Is it bad if I have a good 550W PSU that's pulling 500W from the plug? Is it just asking for trouble/housefire?

Is it bad if I have a good 550W PSU that's pulling 500W from the plug? Is it just asking for trouble/housefire?

(Ryzen 5 1600 + Vega 64, pulling 495W with prime95 and gpu miner running for stress test)

(500/550)*100%=91%
if it's 80+ certified, you might be scraping on the limit. 600W should be safe.

A 550W PSU theoretically can supply up to 550W of DC current. So if your PSU operates at 85% efficiency while drawing 500W from the outlet, it's actually supplying 425W to your PC. 550 - 425 = 125W of headroom, which isn't ideal but it shouldn't cause a catastrophic failure unless it's cheap shit. Also keep in mind that % efficiency varies depending on how many watts the PSU is supplying. Read some reviews on jonny guru and you'll get a better picture than I can explain.

It's platinum rated, supposedly that means it's 91% efficient at 100% load... But I thought that meant it would use 550/.91= 604W from the wall, as opposed to a less efficient PSU which still would supply 550W but it would pull more from the plug??

Jonny Guru said this PSU (super flower leadex platinum) was good, but I just changed my GPU and this consumes shitloads more (like 100W extra)

Pic related

>Apple is the one with the retard expression
Heh

Is your system unstable?
Is your PSU noticeably hot or loud?
If no than it's fine. Good PSUs are designed to work at 100% load 24/7.

If you're drawing 500W during stress testing then I think it's fine. Prime95 on its own stresses your CPU more than any real world application ever will because it concentrates stress testing on floating point calcs. So really your 500W draw is a theoretical maximum. Try running Superposition or 3dmark or encode some video files or whatever and see what your power draw is.

user from is on right track, however if you're planning on scrapping bottom of the barrel you need to pay attention how much power can be supplied on each power rail, especially 12V.
>picture as example
+3.3 and +5V are used to power mobo and drives, most PSUs have very similar current rating for these rails regardless of total power rating.
+12V1 is usually hooked up to CPU VRM, and ATX power; rated at 20A should provide up to 240W
+12V2-4 are split between EPS, Drives, and PCI-E
As you can see PSU rated at 850W can provide only 760W on 12V rails

Super flower is good quality, but I wouldn't sleep easy knowing I might be trying to pull 40A from ~35A rated 12V rail.

Just had a crash

Prime95 = fine (160W)
Ether mining = fine (415W)
Prime 95 + ether mining = fine (495W)
Furmark = fine (425W)
Furmark + Prime 95 = crash (Saw 490 before it went down)

Is it worth connecting the two pcie power inputs to two different modular cables out of the PSU (a single one splits into two, that's how I connected them)? Or is this definite proof that it's not working?

You are literally running two fucking benchmarks before it crashes. You can even run coin mining AND a benchmark and it won't crash. take that as a W and move the hell on. You'll be fine.

These are the detailed specs, supposedly I can pull 45A from 12V...

It DID crash with furmark + prime95, and the GPU wasn't hot then... So it's probably not enough. Worst part is that I got this PSU instead of a seasonic 650W just because the seasonic would've taken a week to arrive

you can try doing that but I don't think it will help much if at all

That does make sense... I was trying to gauge long term stability because I have a one-week return window on GPU... I could return it and get a lower power 1070 or something if it looks like it'll be trouble.

The PSU has a single 12V rail, I'm not sure whatever that means exactly but I think it indicates a second cable won't be making any difference (except maybe more 1W extra lost from another modular cable being plugged in)

Well, that's a pleasant surprise, they rate their PSU by 12V rail. If you're not gonna try hitting it this hard all the time it should be alright.
If you're in yurop you can order seasonic, and return the super flower within 2 weeks of purchase without giving any reason.

I got the PSU one month ago sadly, can only return the GPU (which I got today).

Do some real life tests instead of benchmarks and if it keeps crashing consider that seasonic and sell the super flower on ebay or something.

Or novideo but I didn't say that.

>links to stupid questions general
>gives stupid answer
Sup Forums ladies and gents.

I'm going to try undervolting the vega (supposedly some can keep the same performance)... if it works it'll probably solve my woes

Isn't that the point?

So I underclocked the vega a bit, now the core runs at 1683MHz instead of 1750MHz (down to the level of the air cooled version, this is the liquidshit edition) and I overclocked the HBM from 945MHz to 1100MHz

Now the card consumes 30W less and the temps are 8 degrees lower. And the performance has improved. This should do.

Not quite true. Some power supplies can supply their rating in DC watts, which of course means they draw more AC watts than what is written on them. For example a Corsair rm650w can supply 650W to the system fine (measured in benchmarks), at which point it draws ~740W from the wall.

It is true that you'd be scraping the limit running a 550W with Vega, though. I guess if you run it with a lower power limit, perhaps set up proper voltages to lower the power usage, then you should be fine. Don't try overclocking though.

When I ran prime95 this time it didn't crash (total power at 440W instead of 495W) but VDDC into the GPU started getting unstable and it started randomly downclocking even though the temps were good.

Or was it the other way around, because the CPU was busy not enough data was fed to the GPU and it downclocked because there was less work that could be done? How do I know?

Look up undervolting the Vega. They released it with super high voltages so it can push max clocks, which is why it ended up as a 300W card. Some people undervolted it a lot and had it running using ~200W or something like that, with performance not dropping due to less heat and less throttling.

>(Ryzen 5 1600 + Vega 64, pulling 495W with prime95 and gpu miner running for stress test)

Jesus christ

Doing that now Can do 1680MHz with 1050mV (230W) and 1645 with 1000mV (200W)... I'll see if the latter can survive furmark + prime95

Yup, we got stability now

downclocking from 1.2 to 1V reduced the power consumption by 60W, the fan is now actually silent and it's just 100MHz slower... The HBM is faster now which should help in mining.

In fact I've accidentally created a mining monster. Performance went up from 33MH/s to 43MH/s, while the system power consumption is down to 320W from 415.

In normal usage you would be pulling about 50W less than those stress tests so you're all good.