How many watts do i need?

[Versti]

How many watts will i need for this rig:

 

MB: Gigabyte GA-AB350M-Gaming 3 (rev. 1.0) 

CPU: AMD Ryzen 5 1500x Box 

RAM: HyperX Fury 8GB DDR4 2400MHz HX424C15FB2/8 DIMM 

Storage: Corsair Force Series™ LE200 CSSD-F120GBLE200 120GB SSD and Seagate Barracuda ST1000DM010 1TB

GPU: Gigabyte NVIDIA GeForce GTX 1050Ti G1 Gaming 4G GV-N105TG1 GAMING-4GD 4GB GDDR5

Case: Fractal Design Define C(i know that the case doesn't matter)

[STRMfrmXMN]

8 minutes ago, Versti said:

How many watts will i need for this rig:

 

MB: Gigabyte GA-AB350M-Gaming 3 (rev. 1.0) 

CPU: AMD Ryzen 5 1500x Box 

RAM: HyperX Fury 8GB DDR4 2400MHz HX424C15FB2/8 DIMM 

Storage: Corsair Force Series™ LE200 CSSD-F120GBLE200 120GB SSD and Seagate Barracuda ST1000DM010 1TB

GPU: Gigabyte NVIDIA GeForce GTX 1050Ti G1 Gaming 4G GV-N105TG1 GAMING-4GD 4GB GDDR5

Case: Fractal Design Define C(i know that the case doesn't matter)

That system will draw around 200W under max load. A Corsair CX450 would be fine.

[Versti]

Just now, STRMfrmXMN said:

That system will draw around 200W under max load. A Corsair CX450 would be fine.

will i be able to overclock? like will a 450W PSU be able to handle giving power to my overclocked parts?

[Mira Yurizaki]

2 minutes ago, Versti said:

will i be able to overclock? like will a 450W PSU be able to handle giving power to my overclocked parts?

Yes. Assuming TDP is power consumption (I know it's not), at worst case, you could be pulling 65W. Overclocking by 10% theoretically only increases that by 10%, i.e., it adds 6W.

[System Error Message]

above 9000!

 

probably 300W max power use so at least a decent 450W PSU to handle peaks.

[turkey3_scratch]

2 minutes ago, System Error Message said:

above 9000!

 

probably 300W max power use so at least a decent 450W PSU to handle peaks.

Neh max load will probably be in the 160s.

[pyrojoe34]

8 minutes ago, M.Yurizaki said:

Yes. Assuming TDP is power consumption (I know it's not), at worst case, you could be pulling 65W. Overclocking by 10% theoretically only increases that by 10%, i.e., it adds 6W.

That's not how it works btw. The voltage-power relationship is not linear but has an exponential component which means hypothetically a 10% voltage increase can lead to a 20% power increase, then another 10% increase on top of that can lead to a 50% power increase. (these values are made up but the point is it is not linear.

[System Error Message]

Just now, turkey3_scratch said:

Neh max load will probably be in the 160s.

i've measured power use before when stress testing computers. GPU TDP is very accurate when you are stress testing. Ofcourse game power usage is way lower but you have to remember that there are peaks and max use.

CPU doesnt use much power, GPU is about 150W so add the entire thing together (CPU, motherboard, drives, fans) and you get about 300W total max use, and including for additional peripherals attached and for anything added in the future, 450W gives added space including for any peaks.

[Mira Yurizaki]

Just now, pyrojoe34 said:

That's not how it works btw. The voltage-power relationship is not linear but has an exponential component which means hypothetically a 10% voltage increase can lead to a 20% power increase, then another 10% increase on top of that can lead to a 50% power increase. (these values are made up but the point is it is not linear.

I'm going off Intel's derived equation of P = C * V^2 * f. So while yes, increasing voltage causes an exponential growth, frequency is many orders of magnitude higher than voltage. At 1.5v, increasing it to 1.6v is not going to have much of an an impact than going from 3,600,000,000 to 4,000,000,000 or whatever.

[Mira Yurizaki]

@pyrojoe34

 

Okay, nevermind, I did some math and for the same frequency, yes, the power does grow quite a bit. But if all you're touching is frequency then it's a linear relationship.

[pyrojoe34]

21 minutes ago, M.Yurizaki said:

I'm going off Intel's derived equation of P = C * V^2 * f. So while yes, increasing voltage causes an exponential growth, frequency is many orders of magnitude higher than voltage. At 1.5v, increasing it to 1.6v is not going to have much of an an impact than going from 3,600,000,000 to 4,000,000,000 or whatever.

Yes it will. Plug in some numbers to see:

 

I omitted C since I will assume it doesn't change for the sake of this point.

 

P = 1.5² * 3.6 = 8.1

P = 1.5² * 4.0 = 9 (this is linear 11% increase in freq leads to 11% increase in power)

P = 1.6² * 3.6 = 9.216 (6.7% increase in voltage leads to 13.8% increase in power)

P = 1.6² * 4.0 = 10.24  (11% increase in performance leads to 26.4% increase in power. Just showed this one since you would usually have to increase voltage as well to increase frequency to get a stable OC.)

 

If you want a less hypothetical example:

i7-6800k

Stock: 3.6Ghz (turbo), ~1.15V, 98W

OC1: 4.0Ghz (11% increase), 1.2V (4% increase), 119W (21% increase)

OC2: 4.4Ghz (10% increase from OC1, 22% from stock), 1.45V (21% increase from OC1, 26% from stock), 177W (49% increase from OC1, 81% from stock)

[turkey3_scratch]

20 minutes ago, System Error Message said:

i've measured power use before when stress testing computers. GPU TDP is very accurate when you are stress testing. Ofcourse game power usage is way lower but you have to remember that there are peaks and max use.

CPU doesnt use much power, GPU is about 150W so add the entire thing together (CPU, motherboard, drives, fans) and you get about 300W total max use, and including for additional peripherals attached and for anything added in the future, 450W gives added space including for any peaks.

We're both talking about a GTX 1050Ti, right? Or are you talking about something else? Also, instantaneous spikes are not talked about when speaking of "power consumption". Average power is what's important.

[pyrojoe34]

2 minutes ago, turkey3_scratch said:

We're both talking about a GTX 1050Ti, right? Or are you talking about something else? Also, instantaneous spikes are not talked about when speaking of "power consumption". Average power is what's important.

For estimating power cost sure. But for PSU requirements the peak is what is important. If the peak is higher than the PSU can handle, even if the average is not, then you will have issues.

[turkey3_scratch]

This is not true. The PSUs are made with a lot of tolerance to handle spikes. Average is what is important. Look at a GTX 1050Ti review, the reported power is average. It is not 150W like you say.

 

Also, I'm talking about peak instantaneous not peak average. What are you talking about?

[Beowulff83]

Although  corsair cx450 should be perfectly adequate, I would opt for a cx550 gives a little more flexibility if you decided to upgrade to more power hungry components latter on.  and $ isn't that much more.

[pyrojoe34]

10 minutes ago, turkey3_scratch said:

This is not true. The PSUs are made with a lot of tolerance to handle spikes. Average is what is important. Look at a GTX 1050Ti review, the reported power is average. It is not 150W like you say.

Again this all comes down to the PSU specs. Most are rated for both a peak output and a continuous output. The advertised usually (but not always) wattage refers to the continuous output. For example I have a PSU that's rated for 500W continuous 538W peak. Either way you want to keep your peak below the rated wattage to be safe, getting a PSU that's 50 or 100W higher than another in the same tier is usually just a few bucks more expensive and worth the safety margin IMO.

 

OP: A 450W PSU would be plenty for you and leave lots of headroom for OC and future upgrades. Since a 550W is usually just a few dollars more then you can also opt for that for even more future flexibility.

[turkey3_scratch]

HardOCP reviews are good for wattage. Here: https://www.hardocp.com/article/2016/10/25/msi_geforce_gtx_1050_ti_gaming_x_4g_review/11

 

The test setup is an I7-6700K @4.7Ghz and the MSI GTX 1050Ti. Gaming wattage measured from the wall without overclocking is 162W, with overclocking it is 175W. PSU used is the Corsair CX850M. Going by the review here http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story4&reid=427 efficiency is probably around 84.5%, making gaming wattage 138W for the whole system without OC or about 150W with overclocking.

 

The Ryzen 5 1500X will use a lot less power than the I7 overclocked heavily. So OP's wattage when gaming is probably going to be more around 125-130W.

[System Error Message]

anyone remember one of linus videos where he stuffed loads of GPUs with a PSU that was just enough? It tripped the protection.

You gotta worry about peaks, future upgrades and efficiency.

 

A PSU is most efficient when at a particular use, this is good on both power bill and voltage stability and heat.

Components can be changed or added and use more power.

Peak power that is more than the PSU's wattage can trip protections and blow fuses.

 

So if you can afford a better PSU go for it. Wattage isnt everything on the PSU though but as long as the 12V output is 90% of the rated watts at least and that the PSU has a clean output than it will be decent. Protection features are good.

Sure the gaming wattage is low, but software isnt predictable, games all differ and power use differ. This is why you use stress test and not gaming wattage as every game is different and some games can use more power.

 

So 450W is a good estimate to cover all 3 needs. Even peripherals like usb need power so devices that are attached will draw power too.