Subscriber Discussion

How To Find Equipment Heat Output?

Avatar
Simon Lambert
Nov 16, 2017
IPVMU Certified

Equipping a new server room, the building services consultant wants to know how much cooling he'll need to add to the room. So, he asked me for the heat output of the equipment and the temperature it should be kept at. The new video server, encoders, switches, display drivers, etc. give a total of 5,000W electrical power. The cooling designer says (rule of thumb: 50%) that this means 2500W of heat into the room. Does that ring true?

He is a cooling designer. I'm not, so I bow to his greater knowledge. Nonetheless, my ancient & rusty degree in physics give me the following idea:

energy in = energy out.  (per second, gives power)

Electrical power in = power as heat + sound + light + kinetic + potential energy.

So, does his rule of thumb mean here that 2500W is being turned into heat, and the remaining 2500W is becoming sound, light and movement? If so, that's going to be a dangerous rack to stand near!

All I can imagine in the server rack are spinning HDDs, humming fans, flashing LEDs. Those surely don't take 2500W, so is more than 50% turning into heat? After all, billions of switching transistors do get rather hot.

Regards,
Simon

UI
Undisclosed Integrator #1
Nov 16, 2017

Many products list their temperature output in BTU’s.  I’ve had to provide this before. 

Avatar
Michael Silva
Nov 16, 2017
Silva Consultants

The traditional practice is for the designer of the security/surveillance systems to provide the "faceplate" wattage ratings of all of the equipment in the room to the designer of the HVAC systems (usually a mechanical engineer) and let him or her calculate the actual heat load. This sounds like exactly what you have done.

Trying to calculate the actual heat load in advance is nearly impossible because there are so many variables and the information provided by the equipment manufacturers is not always precise. When using something like a POE switch or lock power supply, some of the power is dissipated in the equipment room, some in the cabling to the field device, and some at the field device itself. Also, the actual power used by most equipment in our industry is almost always far less than the wattage listed on the cut sheets. Because of all these variables, most engineers use rules-of-thumb that error on the side of caution rather than trying to come up with a precise number.

(2)
(1)
Avatar
Simon Lambert
Nov 17, 2017
IPVMU Certified

Thanks, Michael. That all accords with my experiences too.

Avatar
Brian Rhodes
Nov 16, 2017
IPVMU Certified

The new video server, encoders, switches, display drivers, etc. give a total of 5,000W electrical power. The cooling designer says (rule of thumb: 50%) that this means 2500W of heat into the room. Does that ring true?

It would be nice if everything was linear, but it isn't in reality.  Power inefficiency has as much to do with raw voltages being inputted, how much is actually being used, and how they are transformed (or even rectified) for use. 50% efficiency seems low, even for old equipment.  

In terms of cooling, exchanging the heat with ambient air is an option if enough volume of ambient air is available in plenums. Likewise, cooling ducts in the rack itself (often called Close-Coupling) is more efficient than lowering the gross temp in an entire room.

 

(1)
Avatar
Simon Lambert
Nov 17, 2017
IPVMU Certified

Thanks, Brian. The mechanical guy is adding air-con to the room to shift the warm air.

Regarding efficiency, I'm curious that a much higher percentage isn't turned into heat.  This is because of energy conservation. The electrical power doesn't appear to be turned into a great quantity of sound, light or movement. Nor is it leaving the room to power other electrical devices. So, heat seems to be the only remaining form.

I agree that manufacturers' specs do very likely overstate the normal running rate.

U
Undisclosed #2
Nov 16, 2017
IPVMU Certified

So, does his rule of thumb mean here that 2500W is being turned into heat, and the remaining 2500W is becoming sound, light and movement? If so, that's going to be a dangerous rack to stand near!

Here’s a key factor: Are the switches POE or not?

If they are POE then counting half of the POE budget as heat locally lost will lead to overestimating the heat in the server room.

This part of the power coming into the room, (with some loss due to step down), is really just going back out to the ultimate devices.

And we all know how hot some cameras get...

 

Avatar
Simon Lambert
Nov 17, 2017
IPVMU Certified

Thanks, U#2. No PoE switches sending their power out of the room to other devices.

UI
Undisclosed Integrator #1
Nov 17, 2017
(1)
Avatar
Simon Lambert
Nov 17, 2017
IPVMU Certified

Thanks, U#1. Interesting. They seem to say that the input power to the server is what should be used in the calculation of peak heat removal. In their example, 754W into the server PSU gives us 2577 BTU/hr to handle in the room.

So, pretty much what I thought, is it not? Their example suggests that electrical power all turns to heat. Back to an earlier point; what other forms could it reasonably be found to change to? In a server, surely heat is by far the majority.

I'd love to hear how that's wrong! ;)

Avatar
Kevin Nadai
Nov 20, 2017

Their example suggests that electrical power all turns to heat.

Correct. All electric consumed in that room turns to heat, either in that room or someplace else in the case of PoE.

Both sound and light dissipate into heat when they strike materials such as air and objects in the room. The same is true of kinetic (moving) energy, which also converts to heat through friction.

The cooling guy's 50% deduction is not for heat, light, or motion vis-à-vis heat. Rather, he is assuming your power supplies will only run at 50% input/output.

I'm not a scientist or a real engineer, so I am sure someone else in this thread can explain it better than me.

Avatar
Simon Lambert
Nov 20, 2017
IPVMU Certified

Thanks, Kevin. This does beg the question that if, say, an 800W PSU is only 50% efficient, does 400W simply go to heating up the PSU (and therefore room) before the remaining 400W goes into the machine to do its intended work?

U
Undisclosed #2
Nov 20, 2017
IPVMU Certified

This does beg the question that if, say, an 800W PSU is only 50% efficient...

I think Kevin was saying at 50% utilization, not efficiency.  Modern PSU’s operate at much higher efficiencies than 50%.

But to your point, whatever that number is, that would be the amount of waste heat generated by the PSU.  

(1)
(1)
Avatar
Kevin Nadai
Nov 21, 2017

Well said.

 

(1)
Avatar
Simon Lambert
Nov 21, 2017
IPVMU Certified

Thanks. I go along with all of this. The 50% utilization concept makes most sense.

MB
Mark Bottomley
Nov 27, 2017

The face plate power is the maximum possible used - often only seen at start-up. The actual is usually far less as CPUs go idle, or at least avoid hitting their frequency boost, computers do not have the maximal power draw configuration (less RAM, fewer drives, SSDs vs. spinning rust, CD/DVD idle). 50% sounds like a fair de-rating. Even PoE switches that can generate outputs of 500-700W ship most of it outside the room - only their base power plus conversion losses remain in the room. Even then, rarely is all the power required as a PoE maximum output (not PoE+) is 12+W usable and ~15W generated, but things like indoor dome cameras only want ~5-7W so the remainder is not used, not demanded, not converted to heat somewhere. Note also that any UPS will convert ~5-10% of power passing through to heat. In-line UPSs do this with losses in AC/DC/AC conversion and out-of-line UPSs do it by overhead and charging batteries that most of the time are already full.

The only way to confirm what is actually being used is to measure the current. A 3-prong cheater cord with separated conductors allows you to put a current sensing clamp over a single conductor. Some UPSs and distribution panels may also give you the average and momentary and actual power draw.

Depending on where you are located and heating demands, 2.5kW of heat (not a lot - 2 hair dryers on full) could be dumped into the building and let the room run at building ambient as part of the air circulation. Now only an exhaust fan is needed to move the heat out of the room.

This sounds like a small server room - 1-2 racks max. Keeping it at normal room temperatures like 70F should be fine.

 

(1)
Avatar
Simon Lambert
Nov 27, 2017
IPVMU Certified

Thanks, Mark. That all makes good sense.

jw
jim warner
Nov 27, 2017

I like Mark's Answer. The only thing to add is that a meter that actually measures power consumption for 120 V devices is not a big ticket item. You can find out for yourself what the heat load will be by measuring running Watts.

https://www.ebay.com/itm/P3-KILL-A-WATT-Power-Usage-Voltage-Meter-Monitor-P4400-NEW/380172111329?epid=623072101&hash=item5884060de1:g:81YAAOSw5dNWpjhA

You want to measure Watts and not VA to get a conversion to heat. For many devices, they will be the same. But some wall plug size power bricks will be the exception to that. Small power supplies often are less efficient than larger ones.

 

Avatar
Simon Lambert
Nov 28, 2017
IPVMU Certified

Thanks, Jim. I have one of those too. Surprising what numbers we see sometimes.

This server room is at the design stage, so nothing to measure yet. However, it does prompt the idea of taking these power readings at the Factory Acceptance Test before delivery to site, while the mechanical guys might still make alterations.

U
Undisclosed #2
Nov 28, 2017
IPVMU Certified

All I can imagine in the server rack are spinning HDDs, humming fans, flashing LEDs. Those surely don't take 2500W, so is more than 50% turning into heat?

To be sure, the spinning, humming and flashing ultimately end up as heat in the room as well.  (If there are no windows (for the LEDs)). 

On the other hand, if the server room contains racks of switches (MDF?), then some of the power attributed to that equipment will be actually be dissipated in heat outside the room, due to voltage drop on the ethernet cables.  

 

(1)
U
Undisclosed #3
Nov 28, 2017

Many data sheets will have BTU/hr listed. I typically provide this to mechanical engineer when adding my equipment in data closets.

(1)
New discussion

Ask questions and get answers to your physical security questions from IPVM team members and fellow subscribers.

Newest discussions