Computer randomnly shuts down with Dtech "Full" power supply

[SymS]

I recently had to change my power supply, as the capacitor blew up during an heavy task (building Android OS). It's actually the second I blow that way in 2 years, so I asked the seller if I could get a better supply, and they proposed

DTech PW045 "Full" power supply (MAX:550W). Link: http://thesystem.co.th/products/previews.php?code=A0032787

It's working OK most of the time, but maybe once or twice a week my computer will just turn off automatically (and not cleanly, just like a power failure). There's very little CPU at the time of the issue, the temperature is rather low (I'm now in Chiang Mai, the room at 20 C), and I'm behind a Syndome UPS. If I do some heavy duty tasks using all eight cores at 100%, such as video transcoding or building large software stacks, the power supply copes just fine.

My specs are as follows: AMD FX8350 processor, 16 GB RAM, 1TB SATA hard drive + 1 Seagate SSD, GeForce 210 GPU. I'm running Ubuntu 14.04, but I guess the OS does not matter.

This issue never occurred with my previous "cheap" power supply. Any idea what may go wrong here?

I can't really return it to the shop since I've had it for 2 months.

Edited December 23, 2014 by SymS

[ukrules]

It sounds more like a memory problem to me, these can cause seemingly random crashes at any point in time regardless of whether it's doing much or not.

I would download and boot into the memtest memory checker from CD and run some scans overnight - this one : http://www.memtest.org/

I once had a computer which would crash every few days for no apparent reason, there was nothing wrong with the memory - the motherboard didn't like 4 x 8 GB DIMMS, works perfectly with 4x4GB's. I swapped the RAM to a different computer and I've never had a random crash since then.

BTW : 550w is a pretty small power supply, is it definitely good enough to easily power your system ? I tend to over spec items like this just to make sure.

[SymS]

Thanks. I'm running memtest now. I would think a memory problem would crash the system, and not make it power off, but I have seen others suggest the same thing. We will see.

The power supply should be large enough, I measure around 400 watts during heavy lifting, and that includes my two monitors.

[phazey]

Memory or Thermal shutdown, but as above, i'd put money on memory

[thedemon]

Issues with the GPU can also cause sudden shutdown. Maybe worth checking that the GPU fan is spinning freely. Also pull out the card and put it back making sure it is seated securely.

[bkkgooner]

550w seems very low for the heavy duty stuff that you are doing and why are you taking the advice of a Thai salesman?

Go have a search around on forums and get better advice, don't be cheap on the PSU as it is essential

[RKASA]

It is like a loop so many things and they look and act the same. A bad cap on the main board. A ram chip. The GPU. All you can really do is substitute parts until you find it. One time it was a main board for me another time I just had to re paste the fan on the GPU board and yet another time it was paste on the CPU dried out. Cross fingers and try the cheapest stuff first.

[WhizBang]

It's a cheap power supply. You paid 750 baht. Your problem could be the power supply itself. Do you have any other power supply you could swap out to test?

I just built a new computer. I paid around 12000 baht for my PS.

You are getting random shutdowns. If the PS is the source, swapping it out will solve the problem. If the problem persists, the power supply is not the problem.

Edited December 24, 2014 by WhizBang

[phazey]

Issues with the GPU can also cause sudden shutdown. Maybe worth checking that the GPU fan is spinning freely. Also pull out the card and put it back making sure it is seated securely.

True, but you would see graphic corruption if the GPU was on the way out.

[RKASA]

Issues with the GPU can also cause sudden shutdown. Maybe worth checking that the GPU fan is spinning freely. Also pull out the card and put it back making sure it is seated securely.

True, but you would see graphic corruption if the GPU was on the way out.

No not always - if it is tripping it can just shut down the PC like mine did, but it was just a matter of pulling the fan unit off it and cleaning good then re pasting it. I am talking the video card here - same thing can happen and has with the CPU. Yes you might see corruption depending on how it fails, but... It really is a matter of step by step substitution of parts without assumptions.

If you have 4 ram cards try it with one removed then the next and put the other back - if they all did not stop it from happening when one was removed then move on to using the on board video and pull the vid card. If that is it don't assume the card is bad. Try resetting the cooling fan or the heat sink if it does not use a fan.

Most power supplies are pretty bullet proof but if you have a spare try that - one of them is likely to be OK. The main board is the bugger because short of proving all your other parts in a different PC or finding an obvious blown cap on the board it is hard to sub a main board if you don't have one.

edit don't forget to re paste the CPU - it is not that hard to do and basic white paste is more then good enough - the silver BS is sales hype - yes it works better but the difference is like 29mpg as compared to 29.0003mpg so what.

edit one more thing the clear alcohol cleaner at the tech shop for 100 baht you can get at tesco in the beauty supplies for 9 baht - it is nothing more the clear finger nail polish remover with a tech label on the bottle.

Edited December 24, 2014 by RKASA

[Xircal]

There's not much is the way of specs in your link so it's difficult to say what the output on the 12V rail might be. But if it exceeds 18A, then the PSU would shutdown before reaching 20A in the same way a circuit breaker or a fuse would. This is primarily to prevent a fire due to the wire overheating.

With the components you have in the machine already, I'd recommend that you don't skimp on the PSU and go for a well known brand like Enermax, Corsair or similar.

[SymS]

Thanks for all your answers. The memory test succeeded, but memtest86 apparently runs the memory clock at 640MHz, instead of the 1600 MHz set in the BIOS.

I have selected an "Open case" option in the BIOS, that I understand reset settings in case you may changes to the hardware, and it worked for over a week, but I had the problem happen to me again.

I did not have this issue with my earlier power supplies (even cheaper), so I guess the issue must really be with the power supply.