Is my PSU causing crashes/restarts? Broken, or under-provisioned?
I've been experiencing instability with my system, and I'm beginning to suspect that the PSU is either faulty or under-provisioned. I'll get random crashes (reboots, more specifically), mostly when doing something intensive on the system like starting up a game. Looking at all the crash logs, I can't really find any errors that make sense to me. When the system goes down, all the lights & fans all die at the same time as the screen, then after a couple of seconds it comes back on and reboots.
I have a Corsair SF600 SFX PSU which is only 600W, and I'm powering a Ryzen 3700x, a AMD 5700XT GPU, a 1TB M.2 SSD, 32G ddr4 memory, 2 case fans, and a water-cooling pump. Plugging all of that into a calculator says that 600W is exactly enough, but is that right? Or could power-usage spikes be pushing things over the edge?
Try limiting the speeds of your CPU then the GPU and see if the crashing still happens. If you drop the CPU speed and it goes away then something is with your CPU (or memory). If dropping just the GPU stops it then it’s probably the GPU. If you have to drop both and then it’s stable, or dropping either CPU or GPU then it could be the power supply.
Interesting, it's a bit hard to test as the crashes are super un-reproducable. (I don't think I've ever been able to produce a crash with a synthetic benchmark or the like). Any recommendations for testing? Or do I just have to run the system for a while in each state and keep track myself?
It could be the spike in loads that’s causing problems. I tried googling for transient spikes since Nvidia is notorious for it but AMD is better. But they pointed out this.
Also SFX PSUs tend to not have as much capacitance on the output side which could be part of the problem. A higher wattage PSU might not fix the problem, but it could help. But a higher quality one should definitely fix it assuming it’s a PSU problem.
Try running some synthetic tests first for a while and see if it’s stable. I think my PSU is dying from the spikes in load from my 4090. But that’s a 4090.
That would probably cover it but I'd personally go a little higher if I had the budget/space to.
You'll never have too much power and who knows what the power draw of your next upgrade might be? 800W would be better and 1000W would be enough for any PC you're likely to build. Anything higher is likely a waste of cash.
Unless you have an unusually small case (or buy the beefiest PSU), anything should be a drop in replacement. If it's a modular PSU and you just want to plug the same cables in, just be careful they're definitely the same cables. I think Corsair has a cable compatibility chart somewhere on their site.
To my understanding you want some headroom for your PSU for spikes of power draw or moments like when large appliances start up. Ive certainly had psu's as the cause of failures like that so it seems like it could be the case. Maybe you could get a 750w with return policy and try it out. Worse case you can just return it within that period!
Yeah seems like a good move, Corsair makes a 750W version of this PSU that's comparable with the same cables - so I could swap without even having to re-do my cable runs! Seems like it's worth a try. Hopefully I can get it somewhere local though, as I hate returning stuff in the mail. Thanks!
RAM instability can cause exactly what you're describing, especially at load. Run some tests to see if that's the culprit.
How old is your PSU? 600W should be plenty, since your CPU has a power limit of 88W, and your GPU averages 185W (worst spike I read about was 202W). Those two parts are pulling ≈300W at worst, and they're the bulk of your energy consumption.
An old PSU, though, will have failing capacitors and resistors, and it may not function correctly. It's usually recommended to replace them eventually, though I don't know at what frequency.
ETA: is your PSU getting enough airflow? Could it be overheating?
It's about 4 years old, so not new but not crazy old. I'm also in a SFF case so I'm using a PCIE riser, which could also be the issue. Do you know, is there a way to tell if it's the riser or the PSU? I can run memtest to check the memory.
I imagine it's getting enough airflow. It's a sff case, but the fans blow right over the PSU so it's probably fine... Would there be a way to tell if it's overheating?
I also have an SFF build, but as long as it's getting cool airflow, should be fine. Only way to test temps is to get a laser thermometer. PSUs unfortunately lack much in the way of readable sensors.
I have heard riser cables sometimes causing problems. Only way I know to test is try a different one or pull everything out, plug the GPU directly into the slot, and run some bench tests that way.
And speaking of temps, what about your RAM? Is it getting enough airflow? As heat goes up, RAM stability goes down, so make sure it's getting at least a little airflow. I'm paranoid, so I have a downdraft CPU cooler and exhaust fans right above that I know are giving mine some air.
It's a water-cooled SFF build so thermals can be an issue. Maybe somewhere something is getting too hot, it's a decent theory. I'll try building up a bunch of heat in the case to see if that increases instability. Thanks!
Doesn't appear to be psu, that model is pretty reliable and I've never had issues with it. And especially with your specs, that should be well within the power profile of the psu.
There's a bunch of other strange things that can cause it. I would first check if everything is seated properly, especially if you use riser cables or extension cables. I've had a build that would have the display go black randomly, and the fans would start running at max speed, and the only way to return it to normal would be a full reboot. Turns out the riser cable was a bit loose and the problem goes away if you push the gpu firmly into the riser cable's pcie port.
If that doesn't work, I would also try reseating the RAM or CPU. I had another build that was prone to crashes at random, no idea why. Did all the regular troubleshooting things. I suspected it was a CPU issue, so I pulled out the CPU and plugged in my old CPU to test. When I put my new CPU back in, there were no issues anymore. I have no idea what I did differently, so it must have been that the CPU wasn't seated correctly, or something like that.
Anyways, connections would be what I would check first. Pull out the pieces and then reseat them one by one and hope the problem goes away. If problem still persists, I would suspect it's probably the riser cable. This might be a bit difficult with an SFF case, but try to mount the GPU directly onto the mobo, and see if the problem goes away. As I mentioned above, it might just be that the riser cable is a bit loose (so you might still be able to use it as long as you seat the GPU firmly), but it's easier to troubleshoot if we can confirm that the riser cable is the problem
Interesting... I've used 2 different CPUs in this system over the years and the crashing has been across both - so it's probably not that. I think I've also swapped ram at some point, but that's easy so I'll just re-seat it to bu sure. I don't think I've ever touched the GPU so maybe re-seating it & the riser would help. Worth a try, at least. Thanks!
I've heard this in passing and haven't really been able to verify myself, but I gather that SFF PSUs are less resilient to transient power spikes (I.e. from GPUs), which can manifest as system instability under load.
There are several PSU tier lists floating around online (which I have also been meaning to consult for my own systems).
I've never heard this, but OP specified an SFX PSU, not just an SFF PSU. I can see that weakness being a thing for Flex PSUs, since they're often designed to fit in server racks or other tight spaces, but in all the time I've spent in SFF communities, I've never heard people complain about inferiority to ATX PSUs.
Would be curious if anyone knows anything definitive about this.