Z590 with i9-11900K - Frequent freezes/blue screens - bug in Intel?

[unclewebb]

1 hour ago, -rascal- said:

VCCIO/VCCSA shouldn't be so high

I agree. It looks like 8th, 9th, 10th and 11th Gen are all similar. 

 

Quote

DO NOT leave these on auto

I always agree with that advice. 

 

His board on auto settings was setting SA voltage to 0.832 which I think is too low. This could be the root cause of the stability issues he was having. Enabling XMP automatically jacked the SA voltage up to 1.456 V which I think is too high and probably not necessary. Too high is a common issue with many motherboards. A stable computer results in less complaints. Long term CPU survival is probably not as important to motherboard manufacturers. 

 

Here is the mid point.

 

1.144 V = (1.456 V + 0.832 V) / 2

 

If I was going to set the SA voltage manually, I would probably start testing somewhere around here and adjust the SA voltage accordingly. I have DDR-4000 memory so for me, setting SA to 1.25 V works well. I might be stable at 1.20 V but I prefer 100% stable. I do not think a little bit of extra SA voltage is going to kill a chip anytime soon. 

 

Some users on OCN seemed to be running their 11th Gen with the VCCIO set to the Intel spec, 1.05 V. This one does not seem as important as SA. Thanks for posting the IMC info. 

[sh10151]

It's just crazy to me that doing BIOS reset seemingly results in bad TJ max setting at auto, and bad voltages. That just sucks.

 

ASRock does have their like "memory enhancement" stuff turned on by default - but when I was using four sticks of RAM turning it off resulted in my system no longer booting. Maybe it's doing something weird to voltages?

 

Sigh. So disappointing to spend this much money - and I still don't know who's to blame here - is there an issue with C-states? ASRock's BIOS settings/voltage handling? My G.Skill RAM? 

[-rascal-]

41 minutes ago, iransofaraway said:

It's just crazy to me that doing BIOS reset seemingly results in bad TJ max setting at auto, and bad voltages. That just sucks.

 

ASRock does have their like "memory enhancement" stuff turned on by default - but when I was using four sticks of RAM turning it off resulted in my system no longer booting. Maybe it's doing something weird to voltages?

 

Sigh. So disappointing to spend this much money - and I still don't know who's to blame here - is there an issue with C-states? ASRock's BIOS settings/voltage handling? My G.Skill RAM? 

 

I'm leaning towards AsRock and their default BIOS settings being a mess.

Kinda why I haven't touched an AsRock board since ..... 2013?

That said, not all AsRock boards are crap...like the Taichi boards on 3000-series / 5000-series Ryzen is pretty damn good.

More recently, with Z390 / Z490 / Z590 (and the lower tier boards from these generations), AsRock has been dropping the ball big time.

Like they can even hold onto the ball....

[sh10151]

I wish I had a way to repro rather than just "wait a few days and see if it crashes again". But I guess here we are. G.Skill support got back to me, wants me to try the other pair by itself and see if the issues persist, and then they can RMA. But I'm gonna leave them out of the PC for now. I have no way of knowing if they're the cause of my issues, or if it's a C-state thing, or voltage.. or what. I messaged Intel on Twitter, and maybe I'll just link them here to see if they have any insight. The C-state stuff seems to be happening to other i9 users. And sure, I can run my proc at high-perf mode and never let C-states change, but that's lame - and I would expect Intel to patch a fix, or replace the CPU if it's a hardware fault.

 

So here's where I'm at:

• Original G.Skill RAM pair installed in correct slots. 2nd pair removed.
• BIOS defaults loaded, except:
  • XMP profile 1
  • Spread Spectrum disabled (BIOS did this when I set XMP on)
  • ReBAR On
  • TJ Max set to 100C
  • Secure Boot on
• Windows set to high-perf mode

[Clausewitz]

3 hours ago, iransofaraway said:

It's just crazy to me that doing BIOS reset seemingly results in bad TJ max setting at auto, and bad voltages. That just sucks.

 

ASRock does have their like "memory enhancement" stuff turned on by default - but when I was using four sticks of RAM turning it off resulted in my system no longer booting. Maybe it's doing something weird to voltages?

 

Sigh. So disappointing to spend this much money - and I still don't know who's to blame here - is there an issue with C-states? ASRock's BIOS settings/voltage handling? My G.Skill RAM? 

Hello, I am the author of the thread referenced here on overclock.net: https://www.overclock.net/threads/a-possible-defect-in-11th-gen-cpus-i-just-got-done-speaking-to-intel.1792599/page-3#post-28855260

I wanted to state that this issue is not uncommon across the Z590 platform with the 11900K/KF CPUs.

https://www.sweclockers.com/forum/trad/1631356-ny-dator-intel-11900k-samt-asus-prime-z590-a-bsod-hela-tiden
https://linustechtips.com/topic/1330614-11900k-asus-z590-a-crashing/

After countless hours trouble shooting this on multiple CPUs and motherboards I can offer a glimpse into what I believe the problem is and perhaps a temporary fix for you for the time being. 

In general, when the C-states are disabled the CPU will not cause BSODs and will be stable. Several things are happening when the C-states are disabled. The core temperature no longer shows 0c as a "minimum temperature" and additionally the CPU will not BOOST beyond 5.1 if in ABT. 

The theory here is that something is wrong in the ucode(microcode) that is causing the C-states to behave improperly causing the temperature readings to hit 0c. I am unsure why this ALSO fixes the CPU crashing/BSODing on multiple platforms. I believe the CPUs are boosting to 5300 too much or on too many cores that are not strong enough to handle that frequency. Only two cores on an 11900k are rated to hit 5300. When ABT is enabled it allows the cpu to raise the frequency on the cores as it pleases. This could also be a fault of the motherboard manufacturers applying Multicore enhancement in conjunction with ABT. 

Intel specifies that in order for ABT to function properly that the power limits must be observed. 



I am not sure what the ASROCK equivalent of disabling MCE is but you will need to do it if you want to keep ABT enabled.

As it sits ABT will sometimes STILL cause issues. It was added to the platform very late after announcement and the original ucode did NOT have ABT nor did the original ucode observe the cores showing 0c at minimum temperatures.

First, we need to make sure your CPU is not faulty.
 

1. Load the defaults on your motherboard
2. Do not enable XMP
3. Do not enable ABT
4. DISABLE the C-states
5. Run benchmarks/play games and report back. 

[SkilledRebuilds]

My 2cents, I've read the thread.

Outside of already proposed solutions.

Curious if PCIE4 mode could be causing weirdness over setting and using PCIE 3.0. Cos PCs are weird too right .

 

I see random quirk posts come up time to time with this as a cause.. (z590) and PCIE4 GPUs and using PCIE3 can stop some encountered weirdness.

 

Its also easily changed and reversible.

Just 2cents worth of input.

[unclewebb]

3 hours ago, Clausewitz said:

when the C-states are disabled

Have you ever tried testing the individual C states to try to find out which ones are causing problems? The core C states, C1E, C3, C6 and C7, can all be toggled on and off individually. Same goes for the package C states. There might be some of them that are causing problems but it is hard to believe that Intel would release a CPU and all of the C states are causing BSODs or freeze problems.

 

It would be interesting to see what C states can be enabled without the CPU bugging out and reporting 0°C. This and the stability issues seem to be related.      

 

3 hours ago, Clausewitz said:

the CPU will not BOOST beyond 5.1

At default settings, the core C3 C state or higher has to be enabled so the CPU can use the maximum turbo multiplier so it makes sense that you do not get full turbo boost when the C states are disabled. 

 

I am curious if the CPU is stable because the C states are disabled or is the CPU stable because the maximum turbo multiplier has been reduced? Disabling the C states may improve stability but it might be doing this indirectly. Lowering the maximum turbo speed might be what is really improving stability.

 

I would love to do some testing but I like my 10850K too much to ever upgrade to 11th Gen.

4 hours ago, Clausewitz said:

Do not enable XMP

That is not practical advice for @iransofaraway

His motherboard drops the SA voltage way too much when XMP is disabled. 

 

[Guest]

I'd diag with both kits installed since this seems to be memory related, however the lack of input from the original poster makes it difficult to diag.

 

Intel fairs well on odd timings. I'd run the two kits total 64gb of memory with relaxed timings 17-17-17-37 or even as loose as 19-19-19-39 and test either with command rate 1 and 2T. Some other settings here and there.

 

But why would ever having a system with mixed kits give any issues and be of some shock or surprise? 

[sh10151]

@ShrimpBrime I’ve tried to be pretty responsive in here. Is there anything else I can provide or try here? I was working with @unclewebb’s tips. 
 

Regarding separate kits- I was under the understanding that as long as they were the same model it was fine. I’ve been building computers for like 25 years now and never heard of having to buy them all at the same exact time or buy only 4 pack or whatever.
 

I feel like you’re upset with me or something. I’m happy to learn and am just trying to figure this out. 

[Guest]

Just now, iransofaraway said:

@ShrimpBrime I’ve tried to be pretty responsive in here. Is there anything else I can provide or try here? I was working with @unclewebb’s tips. 
 

Regarding separate kits- I was under the understanding that as long as they were the same model it was fine. I’ve been building computers for like 25 years now and never heard of having to buy them all at the same exact time or buy only 4 pack or whatever.
 

I feel like you’re upset with me or something. I’m happy to learn and am just trying to figure this out. 

Nope, I try not to share feelings. I'm too old for that kind of engagement with people I don't know... that said, I'm never intentionally trying to offend, just lay what I see on the table and go from there.

 

Well proper info would include both kits serial numbers and build dates. They do make "revisions" of the same part number of many electronics, memory included.

 

So, in the 20+ years I've been building, this to me seems to be common knowledge, not to mix kits even if the same part number. But for certain anything is possible with electronics so I do always encourage trying anything and everything for a resolve.

 

However, starting small sometimes does the trick. That's at all defaults. If this issues is true both XMP and defaults with all other system settings default, you then have a mixed memory issue, the board and or memory controller doesn't like the memory.

 

Lastly, keeping in mind the memory controller with a higher voltage if set to auto is on purpose and for stability. Otherwise the firmware developers that write and test bios wouldn't have set it that high. In most cases not necessary in most users opinions. It's not like your going for records but in search of stability. True higher voltages bring stability, but also higher temps.

 

Lastly.... I've seen a lot of components lock up and blue screen rigs that had nothing to do with internal devices. In example, just 3 days ago an USBC cable went bad and the board no posted to usb power surge. The cord was hot at the end. In the garbage it went. So yeah. Start simple. Defaults for diag always. 

 

 

[Clausewitz]

38 minutes ago, unclewebb said:

Have you ever tried testing the individual C states to try to find out which ones are causing problems? The core C states, C1E, C3, C6 and C7, can all be toggled on and off individually. Same goes for the package C states. There might be some of them that are causing problems but it is hard to believe that Intel would release a CPU and all of the C states are causing BSODs or freeze problems.

 

It would be interesting to see what C states can be enabled without the CPU bugging out and reporting 0°C. This and the stability issues seem to be related.      

 

At default settings, the core C3 C state or higher has to be enabled so the CPU can use the maximum turbo multiplier so it makes sense that you do not get full turbo boost when the C states are disabled. 

 

I am curious if the CPU is stable because the C states are disabled or is the CPU stable because the maximum turbo multiplier has been reduced? Disabling the C states may improve stability but it might be doing this indirectly. Lowering the maximum turbo speed might be what is really improving stability.

 

I would love to do some testing but I like my 10850K too much to ever upgrade to 11th Gen.

That is not practical advice for @iransofaraway

His motherboard drops the SA voltage way too much when XMP is disabled. 

 

I disagree that it is impractical to run the memory at default settings. Saying he must use the Extreme Memory Profile in order to use his ram and that JEDEC standard settings  - which the sticks will operate at with XMP disabled - is unlikely. 

He doesn't HAVE to overclock the RAM (XMP is overclocking).

It's Residency C7 is causing the 0c issue. However, that doesn't fix the crashing issue. Experimenting with individual C states does not resolve the crashing observed. Only when they are fully disabled does the MACHINE CHECK EXCEPTION and CLOCK WATCHDOG TIMEOUT BSOD situation resolve. 

I have personally experimented with several CPUs now and several motherboards. I do not disagree that the issue is resolved because the boost clocks are limited. I believe that is why the C-states do remedy the ABT BSODing. However, it is also observed that any halt state or abrupt frequency changes cause the CPUs to crash and BSOD or freeze.

This is a major issue at the moment. DELL is having a widespread issue with their 11900k/kfs on the Alienware R12 units: 

Shamino on the ASUS ROG forums is offering up beta BIOSs to try to fix it: https://rog.asus.com/forum/showthread.php?123415-Maximus-13-and-Rocket-Lake-The-Rules-have-Changed#post831877

People are setting SVID behavior to INTEL FAIL SAFE in an effort to stop crashes: https://rog.asus.com/forum/showthread.php?123890-Maximus-13-Hero-new-Bios-0707/page2#post835613

https://www.sweclockers.com/forum/trad/1631356-ny-dator-intel-11900k-samt-asus-prime-z590-a-bsod-hela-tiden

Setting max boost to 5.2 is sometimes fixing it (your theory on the boosting too high is correct): 

This is impacting other motherboards, not just ASUS, not just ASROCK.

Users know ABT is crashing their system: 

My belief is that ABT is pushing cores to 5.3ghz when they cannot handle it. This is creating an even bigger issue when power limits are disabled (MCE equivalents) as the motherboards are implementing their own overclock power limits released and this is pushing the CPU to even further unsustainable and unstable clockspeeds.

Intel FAIL SAFE is pushing too much vcore into the CPU but keeping it stable on some boards with ABT enabled.

The issue is that the ucode is not mature enough, ABT is not mature enough, and sometimes different CPUs can handle the ABT settings as is - that are better binned - creating the feeling that some CPUs are just "bad" when in fact the bin cannot handle it. 

The one thing that is fixing the issue across the board is disabling C-states. There are two BSOD issues at the moment, something relating to frequency changing and the CPU being unstable - and the CPU boosting too high. 

If you guys keep focusing on his ram as the issue he won't find a solution. This is a very wide spread problem pertaining to the CPU. SP 50, SP 70+ SP 90, multiple bins. 

In fact, over on Overclocknet forums users are having a better time with a manual all core clock and greater stability. 
 

[Guest]

7 hours ago, Clausewitz said:

 

If you guys keep focusing on his ram as the issue he won't find a solution. This is a very wide spread problem pertaining to the CPU. SP 50, SP 70+ SP 90, multiple bins. 

In fact, over on Overclocknet forums users are having a better time with a manual all core clock and greater stability. 
 

I can agree 5.3ghz is pushing 14nm pretty hard. But I'm still not convinced (Yet), but you do have a strong argument...

However...

OP (here not other forum) pulled memory, was going to do some testing in that fashion but no word on results from that yet.

 

We've had a lot of good ideas here, but testing feed back is null, I feel we may not see the resolve unless there are some results.... from something.

 

But I can say the last time I saw that particular blue screen, it was from overclocking my memory. Which anyone can do a quick Google search on.

 

But ultimately, I'll kick back and watch. I have lots of respect of OCN and all the guys over there, I've been a member since 07.

[Clausewitz]

5 hours ago, ShrimpBrime said:

I can agree 5.3ghz is pushing 14nm pretty hard. But I'm still not convinced (Yet), but you do have a strong argument...

However...

OP (here not other forum) pulled memory, was going to do some testing in that fashion but no word on results from that yet.

 

We've had a lot of good ideas here, but testing feed back is null, I feel we may not see the resolve unless there are some results.... from something.

 

But I can say the last time I saw that particular blue screen, it was from overclocking my memory. Which anyone can do a quick Google search on.

 

But ultimately, I'll kick back and watch. I have lots of respect of OCN and all the guys over there, I've been a member since 07.

The OP really does need to respond. Even if everything is fine now - just let us know.

[sh10151]

The crashes haven’t happened again but like I said it’s random, I have no way of reproducing it that I’m aware of. I’ve sometimes gone days just fine then I’ll be typing or something and it freezes. 

[Clausewitz]

2 minutes ago, iransofaraway said:

The crashes haven’t happened again but like I said it’s random, I have no way of reproducing it that I’m aware of. I’ve sometimes gone days just fine then I’ll be typing or something and it freezes. 

Well, what are you trying right now? What settings are you using since you haven't had freezing? Is it the same?

[sh10151]

54 minutes ago, Clausewitz said:

Well, what are you trying right now? What settings are you using since you haven't had freezing? Is it the same?

I haven't had a chance to be on my computer today. But what I'm trying is the advice given above of going back to BIOS defaults - and turning off C-State entirely in BIOS. I am also running high perf power plan. I played a little Battlefield V last night without issue. But as I said before too it doesn't seem to happen when system is loaded, only randomly when I'm doing low-CPU/etc tasks - which leads me to believe it is something with the C-state stuff stated by others. So I also wound up putting my other memory kit back in last night just to work one thing at a time - if things are stable with the C-State disabled, then my memory probably isn't contributing to the problem. I ran multiple Windows Memory Diagnostics already and they always come back passing. I could try MemTest too, unless there's some better product now. G.Skill has also emailed me back and is willing to RMA - but wants me to try the pairs separate and try and reproduce the problem.

 

Or maybe this is two issues happening together - memory kit/voltage/etc funkiness, along with Intel's C-State stuff. It's all so fun to troubleshoot.

[sh10151]

Been running my PC so far with stock BIOS settings, C-states disabled, high perf power plan and both memory kits in place. No freezes so far. But as I said earlier they're random, so I'll keep the machine running and report back if it freezes again.

[sh10151]

Checking in again. It's been multiple days and no freezing. It really does seem like the C-State stuff is bugged on Z590 chipsets or something.  @Clausewitz is this looking like a chipset issue or CPU? Or both? I see folks on the other thread with different motherboard, but same issues. And then folks with 10 series CPUs, but Z590 chipsets having issues...

[Clausewitz]

7 hours ago, iransofaraway said:

Checking in again. It's been multiple days and no freezing. It really does seem like the C-State stuff is bugged on Z590 chipsets or something.  @Clausewitz is this looking like a chipset issue or CPU? Or both? I see folks on the other thread with different motherboard, but same issues. And then folks with 10 series CPUs, but Z590 chipsets having issues...

I believe the issue lies within the ucode/microcode. Earlier microcodes didn't have the 0c temperature issue or crashing with C-states enabled. Then the ABT microcode dropped and chaos ensued. Hopefully we get an update soon.

[sh10151]

Intel support basically is saying nothing's wrong with the CPU that they can tell, but are willing to replace it.

 

We'll see what happens. In the meantime I also was going to ask G.Skill if they'll let me swap my two kits for a single 2x32GB kit from them or something...

[Guest]

7 hours ago, iransofaraway said:

Intel support basically is saying nothing's wrong with the CPU that they can tell, but are willing to replace it.

 

We'll see what happens. In the meantime I also was going to ask G.Skill if they'll let me swap my two kits for a single 2x32GB kit from them or something...

There probably isn't anything wrong with the cpu.

 

It is nothing more than a standard proceedure to disable c-states and then overclock. And this brings stability. All the way back to 1st gen Intel core Eye processors. 

 

Still no blue screens? 

Then why RMA anything at all?

Run it as you intend with c-states disabled.

 

You won't be saving much power with it enabled.

Why?

Because you pull the most current (measured in amps, not vcore) when under load.

Gated (parked cores) save very little energy and also lowers performance slightly as well.

 

If memory passes all testing with c-states disabled, might as well just run it.

 

Imo of course.

 

[DarkSmith2]

On 8/18/2021 at 4:28 PM, Mister Woof said:

Personally, I have 2 old Seasonic 850w Focus Gold units that run fine on high end CPUs and GPUs without issue.

Only Seasonic PRIME models are affected.

[Mister Woof]

9 hours ago, DarkSmith2 said:

Only Seasonic PRIME models are affected.

From my reading, it's the Focus Plus units that were affected

 

 

[DarkSmith2]

2 hours ago, Mister Woof said:

From my reading, it's the Focus Plus units that were affected

From PSU Tier List Post:

On 10/23/2019 at 5:15 PM, LukeSavenije said:

[8] Seasonic PRIME based units experience shutdowns with RTX3080/3090 (and possibly RX6900 XT) GPUs, especially ones with unlocked power limit like FE and ASUS Strix. It is recommended, if going with such units, to overprovision wattage, 1kW for RTX3080 and 1.2kW for RTX3090, or power-limit the GPU, or performing a 'dirty' fix of disconnecting a pin of 12V V-sense wire from PSU-side connector of 24-pin motherboard cable (courtesy of Jonny Guru). Units based on post-2018 revisions of Seasonic Focus platform and majority of units by other OEMs are not affected.

So ye, im wrong only post 2018 Models of Focus platform isnt affected.

 

If you have A GX model you are fine, if you have FX, you are not.

[Mister Woof]

6 minutes ago, DarkSmith2 said:

From PSU Tier List Post:
 

Seems to corroborate the thread - post 2018 Focus Platforms aren't affected, but it doesn't say anything about pre-2018.

 

Quote

Old revisions (2017-2018) will shut off on some GPUs (970, 1080 ti, vega 64, Vega VII, 2080 ti). it'll will trigger the overcurrent protection, and will shut down immediately. This is likely due to the high transients the GPUs are known to put out

 

New revisions (2018-2019) have a bigger headroom, but aren't built for it and start to ripple at the loads where the PSU would shut down at with the older revision.

 

Someone mentioned the "seasonic thing" which is what is discussed in the thread as a reason for crashes. I said that it's unlikely because if it was, it would be just tripping shut offs, not crashes.

 

And I do have affected PSUs (Focus bought in 2017) and still haven't had the tripping experience even with high power draw items. But yes if it happened, would result in shutdowns, which would be the same type of experience with a PRIME unit.