Fan Wattage at different speeds?

[chuckk281]

Ivan had a customer question regarding fans in the c7000:

**************************************************************************

Wondering if anyone has handy the wattage used by each fan at each of the 3 different speed settings? (I believe our fans have a low, medium/optimal and high setting).

I have a customer wanting to compare our wattage differences between normal operations and high operation… as a comparison to IBM’s chassis.  Apparently there is a HUGE difference in wattage with the IBM chassis between the two different fan modes… and now customer is curious as to how the HP chassis/fan solution stacks up.

Thanx in advance!

****************************************************************************

Derek chipped in:

***************************************************************************

Ivan – actually our fans are controlled by a thyrister stack and are infinitely variable – the beauty of that is that each fan speed is modulated to meet the actual cooling needs of the enclosure at any point in time, rather than use pole settings to regulate fan performance as IBM do – that is a very inarticulate mechanism, and very ineffective as far as fan performance modulation goes.

Candidly – comparing fans themselves will tell him nothing – comparing overall chassis to chassis on a like for like (as close as IBM will ever get to being like BladeSystem) will provide a much more effective comparison. One simply cannot extrapolate values from single components and hope to see the overall savings we offer over IBM BladeCenter in every respect.

I would really discourage your customer from even going down that road – there are many early reports that tried to focus in on component level power draws, but finally in around 2007 the penny dropped through the industry that the whole system was the key, and that has been the measurement and comparison standard since then.

The other piece of guidance I might offer – please do not allow your customer to try and complete a ‘single blade in each enclosure type comparison’ – BladeSystem comes into its own at around 4 blades – this Sine Nomine (attached) report tells the story even though it is a few generations of processor out of date – the fundamentals still hold true, and in that time IBM have not delivered much innovation to change the overall findings – we on the other hand have continued to deliver key innovations especially in Dynamic Power Capping to this specific point.

Does this help?

*************************************************************************************************

Rick joined the conversation:

I was once told that power consumption by fans is essentially a

function of the cube of their RPMs.  I did some tests on an Integrity

system (rack mount) where I was able to learn some magic incantations

to alter fan speed without having to first alter temperature.  I'm not

sure where I stuck that data - of course that would have entirely

different constants from blade chassis fans...

Derek provided some context:

Actually you are almost correct - it is the change in the volume of a fan that results in a change of the power to the cube of the change in the volume - the initial power draw is a function of air density, fan impeller design, efficiency, casing efficacy, drive efficacy, and of course volume and pressure (incl both system and static pressure).

That sparked some additional comments from Rick:

I'm having a bit of trouble groking "change in the volume of a fan" - do you mean change in the volume of air moved by the fan?  If so, then I guess that at the same RPM I could see a power consumption change with a change in blade pitch (were we to have variable pitch fanblades), and RPM was being used as a shortcut for "increase in air volume"

Derek replied: 

Correct - volumetric displacement. In a simple axial flow fan assuming blade pitch remains unchanged air volume displaced is typically varied either by rotational speed change, or by variation in the blade length - the latter not unless one is looking at the system design from scratch, and looking perhaps at reducing fan blade tip speed and thus noise.

Fans in standard DL's are typically axial flow in impeller design, the fans in BladeSystem somewhat closer to a Frances Vane design, better able to deal with the rotational speed, volumetric displacement requirements, and of course efficiency in a comparatively compact form factor. Think of the impeller in a Bladesystem fan as similar to the turbofan at the front of a commercial Rolls Royce Trent engine, the impeller in a DL fan more typical of a standard household cooling fan.

Does this help make it clearer. We did some detailed C-7000 testing for a commercial customer here in EMEA if you want the data?

Rick responded: 

Yes, thanks.  Although I might have prefered a reference to GE or P&W :smileyhappy::)

And Monty did provide some data specific to the c7000 fans:

**********************************************************************************

I hope this helps in how the c7000 fans work. Questions? Any other input?

For the ten fan rule - the power allocated per fan is 50W DC.  This is the most efficient fan rule.

For the eight fan rule - the power allocated per fan is 80W DC.

So my 500W allocation for ten fans was correct, but the eight fan configuration only needs 640W power allocation.