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

[sh10151]

Hello,

 

I dropped a huge chunk of money on a top-of-the-line system a few months ago:

• SEASONIC PRIME 850W 80+T FM ATX

• NOCTUA NH-D15 CPU COOLER

• WD 1TB BLACK SN850 NVME SSD

• INTEL I9-11900K BOX

• 2x G.SKILL 32GB 2X16 D4 3600 CL16 RJ ( F4-3600C16D-32GVKC )

• ASROCK Z590 TAICHI

• EVGA RTX 3080 FTW3 ULTRA

And I've been having random freezes and weirdness, and I can't track it down. It seems to happen when my system is loaded (like playing a game for a while), but it's also random when it occurs - for example the other day I worked an entire day just fine, but then typing into a form on a website suddenly everything froze (system not under load).

 

I'm using V2.2 of my BIOS. I'm using stock/default BIOS settings, except:

• XMP memory profile

• Resizeable BAR Support enabled (and I flashed my GPU to support it)

• Intel Adaptive Boost Technology enabled

I'm using a large older Corsair case with plenty of airflow. Even have an intake fan blowing on the GPU.

 

I've run Windows memory diagnostic and it says things are fine.

 

I only just recently realized that the G.SKILL RAM I bought isn't on the QVL - but it's passing memory tests - so shouldn't that be fine?

 

I've run out of ideas to try and troubleshoot here. I turned off Resizable BAR thinking maybe it was unstable - and yet again today had another random freeze. I also re-seated by cooler in case it wasn't making contact all the way or something.

I guess next I could run memory at Auto and not XMP and try that?

 

What’s also weird is that if I turn off ASRock’s memory enhancement stuff (which is on by default) my system won’t POST.

 

I also wiped Windows a few days ago, and stuck with trusted drivers from Intel and nVidia only. Nothing else. And yesterday had a blue screen while just sitting there. Running Cinebench would freeze within 10 seconds or so, so I turned off the Intel Adaptive Boost Tech in case the CPU is overheating (I wouldn't think so with my massive cooler?) - and no freezes again so far (and Cinebench runs fine).

 

But this just seems bonkers. Shouldn't this system be stable? I'm not overclocking or anything weird. I'm running out of things to try. If it freezes again with the ABT off - I guess I can turn off resizeable BAR, and then if it still has issues - turn off XMP? But what if it keeps freezing after that?

 

Found someone else on this forum with similar issues but ASUS - could it be my CPU is bad? My motherboard? Maybe the 11 platform just sucks, period? Sigh. 

 

[Skiiwee29]

Judging by your system components, my fear is the PSU. Seasonic units struggle with the Transient responses that a 3080 and 3090 GPU place on them and cause what you're experiencing including random shut downs. Other people on the forum are significantly more smarter than I am with this stuff so ill let them explain it more and either debunk it or confirm.

 

@LukeSavenije @IIIIIIIIII

[LukeSavenije]

try slightly undervolting the GPU and leave it to run a stress test

 

some Seasonic PSUs have indeed shown issues with 3080, but these feel more like you pulling the power cord than a freeze of the system

[Jurrunio]

You're using 2 kits together, try run with only 1 kit.

[Chris Pratt]

1 hour ago, iransofaraway said:

I only just recently realized that the G.SKILL RAM I bought isn't on the QVL - but it's passing memory tests - so shouldn't that be fine?

QVL doesn't really matter.

 

I always like to confirm in situations like this just in case: what slots are you using for the RAM? It should be 2 and 4 in most cases, counting from the CPU side.

[Mister Woof]

I don't think it's PSU-if this is the "Seasonic thing" then it would trigger shutdowns, not freezes or bluescreens.

 

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

 

I have been reading a lot of issues with 11th gen chips, so I'm wondering if there's just not some hardware level faults on them. The other big thread seemed to focus on the ASUS Z590s, so who knows.

 

 

[freeagent]

Sounds like a memory issue.

[Mister Woof]

9 minutes ago, freeagent said:

Sounds like a memory issue.

Yeah, this is probably it.

 

If turning XMP off resolves your issue, then try turning it back on, then bumping up VCCSA/VCCIO (or System Agent/IO Voltage) a tad bit and also DRAM voltage a little to see if it helps stabilize it. If that doesn't work, you may need to drop to 3466 or 3200.

 

This is going to be more than likely necessary if you're running 4x16 from two different kits.

[freeagent]

Did you clear the cmos at all when going from two to four sticks? Clear it again when going back to two. Internal sub timings change a lot between configurations.

[sh10151]

I didn't even realize ya'll had replied! Just checked back and wow. Thanks so much for everyone taking the time to chime in here. I just had another freeze this morning, after the system running fine for a day or more. Was just typing in Chrome and it all freezes. No blue screen or anything. System wasn't under load. But last night played Battlefield V and Cyberpunk without issues. So bizarre.

 

Regarding the memory, yes, you're totally right - I didn't buy the sticks at same time. One pair bought at Microcenter, another bought at Amazon.com a month or so later.

 

I'm gonna try running with the original pair for now. Maybe try and return the second pair, as it's not absolutely necessary.

[unclewebb]

@iransofaraway What Windows power plan are you using? To try to troubleshoot this problem, use the Windows High performance power plan and disable all of the C states in the BIOS. Try this for a day or two and see if you have any issues.

 

Light load crashes are usually caused by the CPU dropping its voltage too low, which can happen when entering or exiting the various C states. Most games do not keep all 8 cores active so these crashes can happen at anytime but are more likely to happen when the CPU is lightly loaded. 

 

After disabling the C states in the BIOS, I would use ThrottleStop to double check what C states are enabled or disabled when idle. 

https://www.techpowerup.com/download/techpowerup-throttlestop/

 

 

 

If this works OK and you want to reduce light load power consumption, try enabling the C states one at a time and do some more testing. Go into the BIOS and enable C3 and then C3 and C6 and then C3, C6 and C7. On older CPUs, it was the package C states that caused the most problems. I would leave those ones disabled.

 

Here is with core C7 working but with the package C states disabled. 

 

 

Freezes seem to be a common issue with Z590 and 11th Gen CPUs. Not yet sure what the problem really is. Hopefully you can do some troubleshooting and figure out if this problem is C state related. 

 

The other 11th Gen problem is the CPU sometimes reporting 0°C. Does HWiNFO show that for the minimum temperature for any of your cores? I think these two problems are related. 

[sh10151]

@unclewebb it's funny you mention that - I'm seeing 0°C on my Core 0:

 

 

Should I contact Intel?

[sh10151]

Just found https://www.overclock.net/threads/a-possible-defect-in-11th-gen-cpus-i-just-got-done-speaking-to-intel.1792599/ too.

[sh10151]

I'm running on two sticks of the original RAM again, in the proper slots. Gonna leave power plan at balanced. I have XMP on, ReBAR on, and Intel Adaptive Boost Technology on. We'll see if this freezes again. I wish I had a way to repro.

 

If it freezes again, I think I'll try disabling ABT first (I think my air cooler is enough for it, but maybe not?) - and then if that still has issues, go to Power Plan High Perf.

 

That sound reasonable?

[sh10151]

I went ahead and turned off ABT. It seems like I can get my machine to reliably crash/blue screen if I have it on and run the multi-core Cinemark benchmark. It gets about 45 seconds in, then freezes and blue screens. When my system has frozen other times (as I described in the original messages) - I never get a blue-screen. Just complete freeze (not even mouse moves). So I think the ABT stuff is just my air cooler not handling it. So now I just have XMP and ReBAR on.

[unclewebb]

34 minutes ago, iransofaraway said:

Gonna leave power plan at balanced.

You can run an Intel CPU at full speed and get fantastic power consumption when it is idle or lightly loaded. The Balanced power plan might be useful if you have a laptop that is running on battery power. For most other situations, there is no point. Here is an example of 10 cores all running at full speed, 5000 MHz. 

 

 

Have you checked yet to see what C states your CPU is using when idle?

 

 

 

 

[sh10151]

44 minutes ago, unclewebb said:

Have you checked yet to see what C states your CPU is using when idle?

@unclewebb

I'm not totally familiar with this program, but how should I best monitor that?

 

It looks like C0% on left is showing 1.5 - and then on right I do not see C3 or C6 at all, only C7. I appreciate your help!

 

I was waiting to try changing power plan to only tweak one variable at a time. The processor should be able to run on balanced without freezing. I think if I keep having issues then disabling C-states/running high perf is a nice workaround, but unacceptable for a pricey new processor (and hopefully Intel can fix it?).

 

I'll switch to High Perf for now.

[unclewebb]

1 hour ago, iransofaraway said:

C0% on left is showing 1.5

When a CPU is actively working on something, it runs in the C0 state. When a CPU has nothing to do and the C states are enabled, it will go into one of the higher numbered C states to save power. 

 

1 hour ago, iransofaraway said:

I do not see C3 or C6 at all, only C7

If the core C7 state is available, there is no reason for the CPU to use core C3 or core C6. If a CPU core has nothing to do, it makes sense for it to go into the highest numbered C state possible so it can save the most amount of power. Your CPU is working correctly.

 

Try switching to the Windows High Performance power plan. I already posted a picture that shows it is OK to run an Intel CPU at full speed. When the C states are enabled. A fast CPU or a slow CPU when idle makes virtually no difference. Run it fast at a consistent speed and see if you have any issues. 

 

If you are still having problems, we can go from there. 

 

I would recommend going into the BIOS and not setting the maximum temperature to 115°C. The default is 100°C. Going beyond this is not great for the long term health of your CPU. 

 

1 hour ago, iransofaraway said:

a nice workaround, but unacceptable

You have to try and troubleshoot the problem first. Lots of 11th Gen owners are not having these problems. It could be your BIOS voltage settings or any number of things. It could be a compatibility problem with your memory. The SA or IO voltages might be set too low. It always takes a while for manufacturers of anything new to get the bugs out. 

 

Increasing the SA and IO voltages are what helped me get my 10850K stable. Run HWiNFO and see what it reports for those two voltages. 

[sh10151]

@unclewebb thanks for the explanation of C-states! I've switched to high perf plan for now. This is a little confusing: "I would recommend going into the BIOS and not setting the maximum temperature to 115°C." - not setting it? It's probably set at whatever the BIOS default is. I'm running entirely BIOS defaults except:

• ReBAR on
• XMP Profile 1
• All fans set to "Full" speed
• Secure boot on

As for SA and IO voltages, I'm seeing 0.825 V average for SA, and 1.056 V average for IO.

[unclewebb]

21 minutes ago, iransofaraway said:

BIOS default

The BIOS default should not be setting the thermal throttling temperature to 115°C. The Intel spec for this setting is 100°C. Look in your BIOS to see if you can find this setting and set it correctly. ThrottleStop should show on the main screen, PROCHOT 100°C not PROCHOT 115°C.

 

It looks like your System Agent voltage is set too low. I think the default IO voltage is 1.05 V. Here is what I use on my Z490 board.

 

 

A little bit of extra voltage for the RAM and lots of voltage for the VCCIO and the VCCSA has made my Z490 board rock solid stable. 

 

One test that I remember doing is fully loading my CPU with Prime95 Small FFTs but only for a minute or less and then stopping it. At default settings, I would always end up with a frozen screen. Not a BSOD, just a frozen screen. I believe it was the extra voltage that completely fixed this. 

 

I am not familiar with 11th Gen or the Z590 chipset. Try looking around overclock.net to see how much IO and SA voltage they are using. I do not think you will find anyone successfully running their SA voltage as low as you are. 

 

Can you post a FIVR screenshot?

 

Edit - Here is a guy on OCN running his 11900K at 5277 MHz. He is not afraid to use some extra SA voltage.

 

 

Another one at 5300 MHz. I think you can see what the pros are running. 

 

 

 

[sh10151]

@unclewebb 

4 minutes ago, unclewebb said:

The BIOS default should not be setting the thermal throttling temperature to 115°C

It is, at least on my board, defaulting to 115. Cleared BIOS and confirmed, its set to "Auto" inside BIOS, which reports 115. I manually set it to 100, per your guidance.

 

What would cause the voltages to be wrong, since I'm running at stock? Is this some ASRock issue or something else?

 

Hopefully this is the screenshot that's helpful to you. Thanks!

 

 

[unclewebb]

Your SA voltage appears to be way too low. If I was having stability issues, I would definitely be going into the BIOS to see if I could find this setting and I would increase it. 

 

3 minutes ago, iransofaraway said:

ASRock issue

Lots of manufacturers do dumb stuff. Maybe someday they will change the default voltage setting for SA. They screwed up the thermal throttling temperature so who knows what else they screwed up. 

 

Have a look around the internet to see what other users and what other boards are using for SA voltage. Other boards are probably setting this higher. 

 

Try my Prime95 start and stop stability test. 

[sh10151]

Okay, I'm dumb. I just realized I wasn't running XMP. I went and enabled that in BIOS (and it also turned off spread spectrum for me), and I rebooted. Here's the updated voltages, which probably are more in line with what you were expecting?

 

I have the "Gear mode" set to Auto (default) - I read there's some newer like "Gear 1" vs "Gear 2" settings for XMP - should I be messing with that? Or just leave at auto?

 

Also @unclewebb do you have a way for me to buy you a beer or coffee  or something since you've been so kind to help me out? Lol.

 

[unclewebb]

I was just reading up on XMP and how some motherboards crank up the SA voltage when XMP is enabled. First SA was too low and now if anything, it might be a little too high. Give that a try for a day or two and see how stability goes. Manually setting SA to 1.30 V might be a good compromise but do not go looking for perfect until you find 100% stability.

 

18 minutes ago, iransofaraway said:

buy you a beer or coffee

Forums are about helping each other out. I think everyone here learned something new today about SA voltage. 

 

How about pay it forward and help someone in your local community. Buy someone that looks hungry something to eat. 

 

Leave Gear mode on Auto at the moment. I will go do some reading about that next. 

[-rascal-]

@unclewebb @iransofaraway

 

Is VCCIO/VCCSA different/higher on 11th Gen Intel?

Because from what I know, for 8th/9th (and 10th) Gen Intel, VCCIO/VCCSA shouldn't be so high...like should not exceed ~1.30V.

 

I'm running a manual CPU and DRAM OC, with my i7-8086K, with VCCIO/VCCSA at .. 1.15V?

If you you run too high of a voltage, you risk degrading the CPU silicon reeeaaal fast.