2 Corsair RM750x PSUs cutting out at 250w below rated power

Joe42

Joe42

Joe42

Soldato
Joined
18 Jan 2005
Posts
4,181
Location
North West West Northamptonshire
I had a 2021 model (I think, no 12v HPWR connector) RM750X cutting out after doing a CPU upgrade and enabling a higher TDP.

It was drawing up to 615w on a watt-meter under max load before it switches off. With conversion losses the max on this unit should be around 840w at the wall if I'm not mistaken, assuming 88% efficiency.
System switches off, won't turn on with the power button, has to be switched off at the wall and back on to power back up.

Temps all good, will run Prime95 or Furmark fine for hours, but not both at the same time.

Had the PSU RMA'd, they sent me a new one different serial number. Exact same problem.

I then added a second watt-meter, readings are within 40w of the other. Found I can run Furmark plus prime95 on 3 cores, but on 4 it cuts out within minutes once the PSU is warm.
This is totally repeatable. PSU outlet is at 25c, minimal fan noise, no heat on the PSU case. It looks like a bad batch but Corsair support aren't buying it. It's also clear they don't load test their refurbs.


If you have this PSU or any similar Corsair model and your system is switching off during heavy use, this might be your issue.

Worse still, if you have this model you may have a 550w PSU instead of a 750w without realizing it. I only found out by accident due to an upgrade.
By the time you find out it might be out of warranty, and you would never suspect a PSU that has been working fine for years, I didn't.

Even worse than this, if you have an AIO and your PSU cuts out under heavy CPU load, I suspect it could damage the CPU as the pump goes off and there is nowhere for the heat to go.

Anyone else had issues like this with 2021 RMx units?
Already seen a post in general hardware that looks to be the same issue.
 
hardware spes would be good to know, if you have a highish end pc the unit may be tripping up due to the spike wattages modern gpus can pull, only for a millisecond but its enough to tip out a psu.
 
Last edited:
mickyflinn said:
What PSU are you using now ?
Click to expand...
The refurb replacement RM750X, which is doing exactly what the original was.

I can game on it if careful and drop detail settings, but over 560w or so at the wall and it cuts out.

2 EPS cables installed, and runs prime95 for hours fine. Wouldn't this eliminate the possibility of a motherboard cut-out?
Running a GPU load shouldn't affect that and I don't think GPUs can shut off and lock out a system like this?
 
wookiee87 said:
hardware spes would be good to know, if you have a highish end pc the unit may be tripping up due to the spike wattages modern gpus can pull, only for a millisecond but its enough to tip out a psu.
Click to expand...
If it's cold it will run prime95 plus furmark for a few minutes before cutting out, when warm it's instant which suggests over current protection to me.

I had thought measuring with a watt-meter makes the specs irrelevant but hadn't thought about spike tripping. But shouldn't the PSU handle this? I would consider it defective.

Specs are:
9700x at 170w tdp.
X670e w/2x pci-e 4.0 ssds
2 sticks of ddr5 6000
6950xt
bluray drive
420mm aio

I have it measured at 615w at the wall max. Fairly confident it's the PSU, again.
 
Joe42 said:
I had thought measuring with a watt-meter makes the specs irrelevant but hadn't thought about spike tripping. But shouldn't the PSU handle this? I would consider it defective.
Click to expand...
Not saying that is your problem, but when the 6950 XT came out, it was notorious for tripping PSUs. Some users had to go 1000+ watt to solve the problem, if I recall correctly.
 
Tetras said:
Not saying that is your problem, but when the 6950 XT came out, it was notorious for tripping PSUs. Some users had to go 1000+ watt to solve the problem, if I recall correctly.
Click to expand...
I had no idea about this. Wouldn't it be instant though and not take several minutes?
Also if I load up Prime95 while furmark is running, it will trip. It never trips with furmark alone. The spike would only be on starting furmark right?

I wonder if that was because this was pre 4090 5090 and it was unusually power hungry for the time, but not really by todays standards. And I would expect a Corsair to be able to handle it.
 
Joe42 said:
I wonder if that was because this was pre 4090 5090 and it was unusually power hungry for the time, but not really by todays standards. And I would expect a Corsair to be able to handle it.
Click to expand...
My understanding is that the 4000 series onwards did something to curb the power spikes, it was the 30 series and RDNA2 that were most well known for it, though PSUs weren't ready then, so there's that too.

Joe42 said:
The spike would only be on starting furmark right?
Click to expand...
I think the highest spike is at the start, yes, but it does spike at other times.

I believe a lot of the highest spikes in reviews are recorded using Furmark.

Joe42 said:
It never trips with furmark alone.
Click to expand...
Not necessarily conclusive because it could just be the baseload of Prime pushing high enough that the graphics card spikes cause the trip.
 
Last edited:
6950xt transient spikes up to 450w because it's so brief your power meter won't register these spikes. And no the spikes will happen throughout the furmark run

www.techpowerup.com

AMD Radeon RX 6950 XT Reference Design Review

The AMD Radeon RX 6950 XT Reference Design card not only looks great, but also achieves excellent performance, matching the GeForce RTX 3090, but at much better pricing. Because of good optimization and surprisingly good energy efficiency, AMD's thermal solution works very well, too.
www.techpowerup.com www.techpowerup.com

Minimum recommended PSU: 1000 W
Click to expand...
 
Last edited:
Thanks for the link and info both.

Even with a max 450w spike, it draws 350w continuous which is what I'm seeing on the watt-meter in my system load, and I have at least 225w of headroom on the PSU so I still feel this is unacceptable.

Plus since when do spikes count as PSU load? The rating is continuous and reviewers found the RM750x would give about 50w over even that.
I'm also confused as to why the PSU is rated at 750w on the 12v rail and in total.

If it was a cheap PSU I wouldn't mind, but for Corsair I expect better.
 
@Joe42 transient spikes are milisecond events that can easily cause powersupplies to trip up, when i had my liquid devil 6900 xt the card would happily pull 525w peak power with sustained power in the high 450w ranges, now i have a 5090 fe and in the screenshot you can see although gpu power is 530w the card did pull a maximum of 736w at some stage under load, its that spike which might be tripping out your psu when you have the cpu and gpu going at 100% load.

also a 9700x at 170w tdp doesnt mean the cpu will stop at that number, to provide 170w the cpu might actually be pulling close if not more like 220-230w to provide 170w limit you have set.

have a look at downloading hwinfo64 and check gpu power and cpu power figures, it might be more than what your psu can handle

imgur.com

imgur.com

Discover the magic of the internet at Imgur, a community powered entertainment destination. Lift your spirits with funny jokes, trending memes, entertaining gifs, inspiring stories, viral videos, and so much more from users.
imgur.com imgur.com
 
Last edited:
Thanks for your comments. This is all true and no doubt is causing the issue. I don't seem to have a rail power sensor on mine in HW64.

However the bottom line is the rating of the PSU is continuous, so it shouldn't be tripping unless that rating is exceeded for a sustained period, and the watt-meters don't lie.
If it was a cheap PSU I could understand, but an expensive premium grade Corsair unit should be able to handle transient spikes, as that's surely part of a normal PC load?

The fact that the unit isn't overheating and outlet temps are low suggests the components can handle it, but over-current protection is tripping too readily.
If they want to rate the unit as 750w continuous, 550w peak that's fine by me, but they don't.

By way of an update, they are supposed to be sending me an 850w replacement, they haven't offered any explanation as to the cause and seem to think this is perfectly acceptable.
 
Joe42 said:
Thanks for your comments. This is all true and no doubt is causing the issue. I don't seem to have a rail power sensor on mine in HW64.

However the bottom line is the rating of the PSU is continuous, so it shouldn't be tripping unless that rating is exceeded for a sustained period, and the watt-meters don't lie.
If it was a cheap PSU I could understand, but an expensive premium grade Corsair unit should be able to handle transient spikes, as that's surely part of a normal PC load?

The fact that the unit isn't overheating and outlet temps are low suggests the components can handle it, but over-current protection is tripping too readily.
If they want to rate the unit as 750w continuous, 550w peak that's fine by me, but they don't.

By way of an update, they are supposed to be sending me an 850w replacement, they haven't offered any explanation as to the cause and seem to think this is perfectly acceptable.
Click to expand...

run a game with hwinfo64 up and post a screenshot of the cpu and gpu information, if the combined load is high then the issue is 100% a underspec'd psu, synthetic tests are the very worst thing you can do with hardware, a test for stability is fine but hammering the components over and over wont do it much good, in regard to corsair sending out a 850w, they are very good in that department, they most likley think its a power issue, 850w is better and should sort you out, i'd have opted for 1000w to be absolutly sure, but maybe set your cpu back to 105w tdp just in case with the new unit.

had a play in a few games and this was my very worst spike wattage on my 5090fe, 953w :eek:

U7BLluN.png
 
Last edited:
Joe42 said:
However the bottom line is the rating of the PSU is continuous, so it shouldn't be tripping unless that rating is exceeded for a sustained period, and the watt-meters don't lie.
If it was a cheap PSU I could understand, but an expensive premium grade Corsair unit should be able to handle transient spikes, as that's surely part of a normal PC load?
Click to expand...
It wasn't considered part of a normal load. At least, not officially, not until the ATX 3.x spec made it mandatory. It does depend on the card and the 6900/6950 happened to be examples on the higher-end of these spikes.

Joe42 said:
The fact that the unit isn't overheating and outlet temps are low suggests the components can handle it, but over-current protection is tripping too readily.
If they want to rate the unit as 750w continuous, 550w peak that's fine by me, but they don't.
Click to expand...
Sometimes that's the case, that the manufacturer was too conservative, but one outcome is that it can cause excessive ripple and push the PSU out of spec, which degrades components over time.
 
Tetras said:
It wasn't considered part of a normal load. At least, not officially, not until the ATX 3.x spec made it mandatory. It does depend on the card and the 6900/6950 happened to be examples on the higher-end of these spikes.

Sometimes that's the case, that the manufacturer was too conservative, but one outcome is that it can cause excessive ripple and push the PSU out of spec, which degrades components over time.
Click to expand...
Do PSU manufacturers really need an industry body to inform them that GPUs draw power spikes and their PSUs need to handle that?

I suspect the increase in warranty length to 12yrs has led them to implement overly conservative OCP, meaning we now have to buy more powerful PSUs than we actually need.
I'm not impressed by this, they should spec their components to handle it, or provide a separate peak rating. Lousy engineering.

Also disappointing none of the reviewers are testing this, meaning none of us know what rating our PSUs actually are. Very difficult to find peak power ratings for GPUs, impossible to find peak ratings for PSUs, so it's a complete lottery.

My GPU rail power just sits at around 287w, and it doesn't expand to show the additional sensor as I don't think it has this one. However TechPowerup measured the spikes at 450w max.
That still doesn't put me near the limit on the continuous rating, assuming the CPU load is less spiky. I have almost 250w of headroom, 150w for the GPU spikes leaves 100w extra for the CPU.
 
You must log in or register to reply here.
Back
Top Bottom