[SOLVED] Random power loss and immediate restoration (another one!)

[slurmsmckenzie]

Hi all, seems that there are a few other people having similar problems to me, thought I'd add my own thread rather than confuse or hijack someone else's!

I have a seemingly random issue where the power is completely cut from my system (as though you'd flicked the power switch on the PSU) an then restored a couple of seconds later - and the motherboard powers back on and the systems boots to the Windows logon screen. I have a Corsair PSU, and I noticed that the other people having similar issues did too - although of course they will likely be popular PSUs so it is probably just that. My system:

Corsair SF750 PSU (plugged into surge-protected 8-gang trailing adapter)
Maximus Gene VII + i7-4790k (not OC'd)
32GB Crucial RAM (not OC'd)
Samsung NVMe SSD (system drive)
3 x WD Red storage drives
RTX 2060 (Pailt StormX)
Asus Xonar PCI-E soundcard

I put this system together in Feb 2019, previously had an ancient CM Silent Pro 700 PSU and an H87 based board which never saw these problems (ironically I changed PSU because I read everywhere that the CM was known to be bad and might be a ticking time-bomb...). Since 2019 I have changed a few things, but this problem has always happened before and after these changes:

1050ti StormX -> 2060 StormX
16GB Crucual RAM -> 32GB
960 EVO SSD -> 980 PRO (I know this is completely superflouous for my system, but I felt like an update and wanted to future proof 😛)
PSU power cable changed from old Seasonic braided -> standard cable that came with SF750

From what I can tell my power requirements are well within the 750W limit of my PSU, plus I read somewhere on this forum that it can go to 900+W before giving up, so it shouldn't be overloaded. Also I can report that the fan doesn't spin beyond power-up when it does so briefly. Nothing else attached to the same trailing power adapter sees any problems, and it only has a few other things plugged into it (nothing major).

I've never seen anything in the event logs etc. other than "the previous shutdown at dd/mm/yyyy hh🇲🇲ss was unexpected" and this usually happens under minimal load (only firefox with <10 tabs open and simple browsing). I've never noticed a pattern in terms of when it happens. I have HD Sentinal running as a service and it has a green tick for all of my drives - plus drive, CPU and GPU temps are all good. These power cut-outs worry me mainly because I have some Veracrypt containers that I use for email etc. and I've had a few warnings regarding corruption as they weren't safely dismounted (lost my email program profile more than once). I have backups, but it is still a pain.

Is it significant that the motherboard boots back up again straight afterwards? I don't want to try it but from experience I'd expect this not to happen if I pulled the power cord and plugged it back in?

One last thing - might seem odd having the mobo + CPU combo I do if I'm not overclocking, but I just wanted the best setup possible on Intel 4th Gen so I can stay on Windows 8.1 until it dies (don't want to get into a debate about Win10, I'm just not a big fan of it). Got the Maximus Gene because I have a micro-ATX case (Silverstone TJ08-E) and I wanted a decent amount of SATA ports. But I did have stability problems at first, and these very forums helped me realise that the Maximus Gene overclocks by default (I think?) and I solved my problems by reducing my voltages (as per https://rog.asus.com/forum/showthread.php?51087-Maximus-VII-Hero-4790K-Temps-and-Voltages-Help). They were different problems too - they were as if someone pressed the reset switch on the board/case, as opposed to a complete cut of power. But I realise it could still be board related.

Any thoughts or advice would be much appreciated!

 

[slurmsmckenzie]

I'm not saying that the SF 750 is a bad psu, just that the motherboard has issues with voltages. The SF will probably work fine for years in a different system.

In some ways I would be happy if it was a motherboard issue, I'd finally do my upgrade and move on. It is a real pity that there isn't any kind of event log etc. within the Asus UEFI BIOS, that would be really helpful. I think I've seen that this is sometimes a feature on "corporate" workstation boards but it isn't on mine anyway.

I bought the Maximus Gene VII two years ago from eBay and paid more than I should for it, I wanted it to see me through until 2023 and give me one of the best platforms for my 4790K during that time. But it has had these random power issues ever since I got it, and I'm getting a bit miffed with it now.

A lot hinges around the TJ08-E TBH - I really like the case and would like to keep it, but the PSU limitation (given that I also want an ODD) is a bummer. And it is pretty cramped in general, although once it is together and working (well, mostly) then that doesn't matter anyway. I really wish that Silverstone would release a "long wheel-base" version of the TJ08-E that is just an inch or two longer with some updated connectivity, but it will never happen.

EDIT - having said that, if the SF750 is actually fine then I could keep the case but I'd need another micro-ATX board and they don't seem quite as fully-featured these days from what I can see (my Gene VII has got 8 SATA ports).

 

[--SID--]

SFX PSU in a mATX case, Do you use extension with the PSU?

 

[slurmsmckenzie]

SFX PSU in a mATX case, Do you use extension with the PSU?

I checked and Corsair say that all of their "type 4" cables are the same and compatible between qualifying PSUs, so I used the longer cables from my RM750i which fit perfectly and seemed to work fine (although maybe this is the issue).

https://www.corsair.com/uk/en/psu-cable-compatibility

If this is a/the problem then I probably will abandon the TJ08-E (sadly) because even with a compact ATX PSU it is too tight (for my liking at least).

 

[Vic 40]

32GB Crucial RAM (not OC'd)

16GB Crucual RAM -> 32GB

Have been focussing on the psu, but you could look at this too, test with only one kit of ram. I'm assuming you added ram.

 

[slurmsmckenzie]

Have been focussing on the psu, but you could look at this too, test with only one kit of ram. I'm assuming you added ram.

Just to clarify with that:
Feb 2019 to Jan 2021 - 16GB (4x4) Crucial RAM - had the random reboot issue
Jan 2021 to present - 32GB (4x8) Crucial RAM - still had same reboot issue

Also, the 16GB (4x4) was used from 2013 to 2019 in an H87 PRO with no such reboot issues.

So I guess that made me think it was unlikely to be RAM, but maybe. Do you think I should try with only 2 DIMMs?

 

[helper800]

Just to clarify with that:
Feb 2019 to Jan 2021 - 16GB (4x4) Crucial RAM - had the random reboot issue
Jan 2021 to present - 32GB (4x8) Crucial RAM - still had same reboot issue

Also, the 16GB (4x4) was used from 2013 to 2019 in an H87 PRO with no such reboot issues.

So I guess that made me think it was unlikely to be RAM, but maybe. Do you think I should try with only 2 DIMMs?

Worth a shot, and easy enough to test.

 

[slurmsmckenzie]

Worth a shot, and easy enough to test.

Yeah, should be fairly easy - with the somewhat cramped TJ08 it does mean removing the HDD cage, but simpler than trying a different PSU TBH!

If it was RAM, would it be expected to power off for a few seconds and back on again the way it does?

 

[Vic 40]

See you use four sticks of same kit. could increase voltage for the ram, like 1.65V (=max) , see if that helps as well. Can also add some voltage to the Memory Controller, like 0.05V and add, to locate the MC voltage, look for something like CPUVTT, DDRVTT, QPI/VTT or VCCIO in the bios. Test with two sticks to take load of the memory controller.

Still think it's a power issue, but it's worth a try.

 

[helper800]

See you use four sticks of same kit. could increase voltage for the ram, like 1.65V (=max) , see if that helps as well. Can also add some voltage to the Memory Controller, like 0.05V and add, to locate the MC voltage, look for something like CPUVTT, DDRVTT, QPI/VTT or VCCIO in the bios. Test with two sticks to take load of the memory controller.

Still think it's a power issue, but it's worth a try.

Even for a limited duration isn't 1.65 a bit much?

 

[Vic 40]

Not with ddr3.

 

[slurmsmckenzie]

See you use four sticks of same kit. could increase voltage for the ram, like 1.65V (=max) , see if that helps as well. Can also add some voltage to the Memory Controller, like 0.05V and add, to locate the MC voltage, look for something like CPUVTT, DDRVTT, QPI/VTT or VCCIO in the bios. Test with two sticks to take load of the memory controller.

I found these settings in the manual, do they sound like they are in the right area?

Eventual DRAM Voltage [Auto]
Allows you to set the DRAM voltage to be reapplied before entering OS. Values depend on the CPU installed.

VTTDDR Voltage [Auto]
Allows you to set the termination voltage for the system memory. Values range from 0.6000V to 1.0000V.

 

[helper800]

Not with ddr3.

Oh that's right, DDR3 ram have higher voltages. Its been a bit since I talked DDR3 RAM.

 

[slurmsmckenzie]

Okay, so I removed DIMMs 2 & 4 and then set the DRAM voltage to be 1.6V (it was previously ~1.5V) and I upped the VTTDR from 0.750V to 0.800V.

I also disabled the setting "CPU Integrated VR Fault Management (Allows you to prevent tripping Fully Integrated Voltage Regulator when doing over-voltage)" just in case, although I'm actually manually reducing it from whatever "auto" sets down to 1.2V. I did this after reading a few threads about this board and K processors, particularly this one: https://rog.asus.com/forum/showthread.php?51087-Maximus-VII-Hero-4790K-Temps-and-Voltages-Help

I was getting BSODs regularly until I reduced the voltage to 1.2V and after that I can only ever remember getting one more BSOD in over two years (caused by enabling VT-d, so I disabled it again). But TBH I don't really know what impact setting the CPU voltage to be 1.2V might be having, it just seemed to stabilize things so I was happy.

I also checked and re-made PSU connections whilst I was in the case, they all looked fine.

One other thing - I had forgotten that I am using a Cooler Master 90degree adapter (with ripple capacitors) for the ATX power connector so that it doesn't foul the HDD cage (only ever used it on this board as the connector is in a different place) - https://www.amazon.co.uk/gp/product/B07H9KPKG8/

And also I'm using a 1->4 SATA adaptor for 3xWD REDs and 1x950 EVO SSD - https://www.amazon.co.uk/gp/product/B00ASGC7K2/

Not sure if either of those look suspect? Had completely forgotten about the ATX 90degree adaptor, maybe I should try removing it for a bit (didn't want to change two things at once this time though with the RAM stuff).

 

[Vic 40]

I found these settings in the manual, do they sound like they are in the right area?

VTTDDR Voltage [Auto]
Allows you to set the termination voltage for the system memory. Values range from 0.6000V to 1.0000V.

This should be the one, up that by 0.005V and see if it help with the memory controller. Might do that 2/3 more times if it doesn't help, so don't overdo it.

Not too worried about the sata adapter, but the Cooler Master 90 degree adapter could maybe add to problems so if you can do without would that be good.

 

[slurmsmckenzie]

Can also add some voltage to the Memory Controller, like 0.05V and add

up that by 0.005V and see if it help with the memory controller

Erm, is it 0.05 or 0.005?? Should I go back in and reduce it? I went from 0.750 to 0.800 (0.05 increase) 😳

EDIT - just in case I went back in and changed it to 0.75625 which was the next increment above 0.750, let me know if I need to put it back up.

 

[Vic 40]

One increment per time is fine. Meant 0.05V btw.

 

[slurmsmckenzie]

One increment per time is fine. Meant 0.05V btw.

Ah great, thanks 🙂 Will put it back up to 0.800 in that case.

 

[slurmsmckenzie]

See you use four sticks of same kit. could increase voltage for the ram, like 1.65V (=max) , see if that helps as well. Can also add some voltage to the Memory Controller, like 0.05V and add, to locate the MC voltage, look for something like CPUVTT, DDRVTT, QPI/VTT or VCCIO in the bios. Test with two sticks to take load of the memory controller.

Still think it's a power issue, but it's worth a try.

Just to update, since making the RAM voltage changes I seem to have been stable 😀
I've been able to have all 4 sticks installed when I set the DRAM voltage to from ~1.5V to 1.6V and upped the VTTDR from 0.750V to 0.800V as per advice from @Vic 40.

Had the computer running for 30 days straight, doing all the normal stuff I'd normally do and I've seen no shutdowns. This is definitely an improvement!

So for me and my system at least, it seems it was an issue with the motherboard configuration and seemingly not the PSU. I guess I'd probably tend to stay away from the top-end enthusiast boards in future for that reason - I had to manually reduce CPU voltages to get it stable from the start and looks like it needed RAM voltages tweaking to keep it stable too. I've read elsewhere that the ROG enthusiast boards can sometimes overclock a K processor by default without any manual changes, and for someone like me to is generally happy at stock and prefers stability I'd probably be best with a more standard board. The reason I went for the Maximus was mainly because it was the most fully featured for a mATX board (most SATA ports, etc.) on Z97 and it is a shame that more features can at times seem to be tied in with default settings which may cause instability "out of the box". But it is a useful lesson to have learned!

Thanks again for the help all!

 

[Vic 40]

Good to hear it works. 😁