PC Shuts Down Abrubtly

[Larry4D]

Hi there,

 

I'm having issues with a new build that've I've put together myself. The machine will abruptly power off when using Cinema 4D and Octane. It completely powers off and all that's left is some lights on the motherboard. It only seems to happen when I'm rendering (which uses both GPUs).

 

Here are some of the steps I've taken:

 

• Checked event viewer but the only log I can find is a critical error "The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly."
• Checked temperatures (everything is watercooled, liquid when rendering doesn't exceed 45c when rendering, each component temperature is fine (both gpus are 53-54c, cpu is about 57c when rendering).
• BIOS and everything is up to date, latest drivers etc.

 

Here are the system specs (nothing is overclocked):

 

• Ryzen 5900x 
• 64GB of DDr4 3600 Ballistix Ram (4 sticks of 16)
• Dual 3090 FE in the EK Quantum Vector Waterblock
• Gigabyte Aorus Master
• All the water cooling components are EK (2 rads, d5 pump etc)
• 1200W ROG Asus Power Supply
• 3x M.2 Samsung 980 Pros (1tb each)
• 2x Samsung 870 SSDs
• 6 Corsair Fans with RGB
• The Corsair Commander Pro for temperature control + Lighting node
• Windows 10

My first instinct is the power supply - I was kinda worried it wasn't enough power (I originally tried to get the 1600 Corsair one but it was out of stock) but everything seems to be within the limits (both gpus at 350W, Ryzen at 220W at full load still leaves 280W spare for everything else?). I've run some tests where I've been rendering and ran Prime95 at the same time to get everything drawing as much as possible but it doesn't shut down. But I can be using C4D and maybe moving the camera which doesn't really use a lot of CPU and the machine turns off. It's happened about 4-5 times now. It doesn't happen when gaming (games only tend to use one GPU).

 

I'm at a bit of a loss to how to troubleshoot this one. Maybe the PSU is overheating? I don't think I've ever heard it's fan turned on (it does occasionally make a noise as it whirs up for half a second). The switch for silent mode is disabled. I'd hate to RMA it because this is my work PC and I can't find any others in stock at the moment.

 

Any help would be appreciated, thank you. 

 

(ps i've attached a photo incase theres anything glaringly obvious I've missed (it shuts down with the case off too))

[HanZie82]

Pretty sure the 3090 have insane peak power requirements, but you should still be good.
Do you have the option of changing PSU's to test another one? That would be the easiest way to rule that one out.

 

(Also please do not use black text.

Some of us use the dark theme and then black text becomes unreadable almost.)

[Larry4D]

34 minutes ago, HanZie82 said:

Pretty sure the 3090 have insane peak power requirements, but you should still be good.
Do you have the option of changing PSU's to test another one? That would be the easiest way to rule that one out.

 

(Also please do not use black text.

Some of us use the dark theme and then black text becomes unreadable almost.)

Unfortunately I don't have access to another power supply to run both cards - I could get the beQuiet 1500W but that's £450~. Worse case I do have a 1000W in another build but would have to disconnect one of the GPUs but that might change the outcome of the test.

 

Yeah I figured even with some peaks, there's still some headroom with a 1200W. 

 

(also sorry, fixing the black text now)

 

39 minutes ago, James Evens said:

Try lowering the RAM speed to something like 2100 MHz (just for troubleshooting). If this doesn't help underclock the GPU.

 

I already had XMP disabled for a little while and this still happened  

 

[Mark Kaine]

has already been said probably psu issue.

 

Sure 1200w should probably be enough, but some psus cant handle the spikes, and your symptoms fit 100%

 

 

you can try undervolting (by a lot) but long term you'll need a better psu that can handle the ampere cards…

 

 

[Larry4D]

4 minutes ago, Mark Kaine said:

has already been said probably psu issue.

 

Sure 1200w should probably be enough, but some psus cant handle the spikes, and your symptoms fit 100%

 

 

you can try undervolting (by a lot) but long term you'll need a better psu that can handle the ampere cards…

 

 

I think this makes sense, as when I trigger movements in C4D both cards would restart the render at the same time, and perhaps this causes a large spike.

 

I've ordered a socket wattage meter, would this be able to tell me if the PC was drawing too much power?

[Mark Kaine]

53 minutes ago, Larry4D said:

I think this makes sense, as when I trigger movements in C4D both cards would restart the render at the same time, and perhaps this causes a large spike.

Yeah, something like that could trigger it, but its not entirely clear what really triggers it, i mean for most people benchmarks seem fine but then they get crashes during games…

 

53 minutes ago, Larry4D said:

I've ordered a socket wattage meter, would this be able to tell me if the PC was drawing too much power?

Not sure, depends on how sensitive it is, those spikes are very quick…

 

I think actually donwclocking the cards with something like afterburner would have better chances to "test" the spiking.

 

core clock -50

power limit -50 or so

 

 

That should most likely stop the crashing …

 

 

Also in case its not clear, this issue isnt because the psu cant provide enough power, its that it cant handle transient spikes well and therefore shuts down under certain circumstances (dont know the exact details, but you can look up "transient power spikes" that should explain some things)

[Larry4D]

3 minutes ago, Mark Kaine said:

Yeah, something like that could trigger it, but its not entirely clear what really triggers it, i mean for most people benchmarks seem fine but then they get crashes during games…

 

Not sure, depends on how sensitive it is, those spikes are very quick…

 

I think actually donwclocking the cards with something like afterburner would have better chances to "test" the spiking.

 

core clock -50

power limit -50 or so

 

 

That should most likely stop the crashing …

 

 

Also in case its not clear, this issue isnt because the psu cant provide enough power, its that it cant handle transient spikes well and therefore shuts down under certain circumstances (dont know the exact details, but you can look up "transient power spikes" that should explain some things)

Ah thank you, that makes sense. I will try this and let you know how it goes! I appreciate the help from everyone too.