PSU Overcurrent Protection

[lemming9]

The specs of a PSU quote it having a +12V1 of 17A but also an OCP of 25A on that rail.

1. Doesn't that mean that the PSU can output significantly more useable power than the +12V1 suggests? An extra of almost 100W before the OCP kicked in.

2. Is this a normal level for OCP compared to the +12V spec?

3. With info like that would it imply that a PSU is not a real +12V dual rail but a single rail?

Thanks.

 

[HiTechObsessed]

Are you sure the 25A OCP is per rail, or is it one fuse for the two rails?

 

[lemming9]

Are you sure the 25A OCP is per rail, or is it one fuse for the two rails?

The official specs clearly state it's per rail. Any ideas?

 

[ko888]

What's the brand and model of the PSU?

The OCP circuit should be tripped when 25 Amps is reached for a period greater than 0.5 second according to the EPS12V Power Supply Design Guide v2.92.

The specifications say that the OCP set point for a +12V rail rated for 17 Amps continuous should be a maximum of 25.5 Amps peak current for a minimum of 500 msec.

 

[lemming9]

What's the brand and model of the PSU?

The OCP circuit should be tripped when 25 Amps is reached for a period greater than 0.5 second according to the EPS12V Power Supply Design Guide v2.92.

The specifications say that the OCP set point for a +12V rail rated for 17 Amps continuous should be a maximum of 25.5 Amps peak current for a minimum of 500 msec.

Thanks for that info. The PSU is an Antec Earthwatts. So given the OCP specifications you quoted it would seem accurate that this PSU has a OCP set at 25A on each of the rails. So if a rail is rated at 17A and has a OCP of 25A what would stop someone from using say 24A on that rail to power a more hungry system?

 

[ko888]

Thanks for that info. The PSU is an Antec Earthwatts. So given the OCP specifications you quoted it would seem accurate that this PSU has a OCP set at 25A on each of the rails. So if a rail is rated at 17A and has a OCP of 25A what would stop someone from using say 24A on that rail to power a more hungry system?

The closer you run the rail nearer to the trip point the greater the chance that you will trigger the OCP circuit since peak current draw of the device may be able to do just that.

 

[lemming9]

Thanks for that info. The PSU is an Antec Earthwatts. So given the OCP specifications you quoted it would seem accurate that this PSU has a OCP set at 25A on each of the rails. So if a rail is rated at 17A and has a OCP of 25A what would stop someone from using say 24A on that rail to power a more hungry system?

The closer you run the rail nearer to the trip point the greater the chance that you will trigger the OCP circuit since peak current draw of the device may be able to do just that.

Triggering the OCP is a possibility if getting all the way up to 25A. Still, 17A to 25A seems like a massive difference. Even going half way to 21A would provide an additional 48W. It almost seems like there is a significant amount of additional power that could be used. But surely there must be some other drawback to going above 17A other than a greater chance of tripping the OCP at 25A?

 

[ko888]

Triggering the OCP is a possibility if getting all the way up to 25A. Still, 17A to 25A seems like a massive difference. Even going half way to 21A would provide an additional 48W. It almost seems like there is a significant amount of additional power that could be used. But surely there must be some other drawback to going above 17A other than a greater chance of tripping the OCP at 25A?

The amount of power you can draw from the second +12V rail is reduced.

You can't exceed the PSU's maximum combined +12 Volt continuous current rating.

E.g. using the Antec EA-500:
Two +12V rails each rated to deliver a maximum load of 17 Amps.
Maximum combined +12 Volt continuous current rating: 34 Amps
If you draw 21 Amps from +12V1 then you are limited to a maximum draw of 13 Amps on the +12V2 rail.

You can't draw more than what the PSU is able to deliver.

 

[lemming9]

Triggering the OCP is a possibility if getting all the way up to 25A. Still, 17A to 25A seems like a massive difference. Even going half way to 21A would provide an additional 48W. It almost seems like there is a significant amount of additional power that could be used. But surely there must be some other drawback to going above 17A other than a greater chance of tripping the OCP at 25A?

The amount of power you can draw from the second +12V rail is reduced.

You can't exceed the PSU's maximum combined +12 Volt continuous current rating.

E.g. using the Antec EA-500:
Two +12V rails each rated to deliver a maximum load of 17 Amps.
Maximum combined +12 Volt continuous current rating: 34 Amps
If you draw 21 Amps from +12V1 then you are limited to a maximum draw of 13 Amps on the +12V2 rail.

You can't draw more than what the PSU is able to deliver.

That's exactly what I thought would be the case but was then reminded that the purpose of having separate rail distributions is to prevent demands made on one rail to be felt on the other.
i.e. that unused Amps on one rail would not get added to support a component on the other rail.

I know the +12V2 power draw is underused by almost 50% leaving what would be a full 7A free, so if that will support the +12V1 rail it would allow for increased power use on the +12V1 from 17A (+ 7A) to 24A which would of course be very good and still below the 25A OCP threshold.

Can you confirm that this will definitely be the case? If I can trouble you for some evidence/links to support this if you know of any it would also be great for peace of mind.
Also, are unused Amps simply added like that or would the actual available Amp be slightly different from how we calculated?

Thanks.

 

[hunter315]

The OCP on a rail is generally set notably higher than what it says on the label, this is to prevent the "trapped power" that single rail PSU marketing FUDs about.

Your PSU has a single 12V source, it then has two bundles of wires coming off of it, each with their own OCP set point, their OCP set points always add up to more than the combined power of the 12V source so that you don't have to perfectly balance your 12V rails to use all the capacity. Look at the EA-500D, it has 2 rails limited at 22A each but only a 37A source, so you can draw up to 22A from a rail or 37A from both rails.

I covered some of the power distribution questions better in a post i wrote up a while ago, the single vs multiple 12V rail one from my sig

 

[lemming9]

The OCP on a rail is generally set notably higher than what it says on the label, this is to prevent the "trapped power" that single rail PSU marketing FUDs about.

Your PSU has a single 12V source, it then has two bundles of wires coming off of it, each with their own OCP set point, their OCP set points always add up to more than the combined power of the 12V source so that you don't have to perfectly balance your 12V rails to use all the capacity. Look at the EA-500D, it has 2 rails limited at 22A each but only a 37A source, so you can draw up to 22A from a rail or 37A from both rails.

I covered some of the power distribution questions better in a post i wrote up a while ago, the single vs multiple 12V rail one from my sig

Again, this is how I figured rail distribution worked. Seems my confusion was one of simply using the wrong terminology to explain the process that unused current from say +12V2 is added to that of +12V1, to which the reply was no, as unused current doesn't transfer. That was obviously a completely correct reply in that it doen't transfer between the rails, but started making me doubt that anything beyond the the quoted load of each rail was possible even given unused power on another rail. Taking the higher OCP threshold per rail into account as well as the combined output from the same source obviously allows some flexibility.

So just to clarify, sticking with the EA-500D:
+12V1 = 22A
+12V2 = 22A
The maximum combined output of the +12V is 444W = 37A
(Obviously both rails can't simultaneously use the full 22A each regardless of this spec as that would exceed the total 500W with just these rails)

So, if we assume an OCP of 30A per rail
and +12V2 22A is using 10A leaving 12A free
the +12V1 22A could in theory use an additional 8A without problem before the OCP is triggered at 30A.
But since the combined source limit is 37A and +12V2 is using 10A this leaves +12V1 with an additional 5A
This is because (i) the +12V2 is underused and (ii) the OCP allows some additional headroom.

That's pretty much my understanding of things. Does that all seem correct?

 

[hunter315]

Yup, thats correct.


Though they generally organize the connectors so that your rails will be pretty well balanced, the CPU will generally be on a different rail than the motherboard and the PCI-e and peripheral cables will be split up among the rails so that unless you are using adapter you won't hit the OCP limit so it is usually pretty had to have extra capacity on one rail while needing more on another, but this was more of a theoretical discussion than a practical one.

 

[lemming9]

My thanks to you both. It's good to know I had the theory right.

The rails on an Antec Eathwatts are not really that balanced at all. +12V2 only powers the CPU and +12V1 does everything else. An average CPU would probably only use 8-10 Amps on the rail leaving a lot extra capacity, around more than 50% extra at stock on that rail.

Now that we know that there is some extra capacity which could be used for the other rail, generally how safe would you think it is to do so? Obviously each rail being under its OCP almost implies it should be safe.

But by my calculation, the combined capacity would be reached before the OCP on the rail was reached.
1. So does reaching the combined capacity rather than the rail OCP also have a protection similar to the OCP of individual rails? I assume there must be? It would only be sensible. i.e. is the combined capacity essentially like the 'rated' capacity of each rail and has a higher OCP itself (maybe at 12V1+12V2), or is the rated combined capacity essentially the combined OCP limit?

2. And more importantly, what if the limit is exceed?
a. Is there a risk to damaging components like GPU, MB, HDD on Rail 1 which is above its rated limit but still below its OCP limit?
b. Is there a risk to damaging the CPU on Rail 2 which is significantly below its capacity?
c. Or is the risk only to damaging the PSU as the safety features such as OCP etc are there specifically to prevent such damage to components?

4. Also, does using an adapter increase the risk further? ie 2 molex to 8-pin PCIe? I assume this doesn't necessarily increase the risk further, but is there an added issue with using an adapter to power a 150W load from 2 molex if there is spare capacity?

Thanks.

 

[ko888]

The Antec EarthWatts PSUs that have two 17 Amp +12V rails are the EA-380, EA-430 and EA-500 80 PLUS standard models all made by Seasonic. These PSUs are true single +12V rail designs despite what the label shows. Teardowns of these models show that there is no OCP protection on each virtual +12V rail.

The EarthWatts models that are made by Delta Electronics and FSP Group do have multiple virtual +12V rails with each +12V rail protected by OCP.

 

[lemming9]

Seems risky that there is no OCP at all on the rails on certain models. Having said that as they are really single rails I assume the maximum 'combined' 12V capacity would have some overall OCP features.
Even on the versions with dual rails that have OCP on each rail, I think the maximum combined capacity would more likely be reached first before the OCP per rail.

In any case, is there a major risk to the system if operating at or close to capacity, or if it was to very occasionally exceed capacity?
Would there be a risk only to the 'overloaded' GPU rail or also the the underloaded CPU rail? i.e. could it damage the GPU/CPU or both?
Or is the risk only to damaging the PSU when the OCP triggers and so no harm in trying and seeing if it works below or near capacity?

Thanks

 

[ko888]

The Antec EA-380, EA-430 and EA-500 that is made by Seasonic don't have OCP (Over Current Protection). They only have OVP (Over Voltage Protection), UVP (Under Current Protection), SCP (Short Circuit Protection), and OPP (Over Power Protection).

SCP and OPP would be what you have to rely on but those only apply to the entire PSU and not individual rails.

If a multi +12V rail PSU is properly designed and implemented there should be no risk to either component.

The multiple protections in the PSU are suppose to protect the attached components and also protect the PSU itself. This is assuming that the circuits are properly designed and implemented.

 

[lemming9]

So with an Antec Earthwatts or any half decent PSU with all those protections there's really no real risk of any damage to components even if the PSU is at or near max +12V capacity or even overloaded?

 

[ko888]

So with an Antec Earthwatts or any half decent PSU with all those protections there's really no real risk of any damage to components even if the PSU is at or near max +12V capacity or even overloaded?

I can't make that guarantee since I'm not the manufacturer that has to honor or deal with whatever may happen. You would have to search for any customer feedback that reports damage to any components when their PSU failed.

Running a modern PSU at or near its maximum +12V rating will lead to a shorter lifespan for the PSU since the heat generated by the high power draw will accelerate the degradation of the electrolytic capacitors. Once the electrolytic capacitors have deteriorated to the point where they are unable to suppress high levels of ripple, it will lead to damage of the electrolytic capacitors on the motherboard, graphics card, hard disk drive controller, etc.

Torture testing of a PSU is usually done by drawing 80% of the PSU's total capacity for a period of 8 hours to determine if there are any adverse effects.

 

[lemming9]

Thanks. I somehow doubt that any manufacturers would consider damage to other components as part of any policy. I was really just asking about likelihood and user experience of any such damage occurring to other components.

I wouldn't mind shortening the lifespan of a PSU, but damage to GPU, MB, CPU and HDD is something I would not like to experiment with if user experience shows it to be a likely even with a good quality PSU.

I'd be drawing well into the 90% range so this was something I wanted to confirm before I tried it on the expectation that PSU Protections would be sufficient.

Thanks.

 

[ko888]

Some PSU manufacturers will replace damaged components if they are able to determine that their PSU caused that damage. This is usually handled on a case by case basis and requires that the system be sent to the PSU manufacturer for analysis by their engineers.

Unfortunately no PSU monitors its own noise and ripple output levels and shuts itself down if it exceeds ATX12V specs. You don't know if any damage is occurring until you begin to experience unexplainable system instability and by then it's too late.

 

[lemming9]

As generous as that could be I think such a claim would be a hassle best avoided.
Thanks for all your replies.