPU degradation" and only this thread and a single post on Overclockers concerning what a poster thought was IMC degradation, which turned out not to be the case.

 

[Markfw]

You can easily configure Raptor Lake to run cool. Mine maxes out at 77c at 5.2ghz on air cooling. At stock settings though it is toasty to be sure.

Also, I'm pretty sure what the OP is experiencing is not degradation. As someone suggested, resetting CMOS likely could have fixed it.

I did a search on Google for "Raptor Lake CPU degradation" and only this thread and a single post on Overclockers concerning what a poster thought was IMC degradation, which turned out not to be the case.

The Hulk replied with what I already know, yes, you can configure them down. But at stock, many users have had a problem, this is what I was pointing out. And Hulk's 180 watt setting is probably what I would choose.

 

[MadRat]

Does the CMOS stabilize the CPU operations? I could see where it would slow down to improve stability and get stuck in that mode.

 

[Wolverine2349]

I have my Core i9 13900K set to manual and static all P cores 5.6GHz and VCORE at 1.325V LLC 6/8 on an MSI Z690 Unify X. All e-cores are disabled. Hyper threading is left enabled. Frequency is static and fixed. No power limit set

Under load using CInebench R23, POUT value peaks at 189 watts and VOUT is like 1.257volts under load with a little VDROOP. Idle voltage is like 1.31 per VOUT value in HWInfo64.

Running CPU-Z Stress test, temps peak at 86C and VOUT value like 1.26 under full load and POUT value at 160 to 170 watts.

Am I at serious risk of degrading my 13900K at these settings.

The consensus Over at overclock.net seems to be Intel spec 253 watt power limit or lower is safe. And I am well below it and never even hit that limit except a brief moment (I think it may have been like 265 watts) trying Prime95 with AVX on when it got to Small FFT of blend for like 30 seconds to a minute before BSOD which was the only test it failed and hit above 100C and had to throttle (But Prime95 is so brutal especially on Intgel CPUs even compared to AMD even especially when I could pass all other tests like Y Cruncher 2 iterations even rough SFT remaining at peak 230 watts POUT value in addition to OCCT Large Data set, Linpack XTREME 1.1.5, and AIDA64 System stability test) Though temps above 85C always could have slight degradation long term if used tons and all the time and tons which I do not.

But these reports like this thread are scary none the less. But is it an overreaction and the reports like here are people who never were fully stable anyways and on the edge right off the bat.

Or does no one really know.

I am cooling it with a Noctua NH-D15S with 2 1080 RPM 140mm fans. Of course I sped them to max 1500 RPM during the hardest of stress tetsing like Y Cruncher SFT and Linpack XTREME and even opened my window in cold weather just to make sure it could pass all tests which it did without WHEAs or BSODs. And even SFT Y Cruncher peak temp of 95C with a cold room. Though Prime95 small FFT part of blend AVX on hit 101C and eventually BSOD as only test that failed.

Though once again in room with warm house 21C (67F which my heat is set at for winter), Cinebench R23 temp only peaks at 94C rarely and averages 90-92C with fan at 1080 RPM staying there when I run it.

Gaming it peaks at like 77C with POUT at like 175 watts though averages more like 130 to 140 watts mostly during a gaming load. And gaming is what I will mostly be doing with my upcoming RTX 4090 when I finally find one with so little/hush/faint or absolutely 0 coil whine.

 

[Markfw]

I have my Core i9 13900K set to manual and static all P cores 5.6GHz and VCORE at 1.325V LLC 6/8 on an MSI Z690 Unify X. All e-cores are disabled. Hyper threading is left enabled. Frequency is static and fixed. No power limit set

Under load using CInebench R23, POUT value peaks at 189 watts and VOUT is like 1.257volts under load with a little VDROOP. Idle voltage is like 1.31 per VOUT value in HWInfo64.

Running CPU-Z Stress test, temps peak at 86C and VOUT value like 1.26 under full load and POUT value at 160 to 170 watts.

Am I at serious risk of degrading my 13900K at these settings.

The consensus Over at overclock.net seems to be Intel spec 253 watt power limit or lower is safe. And I am well below it and never even hit that limit except a brief moment (I think it may have been like 265 watts) trying Prime95 with AVX on when it got to Small FFT of blend for like 30 seconds to a minute before BSOD which was the only test it failed and hit above 100C and had to throttle (But Prime95 is so brutal especially on Intgel CPUs even compared to AMD even especially when I could pass all other tests like Y Cruncher 2 iterations even rough SFT remaining at peak 230 watts POUT value in addition to OCCT Large Data set, Linpack XTREME 1.1.5, and AIDA64 System stability test) Though temps above 85C always could have slight degradation long term if used tons and all the time and tons which I do not.

But these reports like this thread are scary none the less. But is it an overreaction and the reports like here are people who never were fully stable anyways and on the edge right off the bat.

Or does no one really know.

I am cooling it with a Noctua NH-D15S with 2 1080 RPM 140mm fans. Of course I sped them to max 1500 RPM during the hardest of stress tetsing like Y Cruncher SFT and Linpack XTREME and even opened my window in cold weather just to make sure it could pass all tests which it did without WHEAs or BSODs. And even SFT Y Cruncher peak temp of 95C with a cold room. Though Prime95 small FFT part of blend AVX on hit 101C and eventually BSOD as only test that failed.

Though once again in room with warm house 21C (67F which my heat is set at for winter), Cinebench R23 temp only peaks at 94C rarely and averages 90-92C with fan at 1080 RPM staying there when I run it.

Gaming it peaks at like 77C with POUT at like 175 watts though averages more like 130 to 140 watts mostly during a gaming load. And gaming is what I will mostly be doing with my upcoming RTX 4090 when I finally find one with so little/hush/faint or absolutely 0 coil whine.

"disclaimer" I am not an expert at Raptor lake, but from that I read, those settings look good to me. Something close to this is what Intel should have set the chip at, and there would be way less problems in this area.

 

[Hulk]

We had quite a few discussions during the Core2Duo days about this when we were overclocking the crap out of those processors. If I remember correctly the only people that had problems were pushing 1.5V or more into them. Of course the nodes were a lot bigger then and could presumably handle the additional voltage. But I think the heat is what kills and heat is heat big process or small it's a problem.

 

[Wolverine2349]

We had quite a few discussions during the Core2Duo days about this when we were overclocking the crap out of those processors. If I remember correctly the only people that had problems were pushing 1.5V or more into them. Of course the nodes were a lot bigger then and could presumably handle the additional voltage. But I think the heat is what kills and heat is heat big process or small it's a problem.

Yes good point. Though it seems processors running hot has gotten to become acceptable. I remember when it was frowned upon to have tmeps in the low to mid 90s full all core load. Now it is considered well its ok because spec if up to 100C for these 10nm Intel CPUs.

Even AMD CPUs spec if up to 95C and they do that out of the box.

 

[Markfw]

We had quite a few discussions during the Core2Duo days about this when we were overclocking the crap out of those processors. If I remember correctly the only people that had problems were pushing 1.5V or more into them. Of course the nodes were a lot bigger then and could presumably handle the additional voltage. But I think the heat is what kills and heat is heat big process or small it's a problem.

I was one of those overclocking the crap out of core2duo. It converted me to Intel from AMD. I did this for years. I was such an avid overclocker, that what got me elected as a mod here. Until 2017... Anyway, yes, I never killed a cpu, and I put good heatsinks on them. I remember those days. (~2006)

 

[Hulk]

I was one of those overclocking the crap out of core2duo. It converted me to Intel from AMD. I did this for years. I was such an avid overclocker, that what got me elected as a mod here. Until 2017... Anyway, yes, I never killed a cpu, and I put good heatsinks on them. I remember those days. (~2006)

My first C2D was the E6400. Cheap and quite fast for the day.

Yup, Anand had that great preview/review for C2D. I think it was titled "The Empire Strikes Back" or something like that as AMD had been really pushing the P4 beyond it's limits with the Athlon.

Anyway electromitigation was the effect I think we were concerned about. Basically the charge carriers would eventually change the structure of the CPU requiring more volts to maintain a frequency.