Weird wattage spike on the 7800X3D in Open Hardware Monitor

[StoleMyOwnCar]

I'm not sure if this is a problem or just a glitch in the program (hope it's the latter). Unfortunately I didn't take a picture of the actual spike (I restarted the program to test if the "max" value persisted between restarts because I wanted to see if it was from before I updated the BIOS), but I underlined which value I'm talking about in this picture:

What happened was I randomly looked over at it and looked at the max wattage of the CPU Package in OpenHardwareMonitor and noticed that it was around 4800 watts. Yes, 4.8kW. Some of the individual cores had wattages ~180 as well. I don't know when this usage happened. Obviously it didn't actually happen or I think my entire case and desk would have melted and/or burned down. I keep my computer on 24/7 and it has been running stable since I built it about 4 days ago. I haven't had weird crashes in games or stress testing. The motherboard is on the latest bios. Obviously I have heard about the AM5 boards randomly burning CPUs, but as I understand it this was mainly an issue on ASUS boards. I did also notice that I had a max value of CPU package temperature around 86C; I'm not sure if it happened at the same time, but 86C is still within the expected thermal envelope for this chip anyway...

So I don't think it actually pulled 4.8kW, but question is what actually happened, then...? Has anyone seen this weirdness with AM5 and OpenHardwareMonitor themselves? I smelled around in my case after opening the side panel, and then looked at the back of the PCB behind the CPU (or what I could see near the backplate window). I didn't notice any burning smell, nor any bulges or anything odd. Games are still running fine. Stress tests only push it to ~92W, as expected. I'm kind of confused. Is there anything I should test?

 

[KickAssCop]

Seems like a reading error. Nothing in a PC can pull 4.8K watts lol.
Will watch this thread as I am getting motherboard from the same manufacturer.

 

[LigTasm]

CPU burn issue was related to vSOC, as long as you're under 1.300V or so there it'll be fine. This seems just like the program didn't read the sensor correctly for that polling cycle.

 

[StoleMyOwnCar]

CPU burn issue was related to vSOC, as long as you're under 1.300V or so there it'll be fine. This seems just like the program didn't read the sensor correctly for that polling cycle.

Yeah that's what I've read. Unfortunately OpenHardwareMonitor doesn't do SOC voltage or any voltage readings, so I had to download HWmonitor. I then decided to do a stress test.

My SoC voltage is pretty much always right at 1.3, maybe 1.302-1.304 as shown in this picture.

According to this thread I found from googling, another user is having the same showing (well his is higher at 1.318):
https://www.reddit.com/r/AMDHelp/comments/146ujg3/7950x3d_soc_voltage_still_13_on_latest_msi_beta/
But a poster in said thread is saying that this is just likely the motherboard overreporting the voltage.

That's kind of curious. Maybe I should go into the BIOS and set it slightly below 1.3, like 1.29 or something? I find it odd that we're all the way out like 5 months from when this burning stuff happened and they would set it even close to 1.3, which was clearly a problem area.

All of this is probably completely unrelated to the 4.8kW.

Seems like a reading error. Nothing in a PC can pull 4.8K watts lol.
Will watch this thread as I am getting motherboard from the same manufacturer.

Yeah, I kind of thought it was a reading error, but just in case it's a good idea to give the thing a lookover. I've never had a reading like that on my old 5950X the entire time I was running it, as far as I'm aware...

 

[michalrz]

It most likely didn't pull that much.
I think you just had the rare opportunity to see some stray particle from the sun flipping a bit in just the right spot in-memory.
Check system log for MCE (machine check exceptions).

 

[StoleMyOwnCar]

It most likely didn't pull that much.
I think you just had the rare opportunity to see some stray particle from the sun flipping a bit in just the right spot in-memory.
Check system log for MCE (machine check exceptions).

If I listed by Source, would it be in the M's, or does it actually read like something else? I am having trouble locating anything:

I haven't had any blue screens that I can remember. Like I said, I leave this system on 24/7, so I would generally know if I came back and the things that I was expecting to still be pulled up, were not there.

I dunno, I guess something weird just happened with the program. For now I'm going to just leave the system going with both CPUID HWMonitor going and Open Hardware Monitor going. Since OHM doesn't have a voltage readout.

Again, is ~1.298-1.304V on the "CPU NB/SOC" reading on HWMonitor normal? It was my understanding that less than 1.3 was ideal...

 

[michalrz]

If I listed by Source, would it be in the M's, or does it actually read like something else? I am having trouble locating anything:
View attachment 596514

I haven't had any blue screens that I can remember. Like I said, I leave this system on 24/7, so I would generally know if I came back and the things that I was expecting to still be pulled up, were not there.

I dunno, I guess something weird just happened with the program. For now I'm going to just leave the system going with both CPUID HWMonitor going and Open Hardware Monitor going. Since OHM doesn't have a voltage readout.

Again, is ~1.298-1.304V on the "CPU NB/SOC" reading on HWMonitor normal? It was my understanding that less than 1.3 was ideal...

Machine Check Exceptions and other hardware weirdness can occur without Windows needing to stop the machine.
These can be benign, some ECC corrective action happened, etc. Try looking for "WHEA" as the "source".
("Windows Hardware Event Architecture") or just a short "mce" somewhere.

But - that might have made its way to the logs, but might have not, too.
I also assume some of these numbers (voltage readings, motherboard electric stuff) are provided sometimes by some chip on the mobo other than the CPU, so there's that also.

I can't speak to the overvolting issues on some mobos, or the specific voltages your CPU should run at - I haven't seen one IRL yet. But, if you're worried, keep at it. Have the monitor open. Maybe launch some "Live" (non-installing) version of Linux and leave it running overnight, after which open the terminal and type "dmesg -T". It will look like some ancient aliens scrolls, but there might be a human-readable description of a hardware error it noticed.

 

[StoleMyOwnCar]

Machine Check Exceptions and other hardware weirdness can occur without Windows needing to stop the machine.
These can be benign, some ECC corrective action happened, etc. Try looking for "WHEA" as the "source".
("Windows Hardware Event Architecture") or just a short "mce" somewhere.

I believe these are what you are describing (both the "System" heading and the "Administrative Events" heading only show these two). The last instance of a WHEA error across both this X670E rebuild and the X570S I had previously appears to be in 2022. Nothing recent.

I guess it's just a weird sensor error...? Either way, I'll keep the sensor apps open to see if it repeats itself. I have Open Hardware Monitor doing a text log as well, so that I can catch exactly when it happens, and for how long, if it reoccurs. I guess we'll see. None of my games have had any crashing or anything either, so far. Currently starting Starfield...

 

[KickAssCop]

I just dropped by to say Fuck WHEA errors!

 

[StoleMyOwnCar]

Now I'm suddenly having issues. The problem is that I can't tell what could possibly be causing them. I'm just posting in this topic because it's still open and it MIGHT be related. I'm having some "Nvlddmkm" (or similar) crashed and recovered errors in event viewer. I can show them in a bit, but I'm running a memtest86 to rule out memory errors. I'm not sure if that's the actual name of the .sys that's crashing. I tried doing DDU and doing a fresh install of current Nvidia drivers, no luck. I haven't had that weird wattage spike again and all of the CPU temps look fine, though.

I know it's a GPU driver crash, but until I can rule out this motherboard and RAM, since they're both open box, I'm putting a bit more scrutiny on them. The problem is that this error has never popped up on my 3080 Ti before in the past the entire year and a half or more that I have had it. I just started playing Starfield yesterday. It played fine for several hours. Then I think at some point I did get a crash in it. Afterwards I played Divinity Original Sin 2 with my friend for several hours (as we had been doing for about the entire past half week), and all of a sudden I got that crash message. Never happened in DOS2 before. Now I come back to Starfield and it also gives me that error message during weird times, like alt tabbing back in after alt tabbing out, or during loading, or just randomly. The frequency seems to be increasing.,

I'm hoping Starfield didn't kill my 3080 Ti. I know that was an issue with some Amazon game; it was killing GPUs. But this could also be a CPU/MB/RAM issue.

I could try putting my old 5950X into another case and seeing if the game runs fine on it instead after transplanting the NVME, but the problem is the 5950X can't actually utilize my 3080 Ti to the degree the 7800X3D can in Starfield, so even assuming I have no crashes, I can't even be sure that I need to return them... sigh...

Aside from Memtest, I don't know where to even go with this. Thoughts? I could also try buying a new GPU to see if it fixes it, but I'm not sure how many returned 4090s Microcenter needs from me before they get a bit tired of it.

 

[KickAssCop]

Need to do process of elimination.

1) Did you install new drivers recently? Was your windows install new or ported over? Did you clean install drivers using DDU? This should rule out any Nvidia software errors. You can also try older driver version to at least ensure DOS 2 works (even if Starfield doesn't) to isolate the issue.
2) Did you update windows with latest beta (for me this caused instability and I had to roll back especially the last 2 KB updates).
3) Is anything overclocked in your system?
4) Is EXPO running and if yes, is it set to profile 2 in XMP/EXPO?
5) Did you play around with memory timings?
6) If you install only one stick of ram do you have same errors?
7) Do you have latest AMD chipset drivers? Was there a new install of AMD chipset drivers?
8) Did you update motherboard bios? Is this a beta or final bios? Search online if new bios is causing instability?
9) Did you play with SOC voltage at all? If yes, what did you set it at? Starfield is expected to push the system (low system SOC voltage can exhibit weird behaviors from hitching to game crashes)?
10) Is Starfield causing crashes in other PCs as well? Try to check online if this is the case (Steam forums are a good source).
11) If none of the above works then it is quite likely your open box motherboard that might not be playing well with some ram/or specific game situations that MC won't have had the opportunity to test when they put it back on sale.

My 2 cents. It's a bitch to debug such problems and one of the reasons why I was hesitant on buying a 7800X3D platform. Oh well, gonna know what's what next week.

 

[StoleMyOwnCar]

Need to do process of elimination.

1) Did you install new drivers recently? Was your windows install new or ported over? Did you clean install drivers using DDU? This should rule out any Nvidia software errors. You can also try older driver version to at least ensure DOS 2 works (even if Starfield doesn't) to isolate the issue.
2) Did you update windows with latest beta (for me this caused instability and I had to roll back especially the last 2 KB updates).
3) Is anything overclocked in your system?
4) Is EXPO running and if yes, is it set to profile 2 in XMP/EXPO?
5) Did you play around with memory timings?
6) If you install only one stick of ram do you have same errors?
7) Do you have latest AMD chipset drivers? Was there a new install of AMD chipset drivers?
8) Did you update motherboard bios? Is this a beta or final bios? Search online if new bios is causing instability?
9) Did you play with SOC voltage at all? If yes, what did you set it at? Starfield is expected to push the system (low system SOC voltage can exhibit weird behaviors from hitching to game crashes)?
10) Is Starfield causing crashes in other PCs as well? Try to check online if this is the case (Steam forums are a good source).
11) If none of the above works then it is quite likely your open box motherboard that might not be playing well with some ram/or specific game situations that MC won't have had the opportunity to test when they put it back on sale.

My 2 cents. It's a bitch to debug such problems and one of the reasons why I was hesitant on buying a 7800X3D platform. Oh well, gonna know what's what next week.

1. I tried reinstalling the drivers earlier with DDU. I unplugged the network cable so Windows wouldn't autoinstall some. It crashed in the game anyway. The Windows installation is indeed carried over from my old one on the 5950X
2. I'm on Windows 10. I rarely update it and I'm not sure what version it's on. I don't remember opting into any betas.
3. Not really, nope.
4. RAM sticks are in EXPO/XMP 1.
5. Nope not yet, just the default XMP settings the sticks came with.
6. That would be hard to test. I'm not sure if the system would be stable with just one stick in, while trying the same memory profile?
7. That's a good question. The motherboard automatically prompted me to install the chipset drivers at some point, but I'm unsure if they're the latest ones. I can try updating them.
8. I booted the system twice before updating the BIOS just to make sure it would post when first building. After that, I promptly updated it to latest BIOS. 7D70v18 is the version number.
9. I didn't touch it. The SOC voltage seems to be staying at 1.3-ish, which if anything might be uncomfortably high... https://www.reddit.com/r/Amd/comments/12yq4yb/megathread_for_am5_ryzen_7000/jxrow1f/
10. I haven't tested it on any other system yet.
11. So.. there are a lot of issues it could be. memtest86 is on like its 4th pass at the moment, and has had no errors. I'm kind of baffled. Maybe my GPU really is simply dying, and the timing is coincidental. Or maybe it really is simply a driver error.

What I'm sort of worried about is that when this motherboard first started it up, it had some warning message about some failed overclock profile that the previous user probably set. I probably should have popped the battery. Afterwards, it posted fine. Then it posted fine again after I set the XMP profile. But thinking back on it, the BIOS was from March 2023? So perhaps it briefly did spike the CPU with hazardous voltage... which, as I understand it, will kill the thing regardless. What's annoying is that I had played on this MB for multiple days up until this happened yesterday, and had absolutely no issues (again, including Divinity OS 2). The only thing I can think that changed is that I started up and tried Starfield once. I'm worried Starfield somehow permanently damaged my GPU, so now it's slowly failing.

I have 15 days to return the whole set, and honestly I'm kind of debating doing it. With coming into the office, I just don't have the time or energy to troubleshoot this shit all the time. Maybe I should just reduce the graphics and eat the FPS loss on my 5950X. As much as I love the smoothness on the 7800X3D, I would like to just play without having to worry if my investment is a mistake, and all of the motherboards this gen are just such ripoffs at normal prices. Saving money for the next set of motherboards that have these issues ironed out might be for the best.

Does Starfield have a "benchmark loop" mode? This is annoying to try to catch while "playing" whereas in reality I'm not really "playing" but "waiting for an error to happen". I need something Starfield-esque that's demanding and will cause the crash...

 

[KickAssCop]

Bios reset and chipset drivers from AMD website seem like your first order of tests to do. AMD chipset drivers suck ass just like every other software from AMD. So they might have released new chipset drivers that are needed to use Starfield. Same thing happened to me when Jedi Survivor came out and I had to update chipset drivers just to get it to run properly.

Also Windows 10 is deprecated in my opinion. New games with new features desperately need Windows 11. Even performance delta you will notice by loading up Windows 11. I strongly suggest it. There are enough mods to make it look like Windows 10 btw.

Same for bios, make sure it is a stable bios (see forums) and load it up.

Final thing from my side would be to set XMP 2 and not 1. XMP 1 is finicky as fuck all on AMD. I have tried this with 3 completely different set of rams on the MSI motherboard that I had X570 Tomahawk and XMP 1 would always error out, cause hitches, crash games, the works. XMP 2 works fine.

Hopefully once I get my system maybe I can hep more meaningfully.

 

[StoleMyOwnCar]

4 passes, 7 hours of testing, didn't find anything. This is annoying.

Bios reset and chipset drivers from AMD website seem like your first order of tests to do. AMD chipset drivers suck ass just like every other software from AMD. So they might have released new chipset drivers that are needed to use Starfield. Same thing happened to me when Jedi Survivor came out and I had to update chipset drivers just to get it to run properly.

There were only a few things slated for update with the download, but I installed them. Kind of not sure if this will fix anything, but worth a shot I guess.

Also Windows 10 is deprecated in my opinion. New games with new features desperately need Windows 11. Even performance delta you will notice by loading up Windows 11. I strongly suggest it. There are enough mods to make it look like Windows 10 btw.

I don't have a Windows account and Microsoft is being crappy with honoring my Education key from college now. I had to just get an activator because the phone support line didn't work. I'm not sure how I would get 11, and I sure as heck am not paying for it. I'll have to look at my options.

Final thing from my side would be to set XMP 2 and not 1. XMP 1 is finicky as fuck all on AMD. I have tried this with 3 completely different set of rams on the MSI motherboard that I had X570 Tomahawk and XMP 1 would always error out, cause hitches, crash games, the works. XMP 2 works fine.

There is no profile 2, unless you're talking about something else. This is the only profile these sticks came with. Exact model:
https://pcpartpicker.com/product/hR...dr5-6000-cl30-memory-f5-6000j3040g32gx2-tz5nr

One more thing I could try is this:
https://www.reddit.com/r/nvidia/comments/12l01wf/nvlddmkm_4090_crash_solved/jg4ygjh/

They said they had success with changing the file permissions on that .sys file for some reason. I have no idea why that would make any difference, but worth a try. I appreciate the aid.

Hopefully once I get my system maybe I can hep more meaningfully.

https://www.reddit.com/r/Amd/comments/12yq4yb/megathread_for_am5_ryzen_7000/juy8g4w/
One more thing, be careful with your Tomahawk.

 

[KickAssCop]

SOC voltage is first on my list to set manually. 1.25 volts should be good enough.
I am a bit surprised by only having single memory profiles but it also seems like memory is not the issue here.
My bet then is windows 10 and/or Nvidia drivers not working well with windows 10 for this game as possible issues.
I am pretty sure you can get a windows 10/11 key for 10-20 bucks from people here. There was an FS thread. Alternatively, you can get it from grey market sites if you want. I never paid for my windows and no longer remember where I got it from lol.

Correction: I paid for Windows 7 -> upgraded to 8 -> upgrade to 10 -> upgraded to 11 so my key carried on for at least 1 copy of windows. I think I have at least a couple of other copies (1 from a guy who was selling on Hardforums).

 

[StoleMyOwnCar]

Yeah I'm probably overreacting and just assuming it's the motherboard/RAM. It's rock solid in every stress test I've tried so far. I have not had the issue again quite yet after trying out that file permissions thing that I linked in the thread, which is weird. I also updated the chipset drivers though, which like you said

I might want to lower that SoC voltage, though. It's right at 1.3....

What I'm sort of worried about is that when this motherboard first started it up, it had some warning message about some failed overclock profile that the previous user probably set. I probably should have popped the battery. Afterwards, it posted fine. Then it posted fine again after I set the XMP profile. But thinking back on it, the BIOS was from March 2023? So perhaps it briefly did spike the CPU with hazardous voltage... which, as I understand it, will kill the thing regardless.

This part is really what worries me, I guess. Does AMD warranty CPUs that get exploded like that?

Well guess I could try running prime95 at the same time as I'm idling in Starfield while I sleep...

 

[StoleMyOwnCar]

Starfield didn't crash while I was asleep but still crashed while I was playing it a bit after waking up... but that said I don't think there's anything wrong with the CPU or motherboard now. Divinity Original Sin 2 ran for about 5 hours straight with no issues with my friend, so I think that other crash was just random. I also just finished playing Hogwarts Legacy for about 7+ hours straight. Again, no crashes and no issues. Pretty sure it's some Starfield-specific driver crap or something. Might even be the DLSS mod. Worst case, I guess I could try RMAing the GPU to EVGA. Its stock settings might just be unstable in some games, sporadically.

 

[StoleMyOwnCar]

Played games all day today without any crashes. Starfield didn't crash either, after disabling DLSS mod. So... kind of sure that was the issue. About 99% sure there's nothing wrong with CPU or motherboard now. Guess this open box Carbon for 200 bucks is a champ after all.

Only thing is, not sure about that SOC voltage at exactly 1.3... doesn't seem to be causing issues, but isn't it a tad on the high side...?

Edit: Also am discovering that power reporting on Open Hardware Monitor just appears to be unreliable in general:

Not as large of a spike, but there's not really a way for this CPU to go to 130W. HWMonitor shows the correct value. Guess that was the issue.

 

[riev90]

StoleMyOwnCar have you tried using HWInfo?
I haven't used HWMonitor since 2018 (when I started using Ryzen) and HWInfo is more accurate compared to other similar monitoring software.

Also, 1.3v SoC for daily use is safe.
My Ryzen 7500F is running with 1.22v on my overclock Hynix A-Die at 7600 with timings 38-47-47 1T.

 

[KickAssCop]

Oh HWMonitor. Yeah that software is fucked.

Use HWInfo.

 

[StoleMyOwnCar]

StoleMyOwnCar have you tried using HWInfo?
I haven't used HWMonitor since 2018 (when I started using Ryzen) and HWInfo is more accurate compared to other similar monitoring software.

Also, 1.3v SoC for daily use is safe.
My Ryzen 7500F is running with 1.22v on my overclock Hynix A-Die at 7600 with timings 38-47-47 1T.

No I have not; admittedly I usually just open up Open Hardware Monitor and call it a day. It's never given me strange readings when on any other platform, but I guess there's just something odd about AM5. Was not aware that HWMonitor had issues, too. I guess I'll try downloading HWInfo when I get a chance, thank you.

As far as SoC voltage, I decided to just decrease it to 1.25 just in case, and the system has been rock solid regardless. Starfield hasn't crashed. Well it did sort of soft-hang (ie I still had animations but it wouldn't get out of the loading screen) once while loading, but... who cares, right? Still an accomplishment, considering I leave the game open all night and all day and just pick it up at random times.

 

[StoleMyOwnCar]

Welp, computer just restarted randomly. I had it running for a two weeks straight... and had Starfield open much of the time. Been playing for the past 3 days, completed various quests, basically never turned the game off. I alt tabbed out of Starfield and was about to walk away since I'd been looking at the phone for a bit.
The computer looked like it was just shutting off the screen, but next thing I saw was just the flashing black bar indicating startup. There were no errors in the Windows event viewer. There was no blue screen. There was nothing, just randomly decided to restart. Restarted just fine, no warnings or anything.

I'm confused. It's like someone just decided to hit the reset switch on the case. There were no abnormalities, it just restarted, and then booted up completely normally. I could chalk it up to a random event, or maybe it's related to me using Buildzoid's timings. Either way, it's very rare considering it hadn't happened until two weeks after I changed those timings (with the computer running without pause), but it still bugs me that I can't even begin to figure out how to try to troubleshoot this. There's nothing that I can really do since it didn't leave any errors behind or any trail for me to even follow.

 

[riev90]

Can you run zentimings (download the latest one) and then post the screenshot here?
I'm quite curious what is your ram settings.

 

[KickAssCop]

My computer also started giving blue screens. I turned off Memory Context Restore to ensure that it doesn't do it again. Will keep checking if this fixed the issue. I also came back to my computer after a week.

 

[StoleMyOwnCar]

Can you run zentimings (download the latest one) and then post the screenshot here?
I'm quite curious what is your ram settings.

I don't have time at the moment, but this might suffice in the meantime:

I'm just running Buildzoid's low effort Hynix settings, but at 1.39v instead of my ram's default XMP settings of 1.4v.

As I said, it took two weeks to even get one instance of this happening so not sure if it's worth pursuing or if it's even the RAM... I thought RAM usually gave a blue screen when the timings were bad or unstable...

My computer also started giving blue screens. I turned off Memory Context Restore to ensure that it doesn't do it again. Will keep checking if this fixed the issue. I also came back to my computer after a week.

Well I didn't even get a blue screen. This took two weeks to actually happen once... and the thing just outright restarted like I pressed the reset button, as I said. I've basically never seen this before and don't know where to go with troubleshooting.

 

[riev90]

Assuming your kit is Hynix M-Die I would change the timings like this:
tCL 32
tRCDrd 38
tRCDwr 38
tRP 38
trfc 480
tfaw 16
ProcOdt to 43.8 or max 48.
vddio 1.35
vdd & vddq 1.38
vSoC 1.22

PBO set to auto.

Well I didn't even get a blue screen

Thing is some instability on AM5 platform doesn't introduce bluescreen.
Sometimes it restart directly (like in my case, my build successfully passed TM5 (no error) and y-cruncher 2.5b (https://hwbot.org/submission/5356610_riev90_y_cruncher___pi_2.5b_ryzen_5_7500f_1min_13sec_97ms) got the WR on HWBot, but if overclocked using all core to 5300mhz and fixed voltage at 1.32, then it will restart randomly.

 

[StoleMyOwnCar]

Assuming your kit is Hynix M-Die I would change the timings like this:
tCL 32
tRCDrd 38
tRCDwr 38
tRP 38
trfc 480
tfaw 16
ProcOdt to 43.8 or max 48.
vddio 1.35
vdd & vddq 1.38
vSoC 1.22

PBO set to auto.



Thing is some instability on AM5 platform doesn't introduce bluescreen.
Sometimes it restart directly (like in my case, my build successfully passed TM5 (no error) and y-cruncher 2.5b (https://hwbot.org/submission/5356610_riev90_y_cruncher___pi_2.5b_ryzen_5_7500f_1min_13sec_97ms) got the WR on HWBot, but if overclocked using all core to 5300mhz and fixed voltage at 1.32, then it will restart randomly.

Thanks for all of this info. I have my RAM voltage at 1.39 at the moment, because it's dual rank so I just assumed it would need more voltage to be stable at his timings.

To be quite honest, if it only happens once every 2 weeks, I'm not sure how much I care. But if the next boot happens sooner than that, I'm definitely going to give your timings a shot.

I also did notice the kettle plug going to my PSU was about half an inch unplugged (as in, mostly still in, but a little loose-ish...?). I... don't think that makes any difference, but at the moment all I can think of is either a PSU problem or some weird RAM problem. The motherboard is indeed open box, but I have basically every USB port on it saturated, and every NVME slot saturated, and have had 0 issues with any of it (along with the performance), so... kind of doubt it's the issue. I also ran it for 2 weeks before changing the timings without it crashing, but I was kind of confused because I thought DRAM caused blue screens. Thanks for confirming that on AM5 they don't always cause blue screens, because I was worried about a PSU issue, although this PSU is relatively new (EVGA P2 Platinum 1200W).

 

[riev90]

because it's dual rank so I just assumed it would need more voltage to be stable at his timings.

Unlike DDR4, DDR5 has "sweetspot" for all voltage settings.
Sometimes more doesn't mean better / will be stable.
That's why in the ram timings I posted, I adjusted some primary and secondary tighter, to find that sweetspot.

But if the next boot happens sooner than that, I'm definitely going to give your timings a shot.

Don't forget to save your last "stable" timings into your bios or usb drive in case the timings I posted didn't work / boot.

although this PSU is relatively new (EVGA P2 Platinum 1200W).

I would chalk out the PSU from the culprit list. That's a solid unit TBH.

 

[michalrz]

Welp, computer just restarted randomly. I had it running for a two weeks straight... and had Starfield open much of the time. Been playing for the past 3 days, completed various quests, basically never turned the game off. I alt tabbed out of Starfield and was about to walk away since I'd been looking at the phone for a bit.
The computer looked like it was just shutting off the screen, but next thing I saw was just the flashing black bar indicating startup. There were no errors in the Windows event viewer. There was no blue screen. There was nothing, just randomly decided to restart. Restarted just fine, no warnings or anything.

I'm confused. It's like someone just decided to hit the reset switch on the case. There were no abnormalities, it just restarted, and then booted up completely normally. I could chalk it up to a random event, or maybe it's related to me using Buildzoid's timings. Either way, it's very rare considering it hadn't happened until two weeks after I changed those timings (with the computer running without pause), but it still bugs me that I can't even begin to figure out how to try to troubleshoot this. There's nothing that I can really do since it didn't leave any errors behind or any trail for me to even follow.

I'm in shotgun mode at this point, but could you try... unplugging the reboot button connector from the mobo?
Maybe it got nicked somewhere in the case and is just shorting. You were leaving your station, so... vibration.

 

[mnewxcv]

Seems like a reading error. Nothing in a PC can pull 4.8K watts lol.

Introducing, the GeForce rtx 5090.

 

[StoleMyOwnCar]

I'm in shotgun mode at this point, but could you try... unplugging the reboot button connector from the mobo?
Maybe it got nicked somewhere in the case and is just shorting. You were leaving your station, so... vibration.

The computer isn't actually on the same table as the desk I work on. This is an adjustable height desk, and the computer is on a hall table that's right beside it. I think what's more likely to have happened (assuming that it was a power rather than RAM issue) is my slightly loose kettle plug might have gotten just loose enough to cause it to briefly disconnect? But it doesn't really make sense because the computer wasn't pulling much wattage at the time since I was alt tabbed out of Starfield. I've had it running since then and no issues. I'm going to be kind of curious as to how often this occurs on average.

Unlike DDR4, DDR5 has "sweetspot" for all voltage settings.
Sometimes more doesn't mean better / will be stable.
That's why in the ram timings I posted, I adjusted some primary and secondary tighter, to find that sweetspot.

Aren't tighter timings going to be more demanding? I just know that there was a fellow in the motherboard section that was having issues with their DDR5 RAM on the same motherboard. I think they had Samsung B die, and required more voltage than the XMP profile called for, in order to get it stable. I just figured that RAM needed more to get it stable. The default XMP profile for these is actually 1.4V... I turned it down to 1.39. I guess I'll give 1.35 a shot if this happens again, along with your timings.

Don't forget to save your last "stable" timings into your bios or usb drive in case the timings I posted didn't work / boot.

I think the motherboard has some slots to save it in, thanks for the reminder, I'll need to remember to do that, too.

I would chalk out the PSU from the culprit list. That's a solid unit TBH.

Yeah it's a really well rated unit. Like I said all I could think of is that the kettle plug wasn't quite in all the way but... it feels sort of unlikely.

Assuming it's a RAM issue, I really hope we get more BIOS updates that keep getting it more and more stable... I know AM5 is in its infancy.

 

[StoleMyOwnCar]

Hmm... Well, nothing so far (knock on wood lol). I'm wondering what the interval is on this issue, since it's been ~11 days since that random reboot. Throughout Friday and this weekend, I gamed pretty heavily in Starfield and some Hogwarts, too. Doesn't seem like it's a power issue. Cinebench 24 completed with no issues (1123 or whatever I got). This is just bizarre.

Overall if it's this sporadic, I'm fine with keeping things as they are tbh. All I really did since that last reboot was raising the SOC voltage from 1.25 to 1.26. Would be kind of silly if that fixed it.

 

[riev90]

Overall if it's this sporadic, I'm fine with keeping things as they are tbh. All I really did since that last reboot was raising the SOC voltage from 1.25 to 1.26. Would be kind of silly if that fixed it.

Like I said in the previous post:

Unlike DDR4, DDR5 has "sweetspot" for all voltage settings.
Sometimes more doesn't mean better / will be stable.

And that includes finding the right voltage for your SoC also.
Anyway, I hope that the issues fully resolve so you can enjoy your rig without any concern.

 

[StoleMyOwnCar]

Well it's been about a month and my system has had no sporadic shutdowns and or restarts (unless I come home today and it does it). I have manually shut down my computer a couple of times to change out the GPU to my 4090 and make some tweaks, but it's been rock solid.

I just have to comment somewhere, I'm finding this 7800X3D to be an absolutely amazing beast of a CPU.

I'm playing Starfield, have 200+ Firefox tabs open (some of them youtube), have like 20 other programs open, have Media Player Classic decoding a video, and then also am doing Stable Diffusion on the side. Yes, the Stable Diffusion is going slower because the GPU is getting used for Starfield... but all of this is going at the same time without a hiccup. In fact, the system feels snappier than it ever did under my 5950X. I'm not sure if it's the cache or if it's the buildzoid subtiming tweaks on the ram, or what... but this is insane. All of this power from a 60W CPU. I also feel like this Suprim X Liquid 4090 has also somehow made my system feel more stable, but I don't know how or why it would do that, and it could be purely placebo. I'm kind of convinced this CPU is a modern marvel, as an engineer that loves efficiency.