Duff PSU?

[silkstone]

Hi all, So in my system I have this PSU which is rated at 700W, but whenever I put a heavy load on it, I'm getting a 12 V rail reading of 11.3V and system instability (crashes)

I'm running a 8600K (stock) + 1080 + 1080Ti (both at 80% TDP) but when I stress all 3 together, my PC crashes in 10-30 minutes.

From my rough calculations, the whole system should only require 600-650 W at max (not counting the reduced power settings I've put on the video cards)

1080 = 186 W
1080 Ti = 267 W
Rest of the system = 150 W
Total = 600 W

I got this PSU as it received great reviews of Johnny Guru and being Titanium, I thought it'd handle 100% load well, but that doesn't seem to be the case

I only recently upgraded my CPU and motherboard and I wasn't getting this problem before with a heavily overclocked 2500K, so I am a little suspicious that it might also have something to do with the motherboard.
It only happens when I put stress on the CPU, but the temps are fine. When I put the TDP on the VGA cards up to 100% it trips my UPS alarm which is rated at 660W and as adding load on the cpu doesn't trip the alarm, I am pretty certain that I am below 700 W.

Apart from testing my PSU with a digital multi-meter, which is my next step, does anyone have any thoughts or ideas?

 

[CAPSLOCKSTUCK]

I would invest in a watt meter before i went any further.and in the meantime live without the 1080 ( temporarily)

 

[_UV_]

Rest of the system = 150 W

IMO just CPU alone then overclocked consumes that, and rest of the system 100+ above that
also 1080ti (depends on model) actualy can eat up to 400 (according to some reviews)

 

[silkstone]

I would invest in a watt meter before i went any further.and in the meantime live without the 1080 ( temporarily)

I've been looking, but unable to find them. I might have to get one imported. I was able to measure my old i5 2500K + 1080 at the breaker which was drawing about 2A on load (100% TDP). At 240 V that is equal to ~ 500 W

IMO just CPU alone then overclocked consumes that, and rest of the system 100+ above that
also 1080ti (depends on model) actualy can eat up to 400 (according to some reviews)

The API for the miner I am using puts both at 350 W, but I am pretty sure that is just taken from the TDP of each card.

the other reviews put the total system load at ~400 W. That should still leave me with 300 W for the second card

 

[CAPSLOCKSTUCK]

I've been looking, but unable to find them.

you should be able to find one no matter where you live

9 quid in the UK

 

[Bill_Bright]

I'm getting a 12 V rail reading of 11.3V and system instability (crashes)

What are you measuring that with?

Note according to the ATX Form Factor Standard, a ±5% tolerance spread is allowed. That means the ranges following are acceptable:

+12V1DC ±5% = +11.40 to +12.60 V
+5VDC ±5% = +4.75 to +5.25 V
+3.3VDC ±5% = +3.14 to +3.47 V​

As you can see, your +11.30V is too low. If accurate, that is likely your problem, or a major contributor to it.

From my rough calculations, the whole system should only require 600-650 W at max

Because you are running two hungry graphics cards, you need to be more accurate than "rough". I recommend you use the eXtreme PSU calculator to determine your needs. When I plug in your specs, plus a few guesses on my own, as you can see here, at maximum load you are pushing real close to your power supply's capacity.

So combined with your out-of-tolerance voltage, and the heavy load with both hungry graphics cards, you need a new supply.

Note that all PSU calculators pad the results a little to avoid recommending an underpowered supply. But, by far, the eXtreme Outer Vision PSU calculator is the most accurate because it is so flexible with so many input and utilization options. Plus (and this may be the most important factor) they have a group of researchers on staff constantly researching components for us.

 

[silkstone]

you should be able to find one no matter where you live

9 quid in the UK

View attachment 94642

I've found and ordered one. I should have it by the end of the week.

What are you measuring that with?

Note according to the ATX Form Factor Standard, a ±5% tolerance spread is allowed. That means the ranges following are acceptable:

+12V1DC ±5% = +11.40 to +12.60 V​

+5VDC ±5% = +4.75 to +5.25 V​

+3.3VDC ±5% = +3.14 to +3.47 V​

As you can see, your +11.30V is too low. If accurate, that is likely your problem, or a major contributor to it.


Because you are running two hungry graphics cards, you need to be more accurate than "rough". I recommend you use the eXtreme PSU calculator to determine your needs. When I plug in your specs, plus a few guesses on my own, as you can see here, at maximum load you are pushing real close to your power supply's capacity.

So combined with your out-of-tolerance voltage, and the heavy load with both hungry graphics cards, you need a new supply.

Note that all PSU calculators pad the results a little to avoid recommending an underpowered supply. But, by far, the eXtreme Outer Vision PSU calculator is the most accurate because it is so flexible with so many input and utilization options. Plus (and this may be the most important factor) they have a group of researchers on staff constantly researching components for us.

It's measured via on-board sensors. I know they aren't accurate which is why I'll put a multi-meter to the rail tomorrow. However, it does correctly read 12 V when at idle.

Thanks. The PSU calculator recommends a 724 W PSU for my system, with a 675 W max usage. Add to the fact that I'm only running the cards at 75-80% TDP, it leads me to believe that my PSU isn't able to deliver what it is rated for. I'll hold off final judgement until I get a power meter later this week.

 

[Bill_Bright]

which is why I'll put a multi-meter to the rail tomorrow. However, it does correctly read 12 V when at idle.

A multimeter is not conclusive. For one, it is difficult to measure across a full range of loads with a multimeter. Plus, you cannot measure for proper ripple suppression with most multimeters.

A power (watt) meter is not really going to help you much here either. This is because the power meter will tell you the total power the power supply is pulling from the wall. You would have to know the exact efficiency rating of your power supply at that specific load level to determine how much power your computer is demanding.

For example, if your computer (CPU, motherboard, cards, RAM, drives, fans) is currently demanding 500W from your power supply, and your power supply is 90% efficient @ 500W loads, the power supply will be pulling from the wall ~556W since 556 x .9 = 500.4 watts.

While the efficiency rating of your quality PSU is relatively flat, it still varies almost 5 points across all load levels ASSUMING it is working properly. But with your +12V rail apparently being low, it does not seem to be working properly. So we cannot really assume it is meeting its Titanium efficiency ratings either.

BTW, I recommend all computers be on a "good" UPS with AVR. This not for backup power during an outage (though that is a nice bonus feature). It is the AVR (automatic voltage regulation) that makes a "good" UPS with AVR so important. Another bonus is most better UPS have an LCD status display panel (and software) that can tell you exactly the power the connected devices are demanding without having to buy a separate power meter.

 

[silkstone]

A multimeter is not conclusive. For one, it is difficult to measure across a full range of loads with a multimeter. Plus, you cannot measure for proper ripple suppression with most multimeters.

A power (watt) meter is not really going to help you much here either. This is because the power meter will tell you the total power the power supply is pulling from the wall. You would have to know the exact efficiency rating of your power supply at that specific load level to determine how much power your computer is demanding.

For example, if your computer (CPU, motherboard, cards, RAM, drives, fans) is currently demanding 500W from your power supply, and your power supply is 90% efficient @ 500W loads, the power supply will be pulling from the wall ~556W since 556 x .9 = 500.4 watts.

While the efficiency rating of your quality PSU is relatively flat, it still varies almost 5 points across all load levels ASSUMING it is working properly. But with your +12V rail apparently being low, it does not seem to be working properly. So we cannot really assume it is meeting its Titanium efficiency ratings either.

BTW, I recommend all computers be on a "good" UPS with AVR. This not for backup power during an outage (though that is a nice bonus feature). It is the AVR (automatic voltage regulation) that makes a "good" UPS with AVR so important. Another bonus is most better UPS have an LCD status display panel (and software) that can tell you exactly the power the connected devices are demanding without having to buy a separate power meter.

I had similar thoughts. I can go by Johnny Guru's review for the performance at certain loads and work backwards, but If the meter reads below 700 W from the wall, it would be safe to assume the PSU is bad. Luckily it has a 5 year warranty.

I have an oscilloscope at work that would be used to better check the rails, but I don't fancy lugging it home and there aren't any windows 10 drivers for the digital ones we have

I have an APC UPS: http://www.apc.com/shop/id/en/products/APC-Back-UPS-1100VA-230V-AVR-IEC-Sockets/P-BX1100CI. Not great, but better than getting it from the mains.

 

[thebluebumblebee]

when I stress all 3 together, my PC crashes in 10-30 minutes.

IMHO, this is a heat issue, not a capacity issue. Something is getting warm, which is causing more resistance, which causes the crash. I'd check the 24 pin connector to see how warm it's getting under load. JG did have an issue of an incorrectly assembled PCI-e connector, so maybe you have something like that as well. Suggest a thorough inspection of the cables.

Edit: I know that this PSU tested well, but I have to wonder if Andyson actually thought that someone would use it the way you are. If you compare the cabling of this PSU with that of, say, an EVGA SuperNOVA 750 G3, there is a huge difference for such a small capacity difference. If you showed me a picture of the back of the Andyson PSU without telling me that it had a hardwired PCI-e cable, I would have guessed that it was a sub 500 watt PSU. (actually, the cabling for this PSU is essentially identical to EVGA's (SuperFlower) SuperNOVA 550 G3)
Andyson N700 Titanium:

EVGA SuperNOVA 750 G3:

 

[Bill_Bright]

IMHO, this is a heat issue, not a capacity issue.

But one can lead to the other. When you stress a PSU, it can over heat and it becomes unstable. This further increases heat which makes it even more unstable.

But note too, if you do the math, he is using two power hungry graphics cards. When they are pushed and demanding the most power, they are indeed pushing the demands to near capacity of the supply.

Couple that with the low voltage output (which can, again, result in more heat as regulator circuits try to compensate), and you have a further unstable situation.

@silkstone Did you have these problems from the start? Of was it working fine then started to become unstable?

 

[silkstone]

I only had the problem since upgrading to a 8600K last weekend. Before I was on a 2500K at 4.4Ghz. I did occasionally put full load on 2x GPU and 1xCPU, but not 24/7. I certainly didn't notice crashes, just system slowdown (hence the upgrade)

The PSU is cool to the touch and the connectors don't feel hot either. It's not just the 12V rail that is reportedly under voltage though. Below are all my improtant voltages

12 V @ 11.376
5 V @ 4.89
3.3 V @ 3.127

On Idle, they are reported as:

12 V @ 12.024
5 V @ 4.98 V
3.3 V @ 3.206

Again, that's just with the motherboard sensors, so the numbers need to be taken with a pinch of salt. The 3.3 V looks a little low, even at idle, I'll have to look into how I can check that one with a multi-meter too.

 

[Ebo]

What happens if you set the voltage manual to Jg's reading, is it still going to crash ?.

 

[Beastie]

Assuming the 11.3 volt measurement is accurate, this supply voltage is marginally out of spec. So the VRMs on your new mobo might not be able to cope- but not all mobos are equal, so maybe your old one could just cope.
This doesn't necessarily mean your new mobo is defective, if it isn't getting the right voltage supply it isn't its fault if it can't function.

Is the PSU in warranty? The voltage in the JonnyGuru review is 11.96 at full load, whereas it looks like yours is out of ATX spec, therefore not fit for purpose. It should definitely remain in spec up to the max rated power.

However IMO you could really do with a higher wattage PSU, you're pretty much maxing that unit out. Personally I think that 700W is cutting things a bit fine.

 

[Bill_Bright]

The 3.3 V looks a little low, even at idle.

3.206V at idle is okay, but I agree under load is a bit low (though still within acceptable tolerances).

While it would be a rare event when both graphics cards, all RAM sticks, motherboard, CPU, all drives, and all fans are all demanding maximum power at the same moment, the power supply must be capable of supporting that event when it does occur. So I agree, you need a bigger supply - or pull one graphics card and see what happens.

 

[Jetster]

What's the event log say?

 

[silkstone]

What's the event log say?

I know the cause for the crash, it gives the same driver issue each time, but it is caused by a lack of power.

 

[AsRock]

Try just using one card and see if the system runs ok or not and if it still happens you could try a fresh install of the OS and see if the problem goes away or not with a single card.

When i was dealing with Seasonic they did not like any thing lower than 11.35, but software is often wrong.

Maybe a PSU needs a cleaning out with a can of air.

 

[Jetster]

I know the cause for the crash, it gives the same driver issue each time, but it is caused by a lack of power.

Then RMA that PSU and see if it fixes the problem

 

[silkstone]

Try just using one card and see if the system runs ok or not and if it still happens you could try a fresh install of the OS and see if the problem goes away or not with a single card.

When i was dealing with Seasonic they did not like any thing lower than 11.35, but software is often wrong.

Maybe a PSU needs a cleaning out with a can of air.

I know that it'll likely work fine when using 1 card as I'll be well below the rated limit, but I will check it out and see how much the voltage drop is compared to the power draw. That'll give me some indication.

Then RMA that PSU and see if it fixes the problem

Yeah. I'm going to wait for the power meter to try to get an idea of the kind of load I am putting on it first. I don't have time to do an RMA until next weekend anyway, but the warranty is 5 years one-for-one, which is pretty good. The thing is I'll need to convince the service center that there's a problem as, unless something has blown up or is totally non functioning, they don't like to accept RMAs over here. The more evidence I get, the easier the process will be.

 

[SnakeDoctor]

Running the Pc direct power and bypassing the ups , does it still crash ? any difference in 12v rails
As you saying your ups trips thus it is too small for your system
Ups model and Brand?

as posts above show that a 1140 va ups is required for you system

 

[silkstone]

It's an APC 1100 VA UPS. It only trips when I put the TDP of the video cards up to 100%.
I can't imagine the UPS could cause this problem, but it's an easy thing to check so I'll take a look anyway.

 

[Bill_Bright]

It's an APC 1100 VA UPS. It only trips when I put the TDP of the video cards up to 100%.
I can't imagine the UPS could cause this problem

If the load on the UPS is bigger than it carry, then it is supposed to shut down! The UPS has to assume that you have done your homework and don't put a load on it that exceeds its capacity because an excessive load (current demand) could indicate a fault - a short somewhere in the connected devices.

Did you research the specs of your UPS? You did not state which model but note according to the specs for the APC Back-UPS 1100VA 230V, it has a capacity of 660 watts.

Understand, if not specified by the UPS maker, the general formula to convert VA (voltamps) to Watts is VA x PF = W where PF (power factor) is typically .6.

APC uses .6 for a power factor.

So, 1100 x .6 = 660W

So most likely, it is not that your UPS is bad (APC is my preferred brand) it is that "you" over loaded it!

 

[Jetster]

Usually when you think something can't be the problem, it is. Take the UPS out of the equation

 

[SnakeDoctor]

Also when you say Pc crashing are you talking about it rebooting or freezing or just the ups alarm activating /tripping /overloaded