Does this PSU will do enough?

[mclaren85]

So I built an AMD system but I didn't pay attention for psu. I bought a Seasonic 80+ Gold 650w modular type. Do you think this psu can handle all power requirements at max?

My specs are as follows:

Amd Ryzen 9 3900x
Gtx 1660
2tb hdd
500gb nvme
16 gb 3600 mhz ram
asus tuf gaming x570 mainboard.

Thanks in advance!

 

[dirtyferret]

Doubtful you even push that PSU to 50% load while gaming, more than enough power.

 

[Frick]

Yeah. That PSU would drive pretty much anything today.

 

[mclaren85]

Thank you guys, that's a real relief for me

 

[Zareek]

Agreed, more than enough.

 

[Chrispy_]

So I built an AMD system but I didn't pay attention for psu. I bought a Seasonic 80+ Gold 650w modular type. Do you think this psu can handle all power requirements at max?

My specs are as follows:

Amd Ryzen 9 3900x
Gtx 1660
2tb hdd
500gb nvme
16 gb 3600 mhz ram
asus tuf gaming x570 mainboard.

Thanks in advance!

Perfect pairing, IMO.
Your system will use about 325W at max load (125W GPU, 150W CPU, 50W for fans/drives/motherboard)
Power supplies are at their most efficient at 50% load, so you've accidentally hit the sweet spot at exactly 50%

 

[Zach_01]

Perfect pairing, IMO.
Your system will use about 325W at max load (125W GPU, 150W CPU, 50W for fans/drives/motherboard)
Power supplies are at their most efficient at 50% load, so you've accidentally hit the sweet spot at exactly 50%


View attachment 183203

Actually it would be less than 300W
No ZEN2/3 uses all its PPT limit during gaming.
Usually all Zen2/3 parts, from 6 to 16 cores are using around 50~60W (avg) on stock settings.
And the rest system would be 50~100W without GPU.

Mine is around 350~380W

 

[Chrispy_]

Actually it would be less than 300W
No ZEN2/3 uses all its PPT limit during gaming.
Usually all Zen2/3 parts, from 6 to 16 cores are using around 50~60W (avg) on stock settings.
And the rest system would be 50~100W without GPU.

Mine is around 350~380W

Yes I know.

He said "at max", That means CPU-BURNER + OCCT/Furmark synthetic full load testing, not gaming. Closest real-world equivalent to that would be high-quality CPU stream encoding whilst running a GPU-limited game.

My 3900X uses about 70W in modern gaming, according to Ryzen Master, which is barely 40% of the PPT target I've set.

 

[Zach_01]

While that is true we cant be sure its what he ment. There is no practical usage to run this kind of stress combination other than torture.
Yes I've done a similar thing once out of curiosity with CB R20 + Heaven4.0. I never use Prime95/Furmark abusers.

 

[Chrispy_]

While that is true we cant be sure its what he ment. There is no practical usage to run this kind of stress combination other than torture.
Yes I've done a similar thing once out of curiosity with CB R20 + Heaven4.0. I never use Prime95/Furmark abusers.

I just gave you a reasonable, common example; CPU encoding a game stream. Probably loading 6-10 cores depending on game and encode settings which should reach the package power limit for the CPU whist the game runs the GPU at its power limit.

In the context of the thread, I'm going to guess that OP meant "peak possible power his PC is likely to consume". This isn't a thread about gaming power consumption, it's a question about whether his PSU rating will ever be exceeded. Plenty of recommendations on this forum to run combined CPU+GPU synthetic loads for diagnostics - but there are plenty more other common scenarios that can fully load both CPU and GPU either intentionally or accidentally (looking at you, Chrome and Adobe....)

It would 100% suck to have a PSU that could only deliver enough power for gaming loads. It would be crashing/resetting multiple times a day, maybe even multiple times an hour on your typical Windows 10, automatic update, Chrome, Steam, Discord gaming PC. Hell, I can tell when my laptop's doing a background update because all 8 threads get maxed out and the fans ramp up even though it's been asleep for an hour and I haven't even woken it up yet.

 

[mclaren85]

Perfect pairing, IMO.
Your system will use about 325W at max load (125W GPU, 150W CPU, 50W for fans/drives/motherboard)
Power supplies are at their most efficient at 50% load, so you've accidentally hit the sweet spot at exactly 50%


View attachment 183203

Thank you for this great explanation. I will execute the heaviest benchmark test available after building my system.
I trust your technical knowledge, have a nice day!

 

[AleXXX666]

My FOCUS-GX-650 easily worked with 1070 Ti OC (~200W), and Ryzen 7 2700 OC (115W), and loots of fans

 

[Frick]

Perfect pairing, IMO.
Your system will use about 325W at max load (125W GPU, 150W CPU, 50W for fans/drives/motherboard)
Power supplies are at their most efficient at 50% load, so you've accidentally hit the sweet spot at exactly 50%


View attachment 183203

Slightly off topic, but buying a PSU based on 50% load is a waste of money.

 

[AleXXX666]

Slightly off topic, but buying a PSU based on 50% load is a waste of money.

Lmfao, buying PSU based on 100% load is a waste of money, cuz it will die after ~3 yrs of will provide hard undervoltage for your components. 50% load is perfect because then system has enough power to run, you could oc anything you want, or replace components for ANY you want. Of course we are talking about any mid-high end machines here. If you need ms word typewriter or “htpc”-crap then you are happy with cheapest 400-450w psu, or even 350w if you are perfectionist. ;D

 

[Frick]

Lmfao, buying PSU based on 100% load is a waste of money, cuz it will die after ~3 yrs of will provide hard undervoltage for your components. 50% load is perfect because then system has enough power to run, you could oc anything you want, or replace components for ANY you want. Of course we are talking about any mid-high end machines here. If you need ms word typewriter or “htpc”-crap then you are happy with cheapest 400-450w psu, or even 350w if you are perfectionist. ;D

There is a middle ground between 50% and 100%.

 

[Chrispy_]

Slightly off topic, but buying a PSU based on 50% load is a waste of money.

Technically yes. If you're only using 325Watts and the power supply can deliver 650W then you are leaving half of the potential power delivery on the table.

However, the lifespan, efficiency, heat, and noise of a 325W PSU providing 325W would be bad, and I doubt it would survive much beyond the warranty period (if it even reached it at all) operating 24/7 at 100% load, maximum, loud fan speed, and at it's maximum rated component temperature.

The other thing that prevents you from realistically using a 325W-rated PSU to power a PC with a recorded peak TDP of 325W is that the recorded peaks are all actually averages. Even when you look at GPU reviews and the 3080 needs 350W at it's peak, that's only an average of 350W over the sampling period of the driver - probably around 2 seconds based on how fast I observe the Nvidia driver reacting in situations where the driver reports "power limit" as the reason for reducing clocks on my 2070S. When sampling at 0.1 seconds, my 2070S can actually spike multiple times a minute at more than 115% power limit. I would suspect that a shorter sample time might even make that higher still.

Rather than argue about hypotheticals that are difficult to prove because I don't have an easy way to sample power usage faster than every 0.1s, let's talk about all the 3080 launch reviews. Most of them were using i9 or Ryzen9 test benches with very few components other than a CPU, cooler, and GPU. That means that the system power draw numbers are relatively easy to calculate - they should be the PPT of the CPU, the TDP of the chipset, the power rating of the CPU fan, and the total board power of the GPU. In fact, the kill-a-Watt power readings performed by a few reviewers matched that formula almost exactly, once adjusting for the PSU efficiency at the measured power draw levels. That's not really surprising because (providing it's not OC'd, and the BIOS PPT limits are at stock or set manually) all of the power-consuming parts have fixed, known power-draw caps.

So why, then, for an i9+3080 system that was measured to draw only 635W from the wall at full synthetic load (250W PL2 + 350W GPU + 8W Z490 + 7W CPU cooling) fail to run using 750W and 850W Seasonic Platinum power supplies? Linus, Steve, Jay all had to swap out their PSUs for higher-Wattage units for their 3080FE reviews. By your argument, the 750W PSU would only have been running at 82% load, and the 850W PSU would only have been running at 72% load.

Answer: Because instantaneous peak power draw in a system is higher than the average peak power draw in a system, even if only for fraction-of-a-second bursts. The result is bluescreens, crashes, GPU driver restarts. I don't make the rules, that's just how it is; If you don't like it go and argue with the laws of physics, Maybe pick a fight with James Watt, Charles de Coulomb, or Otis Boykin instead of PSU manufacturers, Nvidia, or us forum-dwellers.

There is a middle ground between 50% and 100%.

Well, based on Ampere reviews, 65% might be pushing your luck because even 72% turned out to be unstable for all three of the top three most popular Youtube reviewers (by subscriber count).

Also, if your CPU+GPU combined load is under 250W (pretty common, actually) then you're out of luck because most reputable PSU vendors start at 500W for their lowest model. The most popular gaming spec on Steam appears to be a quad-core i5 and GTX1060, pulling around 200W at full load. That's only 40% of a 500W PSU.

 

[AleXXX666]

Technically yes. If you're only using 325Watts and the power supply can deliver 650W then you are leaving half of the potential power delivery on the table.

However, the lifespan, efficiency, heat, and noise of a 325W PSU providing 325W would be bad, and I doubt it would survive much beyond the warranty period (if it even reached it at all) operating 24/7 at 100% load, maximum, loud fan speed, and at it's maximum rated component temperature.

The other thing that prevents you from realistically using a 325W-rated PSU to power a PC with a recorded peak TDP of 325W is that the recorded peaks are all actually averages. Even when you look at GPU reviews and the 3080 needs 350W at it's peak, that's only an average of 350W over the sampling period of the driver - probably around 2 seconds based on how fast I observe the Nvidia driver reacting in situations where the driver reports "power limit" as the reason for reducing clocks on my 2070S. When sampling at 0.1 seconds, my 2070S can actually spike multiple times a minute at more than 115% power limit. I would suspect that a shorter sample time might even make that higher still.

Rather than argue about hypotheticals that are difficult to prove because I don't have an easy way to sample power usage faster than every 0.1s, let's talk about all the 3080 launch reviews. Most of them were using i9 or Ryzen9 test benches with very few components other than a CPU, cooler, and GPU. That means that the system power draw numbers are relatively easy to calculate - they should be the PPT of the CPU, the TDP of the chipset, the power rating of the CPU fan, and the total board power of the GPU. In fact, the kill-a-Watt power readings performed by a few reviewers matched that formula almost exactly, once adjusting for the PSU efficiency at the measured power draw levels. That's not really surprising because (providing it's not OC'd, and the BIOS PPT limits are at stock or set manually) all of the power-consuming parts have fixed, known power-draw caps.

So why, then, for an i9+3080 system that was measured to draw only 635W from the wall at full synthetic load (250W PL2 + 350W GPU + 8W Z490 + 7W CPU cooling) fail to run using 750W and 850W Seasonic Platinum power supplies? Linus, Steve, Jay all had to swap out their PSUs for higher-Wattage units for their 3080FE reviews. By your argument, the 750W PSU would only have been running at 82% load, and the 850W PSU would only have been running at 72% load.

Answer: Because instantaneous peak power draw in a system is higher than the average peak power draw in a system, even if only for fraction-of-a-second bursts. The result is bluescreens, crashes, GPU driver restarts. I don't make the rules, that's just how it is; If you don't like it go and argue with the laws of physics, Maybe pick a fight with James Watt, Charles de Coulomb, or Otis Boykin instead of PSU manufacturers, Nvidia, or us forum-dwellers.


Well, based on Ampere reviews, 65% might be pushing your luck because even 72% turned out to be unstable for all three of the top three most popular Youtube reviewers (by subscriber count).

Also, if your CPU+GPU combined load is under 250W (pretty common, actually) then you're out of luck because most reputable PSU vendors start at 500W for their lowest model. The most popular gaming spec on Steam appears to be a quad-core i5 and GTX1060, pulling around 200W at full load. That's only 40% of a 500W PSU.

Sorry for off-topic and then some “techno”-tubers say that Seasonic is crap because it is smart enough and have a lot of protections, and some cheaper psu doesn’t have it and succesfully runs that 3080, but could be potential time bomb lol

 

[Chrispy_]

Sorry for off-topic and then some “techno”-tubers say that Seasonic is crap because it is smart enough and have a lot of protections, and some cheaper psu doesn’t have it and succesfully runs that 3080, but could be potential time bomb lol

Well yeah - if the Seasonic 850W PSU is tripping its safeties with a ~615W rated load, then you have to wonder what damage is being done to PSUs by higher-power GPUs with PSUs that lack those safeties?

Hopefully it's nothing more sinister than reduced lifespan.

 

[dirtyferret]

Slightly off topic, but buying a PSU based on 50% load is a waste of money.

agreed and not off topic; it's a minor factor as often the difference between 40%,50%, 60% or even 70% load (on a very good unit) is minor especially once the PSU is operating in a hot PC case and not being tested on a bench at room temperature. That nice efficiency bell curve marketing puts on the PSU box may look slightly different. I also never understood why people freak out if a peak load may hit 60%+ of a PSU rated watts. Now if you have a semi-fanless PSU or an aggressive fan curve and want to keep the fan off and/or noise down that is valid reason. If you are afraid of your PSU peaking past 60% of load, than you purchased/recommended the wrong PSU.

Well yeah - if the Seasonic 850W PSU is tripping its safeties with a ~615W rated load, then you have to wonder what damage is being done to PSUs by higher-power GPUs with PSUs that lack those safeties?

Hopefully it's nothing more sinister than reduced lifespan.

Not really positive what he was saying (hoping he clarify it) and techo-tuber and idiot are redundant terms. I know some AIB partners were pulling power outside of the seasonic PCIE power connector spec and seasonic had to send out adapters to fix the issue (some Asus GTX 970 units come to mind). From my own experience I've seen some (rather minor) Seasonic units be quirky compared to larger OEMs.

 

[Chrispy_]

agreed and not off topic; it's a minor factor as often the difference between 40%,50%, 60% or even 70% load (on a very good unit) is minor especially once the PSU is operating in a hot PC case and not being tested on a bench at room temperature. That nice efficiency bell curve marketing puts on the PSU box may look slightly different. I also never understood why people freak out if a peak load may hit 60%+ of a PSU rated watts. Now if you have a semi-fanless PSU or an aggressive fan curve and want to keep the fan off and/or noise down that is valid reason. If you are afraid of your PSU peaking past 60% of load, than you purchased/recommended the wrong PSU.


Not really positive what he was saying (hoping he clarify it) and techo-tuber and idiot are redundant terms. I know some AIB partners were pulling power outside of the seasonic PCIE power connector spec and seasonic had to send out adapters to fix the issue (some Asus GTX 970 units come to mind). From my own experience I've seen some (rather minor) Seasonic units be quirky compared to larger OEMs.

Definitely no need to get hung up on hitting exactly 50%. One of my systems at home is using a 530W PSU with a 3600XT and 5700XT which would need a 650W PSU if I followed my own advice to the letter, but it's not overclocked and it's closer to 50% than 100% so I'm satisified that it'll be absolutely fine for many years.

My rule of thumb for buying PSUs in the last decade or so has been to add the list* TDP of the CPU and GPU together and then double it. That gives you headroom to upgrade to a hungrier graphics card in the future and should mean that your peak load stays under 70% of the PSU's rating. (* - list TDP is absolute bullshit for Intel, regularly using 250W on a 125W TDP and still somewhat bullshit for AMD with their 65W CPUs using 88W and 105W CPUs using 142W)

 

[AleXXX666]

Definitely no need to get hung up on hitting exactly 50%. One of my systems at home is using a 530W PSU with a 3600XT and 5700XT which would need a 650W PSU if I followed my own advice to the letter, but it's not overclocked and it's closer to 50% than 100% so I'm satisified that it'll be absolutely fine for many years.

My rule of thumb for buying PSUs in the last decade or so has been to add the list* TDP of the CPU and GPU together and then double it. That gives you headroom to upgrade to a hungrier graphics card in the future and should mean that your peak load stays under 70% of the PSU's rating. (* - list TDP is absolute bullshit for Intel, regularly using 250W on a 125W TDP and still somewhat bullshit for AMD with their 65W CPUs using 88W and 105W CPUs using 142W)

The thing is, TDP for both cpu makers is stated WITHOUT turbo boost. That’s why on sh*tty base frequency we have pretty efficient cpu, but why the hell they can’t list turbo tdp, i don’t know. F*ckin marketing.

 

[blobster21]

I bought a Seasonic 80+ Gold 650w modular type. Do you think this psu can handle all power requirements at max?