WHEA-Logger Event 19 errors on 3900X

[erek]

"Version 3401 Beta Version
2021/03/20 20.43 MBytes
ROG CROSSHAIR VIII HERO(WI-FI) BIOS 3401
"- Update AMD AM4 AGESA V2 PI 1.2.0.1 Patch A
- Fix USB connectivity issue
Before running the USB BIOS Flashback tool, please rename the BIOS file (C8HW.CAP) using BIOSRenamer.

Please note that this is a beta BIOS version of the motherboard which is still undergoing final testing before its official release. The UEFI, its firmware and all content found on it are provided on an “as is” and “as available” basis. ASUS does not give any warranties, whether express or limited, as to the suitability, compatibility, or usability of the UEFI, its firmware or any of its content. Except as provided in the Product warranty and to the maximum extent permitted by law, ASUS is not responsible for direct, special, incidental or consequential damages resulting from using this beta BIOS.""

 

[Deleted member 205776]

That's the same latest BIOS of my ROG Strix B550-F. I avoid beta BIOSes so I'll sit this one out.

 

[tabascosauz]

That's the same latest BIOS of my ROG Strix B550-F. I avoid beta BIOSes so I'll sit this one out.

At this point they're only really actively working on firmware for the 5000 chips. If you have had Bus/Interconnect on your 3900X since around 1.0.0.2, then chances are it's not going to go away unless you change something. I've not had any experience where firmware has successfully rectified Bus/Interconnect that's a visibly hardware issue. Yes, better AGESA has fixed glaring problems where the chip is spitting WHEAs every minute like a machine gun, but that's just clearly abnormal behaviour.

What is your VDDG IOD and VDDP? Take them down a few notches see if it helps, especially IOD, whereas VDDP can stay at 0.9V

With any change you make, itll have to be a long term game, Bus/Interconnect only rears its head whenever it damn well pleases

 

[Deleted member 205776]

At this point they're only really actively working on firmware for the 5000 chips. If you have had Bus/Interconnect on your 3900X since around 1.0.0.2, then chances are it's not going to go away unless you change something.

What is your VDDG IOD and VDDP? Take them down a few notches see if it helps, especially IOD, VDDP can stay at 0.9V

I'll check when I'm home.

All I know is that my board auto set 1.1 for the SoC and I recall seeing a 1.0477 in Ryzen Master

 

[tabascosauz]

I'll check when I'm home.

All I know is that my board auto set 1.1 for the SoC and I recall seeing a 1.0477 in Ryzen Master

Good rule of thumb to keep any VDDG at least 0.05V below VSOC or else the board may ignore what you set and do whatever the hell it wants.

The 1.05V is probably VDDG CCD, that's fine. If it's VDDG IOD, no reason it should have to be higher than 0.95V @ 3600.

 

[Mussels]

Good rule of thumb to keep any VDDG at least 0.05V below VSOC or else the board may ignore what you set and do whatever the hell it wants.

The 1.05V is probably VDDG CCD, that's fine. If it's VDDG IOD, no reason it should have to be higher than 0.95V @ 3600.

i dont even know what those voltages are or do, i'm super good at this stuff

 

[tabascosauz]

i dont even know what those voltages are or do, i'm super good at this stuff

It's just stuff derived from VSOC. Don't poke if it's not broke, that sorta deal. My 5900X works fine with the same stock Asus CLDOs, so I don't touch it.

The traditional wisdom from somewhere is to leave CCD where it is usually and trim unnecessary IOD. Some say raise VDDP over 3600, I wouldn't know. All I know is that this is the only thing that for sure eliminated Bus/Interconnect from my ailing 3700X in the last few months i had it. ymmv

 

[Deleted member 193792]

1.2.0.2 should have the fix, that's the one with C-state changes (for the USB issues, but they may help you)

Will MSI X470 mobos get that AGESA too?

 

[Mussels]

Will MSI X470 mobos get that AGESA too?

Slower, but they should

 

[Deleted member 205776]

It's just stuff derived from VSOC. Don't poke if it's not broke, that sorta deal. My 5900X works fine with the same stock Asus CLDOs, so I don't touch it.

The traditional wisdom from somewhere is to leave CCD where it is usually and trim unnecessary IOD. Some say raise VDDP over 3600, I wouldn't know. All I know is that this is the only thing that for sure eliminated Bus/Interconnect from my ailing 3700X in the last few months i had it. ymmv

Ok I'm home.

I think my BIOS auto set both VDDG CCD and VDDG IOD to 1.050v, because Ryzen Master was showing 1.0477v before.

Now that I left VDDG CCD as is (1.050v) and set VDDG IOD to 0.950v, while VDDP is at 0.900v still, it shows this:

I guess it only shows VDDG IOD? If so, that means my board auto set the IOD to 1.050v. Wtf.

I also enabled C-states to try my luck again. If I still get them, I'll disable c-states and see if only the voltage tweaks fixed it. If this doesn't fix it, I dunno. Can't RMA.

 

[tabascosauz]

Ok I'm home.

I think my BIOS auto set both VDDG CCD and VDDG IOD to 1.050v, because Ryzen Master was showing 1.0477v before.

Now that I left VDDG CCD as is (1.050v) and set VDDG IOD to 0.950v, while VDDP is at 0.900v still, it shows this:

I guess it only shows VDDG IOD? If so, that means my board auto set the IOD to 1.050v. Wtf.

I also enabled C-states to try my luck again. If I still get them, I'll disable c-states and see if only the voltage tweaks fixed it. If this doesn't fix it, I dunno. Can't RMA.

At some point they removed the ability for software to read separate IOD and CCD in Windows on Ryzen 3000. Now only Ryzen 5000 can do so. What you set in BIOS still applies, you just have to go to BIOS to see what you set.

Probably still want to leave C-states of, doesn't negatively affect performance just draws a bit more power at idle. I'm curious why not RMA?

 

[Deleted member 205776]

At some point they removed the ability for software to read separate IOD and CCD in Windows on Ryzen 3000. Now only Ryzen 5000 can do so. What you set in BIOS still applies, you just have to go to BIOS to see what you set.

Probably still want to leave C-states of, doesn't negatively affect performance just draws a bit more power at idle. I'm curious why not RMA?

Because dumb me sort of destroyed the serial sticker on the packaging and now they won't accept it. Already tried to RMA it back when I was pulling my hair out with the Gigabyte mobos.

Drawing less power at idle would be a bonus for me, so it would be nice if they work. If I start seeing even one WHEA error, I'm disabling them and testing without. I'll do the "true idle test" at night, where I close all non-essential programs and let it idle overnight.

 

[Selaya]

[ ... ]
Probably still want to leave C-states of, doesn't negatively affect performance just draws a bit more power at idle. [ ... ]

Stupid question but if my cores (of a 3900X) are idling at 2,200 MHz, does that mean C-states are enabled?

 

[Deleted member 205776]

Stupid question but if my cores (of a 3900X) are idling at 2,200 MHz, does that mean C-states are enabled?

If you see a core's state be "Sleep" in Ryzen Master, you have C-States on.

 

[Selaya]

Well you see I'm running 7 and Ryzen Master doesn't work on 7 ...
(took me a force-install, BSOD on boot and safemode uninstall to learn that lesson. xd)

 

[Deleted member 205776]

Well you see I'm running 7

On a serious note, if "Global C-State Control" is either Auto/Enabled in the BIOS then C-States are on. Best way to tell I think. I'm not aware of any tools that would immediately tell you aside from Ryzen Master

 

[tabascosauz]

Because dumb me sort of destroyed the serial sticker on the packaging and now they won't accept it. Already tried to RMA it back when I was pulling my hair out with the Gigabyte mobos.

Drawing less power at idle would be a bonus for me, so it would be nice if they work. If I start seeing even one WHEA error, I'm disabling them and testing without. I'll do the "true idle test" at night, where I close all non-essential programs and let it idle overnight.

Interesting. Were you trying to RMA with the retailer instead of AMD?

AFAIK when you RMA with AMD themselves they only care that the relevant info on the heatspreader is legible and there is no physical damage. I don't think you even need the box, you just need the tray and appropriate protective packaging to send it back. But you may need to follow their specified troubleshooting procedures before you can send it back.

Stupid question but if my cores (of a 3900X) are idling at 2,200 MHz, does that mean C-states are enabled?

If that's from HWInfo, then no that indicates more what power plan you're using. On 1CCD parts, idle SVI2 Vcore below 0.9V can suggest cores C-states on, still not sure myself about 2CCD.

 

[Deleted member 205776]

Interesting. Were you trying to RMA with the retailer instead of AMD?

AFAIK when you RMA with AMD themselves they only care that the relevant info on the heatspreader is legible and there is no physical damage. I don't think you even need the box, you just need the tray and appropriate protective packaging to send it back. But you may need to follow their specified troubleshooting procedures before you can send it back.

Yes the retailer.

I might try with AMD if this problem doesn't go away and starts causing BSODs and whatnow. So far, no system corruption at all. Bizzare how a bus/interconnect error showed up after 2 months of being rock solid stable.

 

[mb194dc]

Well you see I'm running 7 and Ryzen Master doesn't work on 7 ...
(took me a force-install, BSOD on boot and safemode uninstall to learn that lesson. xd)

The original versions of Ryzen Master do and you can change clocks on the fly, without reboot. On Ryzen 1 anyway. I dual boot and it works on Both. Run my Ryzen at stock and just overclock it for gaming when need more power.

 

[tabascosauz]

Yes the retailer.

I might try with AMD if this problem doesn't go away and starts causing BSODs and whatnow. So far, no system corruption at all. Bizzare how a bus/interconnect error showed up after 2 months of being rock solid stable.

I mean if you've done all your testing then the RAM is probably as stable as you say. Unfortunately, the Infinity Fabric always has its own ideas.

The only surefire test for IF is time. Anecdotally, LinX works for the UMC and perhaps IF by extension, and randomx-stress has been suggested for IF stability under load, but that's about it.

This happened all the time with my 3700X. Weeks or months no problem, then some Bus/Interconnect, my SN750 disappears, I reboot and no POST until I wipe my BIOS settings or reflash. Only thing that worked was VDDG IOD.

 

[Deleted member 205776]

RAM is stock XMP except tRFC halved from 630 to 324.

 

[tabascosauz]

RAM is stock XMP except tRFC halved from 630 to 324.

Did you apply the VDDGs yet? I got mine backwards. I have VDDG CCD @ 0.95 and VDDG IOD @ 1.047. Yeah, I'm stupid, I can never seem to remember which is which.

The guide does suggest lower IOD, however. If you have unrestricted access to a clear CMOS button or the jumper pins, I'd give both ways a try.

Or you could run them both at 0.95V, see where that gets you. There's some anecdotal merit out there to running 3 all of them, both VDDGs and VDDP, as low as you can go. BZ says 3600 shouldn't need more than 0.9V for all 3.

 

[Deleted member 205776]

Did you apply the VDDGs yet? I got mine backwards. I have VDDG CCD @ 0.95 and VDDG IOD @ 1.047. Yeah, I'm stupid, I can never seem to remember which is which.

The guide does suggest lower IOD, however. If you have unrestricted access to a clear CMOS button or the jumper pins, I'd give both ways a try.

Or you could run them both at 0.95V, see where that gets you. There's some anecdotal merit out there to running 3 all of them, both VDDGs and VDDP, as low as you can go. BZ says 3600 shouldn't need more than 0.9V for all 3.

I have this config right now:
VDDG CCD = 1.050v
VDDG IOD = 0.950v
CLDO VDDP = 0.900v

with C-states on, and the overnight idle test passed with flying colours, no WHEA errors or reboots when it usually would've gave me one. Still too early to declare victory, I'll keep doing more idle tests every night from now on.

EDIT: So, I'm only day 2 into testing the -100mV reduction on VDDG IOD and the effect it has on WHEA/idle reboots with C-states on, but so far, all seems well.

If this is stable finally, would there be any advantage to me lowering SoC , CLDO VDDP, and both VDDGs, as low as possible? Or should I just keep my current config of SoC 1.1, VDDG CCD 1.05, VDDG IOD 0.950 & VDDP 0.900?

 

[Bpw87]

So tonight I randomly have the same WHEA 19 happening to my system with 3 shut downs in not idle states but very low processor duties.

I have a 5600x, b450-i Strix not sure on the bios version but it was the latest stable version as of the start of February. I had global c states force enabled, I put it back to auto prior to seeing this post as troubleshooting as that's the last change I made to the bios. In saying that it was multiple weeks ago and no errors till today. I do a lot of RAM overclocking and running 3933mhz 17 17 17 34 at 1.385v with 1966.7 fclk. Pretty much every sub timings been changed and a lot of the other settings. SoC is at 1.1v, VDDP is 0.9v. My ram oc was 100% avx1 stable, many many hours of aida64 and memtest including hours of 100% load on the cpu and gpu while stressing the ram.

Initially I thought my issue was tied to my portable ssd as it was being used when I first got the error. It also seems like it slowed down to usb2.0 speeds around 20 Mbs when it's a 3.1 gen2 and regularly sees 500Mbs.

I'll try disabling the c states tomorrow morning. If that doesn't work I'll try a new profile with default bios. At the moment I'm worried I might corrupt something so I'm trying to limit doing anything important.

 

[Space Lynx]

So tonight I randomly have the same WHEA 19 happening to my system with 3 shut downs in not idle states but very low processor duties.

I have a 5600x, b450-i Strix not sure on the bios version but it was the latest stable version as of the start of February. I had global c states force enabled, I put it back to auto prior to seeing this post as troubleshooting as that's the last change I made to the bios. In saying that it was multiple weeks ago and no errors till today. I do a lot of RAM overclocking and running 3933mhz 17 17 17 34 at 1.385v with 1966.7 fclk. Pretty much every sub timings been changed and a lot of the other settings. SoC is at 1.1v, VDDP is 0.9v. My ram oc was 100% avx1 stable, many many hours of aida64 and memtest including hours of 100% load on the cpu and gpu while stressing the ram.

Initially I thought my issue was tied to my portable ssd as it was being used when I first got the error. It also seems like it slowed down to usb2.0 speeds around 20 Mbs when it's a 3.1 gen2 and regularly sees 500Mbs.

I'll try disabling the c states tomorrow morning. If that doesn't work I'll try a new profile with default bios. At the moment I'm worried I might corrupt something so I'm trying to limit doing anything important.

go back to xmp ram. your issues will probably go away.