My PC keeps turning on and off every 3 seconds. Is my motherboard faulty?

[Valantar]

the nominal/optimal load for PSU's are 50%, ie. where they are the most efficient.

Given that modern 80+ Gold or above designs vary in efficiency by a few percent between 20 and 100% load, it's hard to definitively say they have an "optimal" load at all. And I don't mean only high-end designs - look at something like the Bitfenix Formula Gold 750W, which TPU reviewed back in 2017, which should be reasonably representative of a middle-of-the-pack 80+ Gold design these days.

While efficiency drops precariously below 20% load (which matters little in terms of PSU wear or lifespan simply because the thermal output at those low loads are small anyhow, but still is worth improving - and is improving in high-end units thanks to 80+ Titanium finally implementing a 10% load efficiency requirement), it hits >89.5% at 20% load (150W), slightly above 92% at ~350W, and slowly dips down to 90% again at ~825W, or a 10% overload. To sum up, there's a maximum 3% variance in efficiency when your PC is not idling. Claiming that the 50% point somehow stands out is not really accurate on modern PSUs. A 2-3% variance in efficiency hardly matters at all - it's small enough that the heat output scales pretty much linearly with the load applied throughout this part of the load range.

Beyond that, the nominal load of an X watt PSU is supposed to be X. If it can't handle that output, it should neither be named nor rated to do so. And as seen in most good PSU reviews, modern PSU designs can generally handle higher loads than what their labels say, not lower (though this will likely lead to a shortened lifespan, it should not cause the unit to immediately break). Nobody should recommend a 100% PSU load for extended periods of time, but 80% when gaming is perfectly fine - particularly as you're very unlikely to be gaming 24/7. Even gaming PCs spend the majority of their in-use time idling or on the desktop.

 

[EarthDog]

Given that modern 80+ Gold or above designs vary in efficiency by 5-8% between 20 and 100% load, it's hard to definitively say they have an "optimal" load at all.

At most, 3% (for 120V, 4% for 230V). Otherwise it doesn't qualify for the tier rating.

 

[Valantar]

At most, 3%.

View attachment 137093

Heh, thanks, I just (don't know if it happened before or after your post tho) edited that part to say "a few percent" I couldn't be bothered to look up a table, so thanks anyhow!

 

[EarthDog]

Heh, thanks, I just (don't know if it happened before or after your post tho) edited that part to say "a few percent" I couldn't be bothered to look up a table, so thanks anyhow!

Your other points remain... just tightening it up a bit.

 

[Valantar]

We're veering a bit OT here, but nonetheless I mocked up a little graph to show just how misleading these cropped/non-zero Y axis efficiency graphs can be, with data approximated from the TPU graph above. Mostly meant as a response to @skizzo, but we can all need the reminder once in a while. The first is the type of graph you see in reviews:

The second is the same graph, just with 0% efficiency plotted in at 0 Watts to force Excel to not crop the Y axis.

That ... is essentially a flat line past 20% load.

While they make the relatively small differences between units easier to understand, these types of graphs nonetheless work to perpetuate the false belief that modern PSUs have dramatic changes in efficiency within their rated load range. Sure, 0-20% load is a bit of a free-for-all, but even there most 80+ Gold units exceed 80%. And beyond 20% load, the graph is essentially flat. It's time to stop this "50% load is what you should aim for" nonsense.

 

[skizzo]

Given that modern 80+ Gold or above designs vary in efficiency by a few percent between 20 and 100% load, it's hard to definitively say they have an "optimal" load at all. And I don't mean only high-end designs - look at something like the Bitfenix Formula Gold 750W, which TPU reviewed back in 2017, which should be reasonably representative of a middle-of-the-pack 80+ Gold design these days.

While efficiency drops precariously below 20% load (which matters little in terms of PSU wear or lifespan simply because the thermal output at those low loads are small anyhow, but still is worth improving - and is improving in high-end units thanks to 80+ Titanium finally implementing a 10% load efficiency requirement), it hits >89.5% at 20% load (150W), slightly above 92% at ~350W, and slowly dips down to 90% again at ~825W, or a 10% overload. To sum up, there's a maximum 3% variance in efficiency when your PC is not idling. Claiming that the 50% point somehow stands out is not really accurate on modern PSUs. A 2-3% variance in efficiency hardly matters at all - it's small enough that the heat output scales pretty much linearly with the load applied throughout this part of the load range.

Beyond that, the nominal load of an X watt PSU is supposed to be X. If it can't handle that output, it should neither be named nor rated to do so. And as seen in most good PSU reviews, modern PSU designs can generally handle higher loads than what their labels say, not lower (though this will likely lead to a shortened lifespan, it should not cause the unit to immediately break). Nobody should recommend a 100% PSU load for extended periods of time, but 80% when gaming is perfectly fine - particularly as you're very unlikely to be gaming 24/7. Even gaming PCs spend the majority of their in-use time idling or on the desktop.

I can't disagree with your response here, it's valid, but isn't mine also? Which seems to be supported by that graph. The absolute most efficient load on that PSU is between 300W and 400W which aligns with 50% of power rating which is around 375W on 750W rated PSU. So I think perhaps I used too much of an absolute statement the first time or referred to the wrong thing. I am referring to what I think is supported by that graph....the PSU is at its highest effiency at around a 50% load of what it is rated for. For example, a 1000W PSU would be most efficient around 500W load

We're veering a bit OT here, but nonetheless I mocked up a little graph to show just how misleading these cropped/non-zero Y axis efficiency graphs can be, with data approximated from the TPU graph above. Mostly meant as a response to @skizzo, but we can all need the reminder once in a while. The first is the type of graph you see in reviews:

The second is the same graph, just with 0% efficiency plotted in at 0 Watts to force Excel to not crop the Y axis.

That ... is essentially a flat line past 20% load.

While they make the relatively small differences between units easier to understand, these types of graphs nonetheless work to perpetuate the false belief that modern PSUs have dramatic changes in efficiency within their rated load range. Sure, 0-20% load is a bit of a free-for-all, but even there most 80+ Gold units exceed 80%. And beyond 20% load, the graph is essentially flat. It's time to stop this "50% load is what you should aim for" nonsense.

OK I have never read about or seen things like that 2nd graph you attached. Honestly having a difficult time grasping the difference between the two graphs, or i should say, what they are measuring differently to produce such a dramatic different curve. If both measure efficiency vs load I'm having trouble understanding your explanation on the difference. Sorry, I'm the slow guy who needs a "talk to me slow" or "can you repeat that" again kinda explanation lol

edit, so I get what you mean about your statement saying the measurable difference is so small it doesn't matter "in real world scenarios". ie running at 80% load might cost a few % less efficient, it's so small it's not going to be very noticeable to the user if the PSU will fail earlier because of this use case. If that is what you are getting at, I do understand. but regardless it's still true that technically it is most efficient at around that 50% load range

 

[EarthDog]

OK I have never read about or seen things like that 2nd graph you attached. Honestly having a difficult time grasping the difference between the two graphs, or i should say, what they are measuring differently to produce such a dramatic different curve. If both measure efficiency vs load I'm having trouble understanding your explanation on the difference. Sorry, I'm the slow guy who needs a "talk to me slow" or "can you repeat that" again kinda explanation lol

Look at the left side of the graphs.. Notice one starts at zero while the other starts at 78%? With the latter, the curve is more pronounced throughout the power range (20%-100%) due to the axis values being cropped.

If you look at the first graph and do not notice/take into account the axis values, then the curve looks a lot more 'peaky' than flat like it is (literally a 3% difference for 120V). Shrinking the range increases the granularity of the data and extenuates the curve.

It's time to stop this "50% load is what you should aim for" nonsense.

THIS.

edit, so I get what you mean about your statement saying the measurable difference is so small it doesn't matter "in real world scenarios". ie running at 60% load might cost a few % less effient, it's so small it's not going to be very noticeable to the user if the PSU will fail earlier because of this use case. If that is what you are getting at, I do understand.

Yeah, while I understand the thinking behind it (running a PSU at less power is less stress on the components), users are already building in plenty (if not too much) headroom already. I fully expect any quality PSU to be able to output its nameplate values for at least the life of its warranty. Of course, nobody does or should run 100% all the time, but to lop it in half for efficiency sake or a longer life is really splitting a split hair. Save your 9royal you, not you specifically) damn money and get an appropriately sized unit instead. I stick between 60-80% use, closer to the latter (I run a 750W unit and a 7960x (16/32t at 4.4. GHz along with a RTX 2080 Ti - also overclocked... my fan doesn't even spin up!).

 

[Bones]

the nominal/optimal load for PSU's are 50%, ie. where they are the most efficient.

I have to disagree but hey, to each his/her own right?
Why of course.

 

[lexluthermiester]

I think this thread has run it's course. The OP solved their problem on the first page and here things are...

 

[EarthDog]

I think this thread has run it's course. The OP solved their problem on the first page and here things are...

The OP has been responding to the great information here as of this morning... his issue is resolved, yes, then slizzo chimed in... (a repeat of info shared recently here at this site)..lol.

 

[Tatty_Two]

Thanks everybody, thread...………………….