• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

750-850W PSU upgrade

The GPU can take some undervolting too.
I'm running it between 1050~1070mV from default of 1150mV. This can bring the clock up within the same power limit.
Having it at 382W I also cap frequency at around 2900~2950MHz.

Yes at 2300MHz with 1150mV the TBP is down to 300W, no matter what the power limit is, I've tried it right now but that is totally untuned.
If you set GPU voltage at 1060mV you will be able to raise frequency at 2550MHz and still have 300W total GPU card power consumption.

Here

View attachment 381848

Most likely you could even go lower than 1050mV and raise frequency even more and still have 300W TBP.
You will have to find it with trial and error of how low you can go on voltage and how high on frequency to sustain 300W and keep your games stable without glitches and/or crashing. But at least on 1050~1060mV with 2550~2600MHz (~300W) should be rock stable.

I like to tune...
Its easy and fun to get the most out of any power configuration you choose.
Oh I did undervolt it. Not by much - best I managed thus far is 1100mV. Anything under that eventually crashed the 3D Mark's stress test. That's with the -10% power limit.
I guess will try again once I get the new PSU and confirm there are no crashes though.

It's probably about transient power spikes, of which RDNA 2 and Ampere were famous for. It's less common on RDNA 3, but could be the reason of your shutdown. Did you run your card at stock?
I'm not sure, I thought a card will not eat more than the specs say...
And no, I didn't, or not for a long time. That's the other thing, these shutdowns are very random looking. Nothing you could catch onto for troubleshooting.
 
Oh I did undervolt it. Not by much - best I managed thus far is 1100mV. Anything under that eventually crashed the 3D Mark's stress test. That's with the -10% power limit.
I guess will try again once I get the new PSU and confirm there are no crashes though.


I'm not sure, I thought a card will not eat more than the specs say...
And no, I didn't, or not for a long time. That's the other thing, these shutdowns are very random looking. Nothing you could catch onto for troubleshooting.
You better test the undervolt with your games. I always do it this way.
Having mine at 2900MHz some games require 1060~1070mV and others 1040~1050mV.
The point to me is to have the best on my games and not on benchmarks.

Anyway having the GPU around 300W with the 5800X the total system power (PSU out power) by average should be around 450W (CPU+GPU+rest of system) with peaks below 550W.
In theory a 650W PSU should be plenty enough, but also is a matter of PSU wear (if any).
 
Oh I did undervolt it. Not by much - best I managed thus far is 1100mV. Anything under that eventually crashed the 3D Mark's stress test. That's with the -10% power limit.
I guess will try again once I get the new PSU and confirm there are no crashes though.


I'm not sure, I thought a card will not eat more than the specs say...
And no, I didn't, or not for a long time. That's the other thing, these shutdowns are very random looking. Nothing you could catch onto for troubleshooting.
12v Rail, looks like it's capable of 54a 648w.

Did you peak over 55a of current? OCP would have kicked in for that maybe. Do some math.
 
12v Rail, looks like it's capable of 54a 648w.

Did you peak over 55a of current? OCP would have kicked in for that maybe. Do some math.
Never hurts to check this out but I am pretty confident that this system with @Octopuss configuration can't peak current over 42~43A on the 12V rail.

My similar system with 5900X + same GPU at 380W is well below 50A (48A peak) with 39~40A average on 12V rail.
Should be something else causing this, unless the PSU is starting to slide away
 
Last edited:
I'm not sure, I thought a card will not eat more than the specs say...
And no, I didn't, or not for a long time. That's the other thing, these shutdowns are very random looking. Nothing you could catch onto for troubleshooting.
Gotta look at the "spikes" data in reviews.

The factory TDP only means a consistent power level, but there's always a few ms spikes that your PSU has to handle just the same.
 
Gotta look at the "spikes" data in reviews.

The factory TDP only means a consistent power level, but there's always a few ms spikes that your PSU has to handle just the same.
Right...
At 380W max power limit my XTX can spike up to 490W with default voltage (1150mV)
With 1060mV and same 380W spikes are reduced a bit (470~480W).

With 300W and 1100mV spikes shouldn't be more than 390~400W.
I can add roughly another 150W for CPU+rest of system (on peak) and that brings us to 550W (max/peak) that I was estimating on post #27

@Octopuss check from HWiNFO the sensors below

1. Total Board Power (TPB): This shouldn't exceed the configured power limit more than 5W
2. GPU Power Maximum: Peak GPU power including spikes (shouldn't exceed 400W)
3. GPU PPT Limit (Short): Your configured TBP from adrenalin (~315W with standard VBIOS and -10%)

Marked in yellow the 2 and 3.
1 (TBP) is above 2 (peak).

Untitled_242.png
 
Last edited:
I certainly hope it won't happen with the new PSU.
I'll know more in a few days I guess.
 
Personally using a Corsair SF1000L, seen other people complain about it being noisy, but doesn't bother me any. Though I mostly picked it because larger PSU's don't fit in my case.
 
It looks like I might have a bigger problem somewhere.
I have just had even weirder thing happen twice in a row soon after I turned the PC on: the screen would start getting corrupted (like, hm, random parts of it turning into lots of small black squares) and flicker, eventually turning black. However, the system was still running normally, keyboard and mouse remained functional, but the monitor turned completely black. I had to reboot the PC. WTF. Then it happened again after a short while. I tried replugging the DP cable and then replugging it to another DP port on the graphic card, which didn't do anything at first , but then the screen flickered momentarily and then eventually got image back.
There were no driver errors and nothing in the event log.
 
It looks like I might have a bigger problem somewhere.
I have just had even weirder thing happen twice in a row soon after I turned the PC on: the screen would start getting corrupted (like, hm, random parts of it turning into lots of small black squares) and flicker, eventually turning black. However, the system was still running normally, keyboard and mouse remained functional, but the monitor turned completely black. I had to reboot the PC. WTF. Then it happened again after a short while. I tried replugging the DP cable and then replugging it to another DP port on the graphic card, which didn't do anything at first , but then the screen flickered momentarily and then eventually got image back.
There were no driver errors and nothing in the event log.
The small black squares sound like dying VRAM to me. :(
 
That's weird, why would it? The card is only like half a year old. I bought it from a friend who stopped gaming/streaming for family reasons, and the card is in mint condition with warranty and everything (and I know for a fact he wouldn't sell a defective thing, not when it has warranty and he could easily have it replaced if there was a problem).

It also happened right after booting up, not gaming or any load - but then it didn't happen when running 3D Mark for a few hours. It's just weird.

The only thing I did to it was repasting it (which I do out of a habit with everything), but that was several days ago.
 
That's weird, why would it? The card is only like half a year old. I bought it from a friend who stopped gaming/streaming for family reasons, and the card is in mint condition with warranty and everything (and I know for a fact he wouldn't sell a defective thing, not when it has warranty and he could easily have it replaced if there was a problem).

It also happened right after booting up, not gaming or any load - but then it didn't happen when running 3D Mark for a few hours. It's just weird.

The only thing I did to it was repasting it (which I do out of a habit with everything), but that was several days ago.
It might be a coincidence, but my 2070 also died shortly after I repasted it. I tend to have an "if it's not broken, don't fix it" approach. Often, there's more harm in fiddling with stuff that shouldn't be fiddled with.

Did you also do the pads on the VRAM?
 
It looks like I might have a bigger problem somewhere.
I have just had even weirder thing happen twice in a row soon after I turned the PC on: the screen would start getting corrupted (like, hm, random parts of it turning into lots of small black squares) and flicker, eventually turning black. However, the system was still running normally, keyboard and mouse remained functional, but the monitor turned completely black. I had to reboot the PC. WTF. Then it happened again after a short while. I tried replugging the DP cable and then replugging it to another DP port on the graphic card, which didn't do anything at first , but then the screen flickered momentarily and then eventually got image back.
There were no driver errors and nothing in the event log.

Could be driver related, or some other software shenanigans.
 
I brought the new PSU home, installed it, turned the PC on, and righ as desktop was supposed to show up, the screen instantly turned black. Reboot, same story. I was ready to get desperate, but then got an idea, took my wife's 60Hz monitor, plugged it in, and sure enough, worked like a charm.

What the fucking hell does Cooler Master has instead of quality control? The €600 monitor I bought is clearly completely broken garbage!!!
I have no idea what's going on with it, because aside from the horrible colour uniformity it worked for several days, but then the corruption started at completely random times.

Anyway, return initiated.
 
Back
Top