Are people just massively overspeccing PSU‘s?

[PeteyPete]

Hey,

I am going for a mini-ITX build, planned parts are a 5600x and a 3070.

I looked around online and I have seen people recommend anywhere from 550 to 'at least 750' watts for similar builds and this is where I am confused somewhat: A 3070 has a TDP of 220 Watt, the 5600x a TDP of 65 Watt which would easily fit even a 400 Watt psu. Now yes, the 30‘ series cards do exhibit some weird power spikes, but even in Igor‘s lab‘s torture testing, the 3070 maxed out at ~305 Watt spikes iirc which would bring the Cpu+Gpu power draw to 370 Watts.

Even assuming the 5600x somehow uses double its rated TDP at 130 Watt, this would mean 435 Watt peak, thus, with a 500 Watt psu, leaving 65 Watt for the rest of the system. I am not very knowledgeable about this, but storage+fans or even a water pump would never utilise this kind of power right?

Lastly, even if we say the 3070 might spike to 350 and the 5600x consumed 142 Watt (which from my understanding is the max power draw on AM4) AT THE EXACT SIME TIME by ridiculous chance, all decent PSU‘s seem to be able to deliver about 50 Watt more than continuously for short bursts of time which would still keep the system running in this incredibly unlikely scenario.

 

Oh and even if there was not enough power, the PC would just crash right?

 

(On the basis of these thoughts I bought a 500 Watt (B-Tier on the PSU Tier list) power supply but am now questioning that decision — that’s why I‘m here)

 

[Master Singleton]

OP out of curiosity what B tier PSU did you actually buy? 

[da na]

Welcome to the forums!
Kinda like SSDs, it's not profitable to make the small capacities like 60-120gb anymore. If you're going to be making a power supply, might as well build it with decent components. Think how many more people would buy a $40 500w PSU than a $40 200w PSU, that's basically just the cost of the components inside. (I just made those numbers up but the point still stands.) There are of course exceptions but that is generally why

[Origami Cactus]

You are fine. The 750w PSU recommendations are for people who use intel toasters as a cpu, those pull over 200 watts by themselves, so you can see why a 500w psu wouldn't work out with that system.

The 5600x doesn't use anywhere near the AM4 limit of 140w, it is more like 75w max, so you will be completely fine.

[AndreiArgeanu]

Firstly, generally speaking PSU's run best at 80% capacity, that's about where they're most efficient. If you run a PSU constantly at 100% it's life could be significantly shortened although it'll still probably last outside it's warranty period. Second if you ever decide you want to upgrade, either to a better CPU or GPU you may end up having difficulties finding an upgrade. NVIDIA 4000 series is expected to require even more power than 3000 series.

 

Additionally 65W might no be enough for the rest of the system, memory can draw at least 3-5W per stick or even more if you're running XMP even more possibly. A motherboard can use about 30-50W or so of power, a single SSD about 8W, and hard drives a bit more than that, and a standard fan about 2W, with RBG ones drawing even more.

 

You'll probably be fine as long as the PSU is properly cooled and you don't live in a really hot climate. Realistically the CPU is probably going to stay within it's 65W power package. Worst case scenario you have to do some under volting.

 

23 minutes ago, PeteyPete said:

Oh and even if there was not enough power, the PC would just crash right?

Yes and no. Generally speaking PSU's will be able to output slightly more power than what's written on the box before any protections kick in, nonetheless most quality power supplies have protections for too high power draw and temperatures, so nothing will explode, unless you bought a gigabyte GP P750GM.

[TukangUsapEmenq]

In engineering, usually some people would just over spec-ing, like (these are analogies, but close enough still), how civil engineer adds additional material for the sake of 'some additional weight' (like, a single story home but you want to build the 2nd or third story) for the future, or electricians who use bigger-size cable on the house wiring simply because if the power's too high on the smaller cable it could heat up and, goodbye to your home, to a dragon. 

 

It adds an additional layer of safety, and reduces the amount of worrying, and of course, simply improving the reliability since it won't run like on 100% all the time. In this case, it's exactly like you won't run your car engine revving close to it's redline at all times, right?

 

21 minutes ago, PeteyPete said:

with a 500 Watt psu, leaving 65 Watt for the rest of the system. I am not very knowledgeable about this, but storage+fans or even a water pump would never utilise this kind of power right?

Afaik, PSU would have the best power efficiency when they actually have the power load on 60-80% of the rated power (correct me if I wrong, I don't remember exactly how much). And either, you won't blow a component or two on your PSU if in the case of power spiking, because it's still below the rated maximum power. I would be a lot comfortable if my PSU got more headroom than to simply hanging around on it's max. rated power, simply because it won't explode (or simply, slowly damaged) because of sudden overload that we wouldn't know somehow if it even happens.

[Alex Atkin UK]

17 minutes ago, AndreiArgeanu said:

Firstly, generally speaking PSU's run best at 80% capacity, that's about where they're most efficient.

Actually Gamers Nexus started their testing and seem to find 40% is the sweet-spot for optimum efficiency though yes, to avoid paying insane prices for overkill PSUs then 80% is generally okay.

 

Most if not all good PSU vendors will put the efficiency curve on their box (annoyingly not always on their website, though retailers may show it).

 

There's also the fact 30x0 series card tend to have very spiky power loads, so while the average may seem fine it can be good to have a higher spec PSU to avoid potentially triggering its over-current protection.  Plus in theory a less stressed PSU will have a longer lifespan, be quieter, produce less heat and you won't need to replace it again if you upgrade to a beefier GPU.

A typical example of an efficiency curve.

[SolarNova]

As mentioned above, its usually down to efficiency and PSU longevity.

Running a PSU at 40-60% will keep it in the most efficient zone, and allow for the best longevity.

 

My old Seasonic Platinum 1kw lasted about 10 years before it started to fail (random crashes), and it was run well below its maximum the entire time.

If i hadnt overspeced my PSU choice I likely would have had to replace the PSU years sooner.

[Dogzilla07]

@Alex Atkin UKThat is not a typical example of an efficiency curve. They differ wildly from PSU platform to platform. @SolarNovathe amount of efficiency for most modern PSUs is irrelevantly small in the "perfect range" 40-60%. While leaving some room is beneficial noise wise depending on the fan curve, in very small form factor situations, or when using 500w TDP GPUs of latest and next-gen on older PSUs.

 

Other than that there is no need to leave more than 20-30% extra room. Longevity is not directly correlated with efficiency outside two fringe situations:

 

1. Very small form factor (and even here it's more fan noise issue)

2. after 12-15 years of usage.

[Alex Atkin UK]

2 minutes ago, Dogzilla07 said:

@Alex Atkin UKThat is not a typical example of an efficiency curve. They differ wildly from PSU platform to platform. @SolarNovathe amount of efficiency for most PSUs is irrelevantly small in the 40-60%. While leaving some room is beneficial noise wise depending on the fan curve, and in very small form factor situations, other than that there is no need to leave more than 20-30% extra room. Longevity is not directly correlated with efficiency outside two fringe situations:

 

1. Very small form factor 

2. after 12-15 years of usage.

Its typical of a good PSU and its also a typical example of what an efficiency graph looks like.

 

I never said every PSU has the exact same curve, in fact I explicitly said you need to look on the box or retailer website.

 

Longevity is inherently related, unless they over-spec the components of the lower models (sometimes they're identical just artificially limited) as the hotter it runs, the quicker components degrade.  Thus why I said "in theory" and I've owned plenty of PSUs for longer than 15 years (though these days I would prefer to replace within 10).

[Dogzilla07]

10 minutes ago, Alex Atkin UK said:

Its typical of a good PSU and its also a typical example of what an efficient graph looks like.

 

I never said every PSU has the exact same curve, in fact I explicitly said you need to look on the box or retailer website.

It simply is not "typical" the line for % difference is flatter towards the right. For most modern Gold/Platinum PSUs there is no effective practical difference (<1.5-2%). Especially when looking at Cybenetics testing, as most modern Gold PSUs are platinum in everything but the name.

 

We agree in essence you mentioned 40% being mathematical sweet-spot, and up to 80% being typically ok (as you said) lines up exactly with I wrote. So we agree in essence.

 

10 minutes ago, Alex Atkin UK said:

as the hotter it runs, the quicker components degrade. 

Nope, they don't run hotter+degrade if efficiency is slightly worse in the 1-2% range I mentioned. it's too little of a difference. The fan just runs more noisy. Most PSUs are tested at 50c, some at 40c (Seasonic, and then they drop 80% effective wattage at 40c). What you're saying does not apply outside FlexATX, SFX psus (and even there not 1:1).

[Rexper]

The RTX 30 series cards exhibit very large power spikes. These spikes are so quick that most reviewers don't even see them with their testing equipment. They are also so high they trip protections on some PSUs (especially <750w units.. Though, this is more an issue for the RTX 3080/3090.

 

Your PSU would probably work fine. If not, protections would safely shut it down. You can worry then.

 

The efficiency and lifetime argument is a myth. The price you spend to get a higher wattage PSU so it's running at "optimal efficiency" with your PC isn't even covered by the efficiency savings.

But those myths still float around which is why we see such high wattage recommendations.

[PeteyPete]

1 hour ago, Master Singleton said:

OP out of curiosity what B tier PSU did you actually buy? 

The BeQuiet SFX-L Power 500

[PeteyPete]

2 minutes ago, Rexper said:

The RTX 30 series cards exhibit very large power spikes. These spikes are so quick that most reviewers don't even see them with their testing equipment. They are also so high they trip protections on some PSUs (especially <750w units.. Though, this is more an issue for the RTX 3080/3090.

 

Your PSU would probably work fine. If not, protections would safely shut it down. You can worry then.

 

The efficiency and lifetime argument is a myth. The price you spend to get a higher wattage PSU so your PC fits in the "optimum efficiency point" isn't even covered by the efficiency savings.

But those myths still float around which is why we see such high wattage recommendations.

I know, that’s why I looked up Igor‘s review, he does incredibly in depth analyses from my understanding, and yes — The 3070 does spike to way higher than TDP, but still only to 305 Watt.

[Dogzilla07]

@RexperThe Spike issue fiasco was mostly Seasonic first generation focus units, which has been solved now, there's a few other PSUs that had the problem, it's written on the cultist tier list:

 

https://cultists.network/140/psu-tier-list/

 

Now, just to clarify what I'm saying is as of today, for the parts out as of today, and in the past. How things are gonna pan out with ATX 3.0 Gen 5 spec and next Gen 500-600W GPUs and MCM GPUs, who knows.

[Alex Atkin UK]

Just now, Dogzilla07 said:

It simply is not "typical" the line for % difference is flatter towards the right. For most modern Gold/Platinum PSUs there is no discernible (<1.5-2%). Especially when looking at Cybenetics testing, as most modern Gold PSUs are platinum in everything but the name.

I literally picked the graph of my own PSU as it was easy to find (and its a popular model) and you're being all pedantic over the word "typical" when all I was pointing out is that they are more efficient around 40-50% and this is what a graph looks like.

If most PSUs follow that graph or not is irrelevant as I said to check any specific models graph.

[Rexper]

10 minutes ago, PeteyPete said:

I know, that’s why I looked up Igor‘s review, he does incredibly in depth analyses from my understanding, and yes — The 3070 does spike to way higher than TDP, but still only to 305 Watt.

Yeah, they're very in depth.

His test shows 370w spikes (1-5ms in length) on an aftermarket RTX 3070.

Even if everything in your computer peaked in power at the exact same millisecond, your PSU would still handle it.

[dilpickle]

It's difficult if not impossible to spec out the exact amount of watts you need for a system. You want some breathing room and you never know when you might add a new component. PSUs are also most efficient anywhere from 40-80% usage so you want to stay in that range.

 

Also a PSU is one component that can last you thru multiple builds so its good to get more than you need right now. (Although we happen to be in a transition to an updated standard right now)

[Alex Atkin UK]

14 hours ago, dilpickle said:

Also a PSU is one component that can last you thru multiple builds so its good to get more than you need right now. (Although we happen to be in a transition to an updated standard right now)

It looks like the data channel is optional though, given at least some 3090 Ti cards are using the new connector with 3 PCIe adapters.

 

Presumably its just a safety precaution going forward so people can't overload their PSUs.

[Master Singleton]

On 4/3/2022 at 11:48 PM, PeteyPete said:

The BeQuiet SFX-L Power 500

The only downside of this particular PSU is the smallish heatsink BeQuite used and its not particularly efficient at passively cooling the PSU and at higher loads the fan kicks in and might reduce the lifespan of the fan bearings if you plan on running the system at higher loads for extended periods,