The physics and thermodynamics of the nMP and future desktops thread

[Gav Mack]

I bet this won’t be a IEEE standard anytime soon. Why? Look at the Macbook Air and the Ultrabooks. The latter are blatant ripoffs from Apple’s design. I bet Apple is not amused by this ripoffs (as you can see with the Samsung lawsuits).
So I guess, Apple would rather not see their design become a standard for everybody … just so that everybody can build cheap nMP lookalike clones with standard components.

I'll have a buck or pound on it happening within the next few years if the efficiency is proved beyond a doubt. The concept of a central thermal core is certainly not unique in industry, it's Apple's innovation in applying this design and they will no doubt try and protect their intellectual rights. However the market has a habit of taking that away from the dominant IP rights holder to a particular PC design, you only have to look at what happened to IBM thinking they had the hardware market sewn up with patents for the original PC letting Microsoft have the bum deal with the software.

Besides I don't expect a plethora of new Mac Pro clones, more like the central core concept adapted to fit other designs such as desktop and server enclosures. Apple will no doubt litigate against blatant rip offs but going after the rest of the market because it saves energy and improves efficiency I can't see them doing because it benefits everyone. Samsung definitely would try though!

 

[-hh]

The thermodynamics in the central cooling core of the nMP is straightforward: Air flow carries heat and hot air has a tendency to rise.

True, but the 'trashcan' is simply not tall enough to have a good enough temperature delta to get the convection rise - if one really wanted to rely on this, it would become a floorstanding PC that's 1.5m tall.

...The cylindrical shape eliminates the dead corners for the air flow in the rectangular shape...

To a certain degree, as it can generally be considered a good match for a fan of the same diameter.

However, the components within are still flat (we don't yet have curved topology CPUs or GPUs), so there's still going to be corners and dead spots - - see the Wankel engine design for a well-known example - - so the relative significance of its cylindrical shape is far from being a foregone conclusion that we can claim success on.

- -

Overall, the "design goodness" that I see in the nMP is that they appeared to have had maximized the diameter of their cooling fan for their overall package, which allows for slower rotation speeds for air mass moved which makes it quieter.

However, that's looking at the system on an individualized level, not at the system level. Because Apple also concurrently removed provisions for things like internal storage, additional PCIe cards and the like, the design choice has pragmatically forced there to be a much higher likelihood of a complete system now having additional external expansion chassis to be provisioned, which will invariably have their own thermal budgets to be managed...and with that, also an acoustic budget too.

YMMV on one's particular use case of interest, but all of these are very typically going to end up being provisioned with a smaller-than-nMP's diameter cooling fan, which means that whatever the nMP's big diamter fan design gained will be offset by higher noise levels from said small diameter fans on these externals...

...and since these fans in externals are probably not going to have as much of an intelligent feedback loop for their speed control, they'll probably just be set up to run at a fixed speed regardless of actual internal temperature...

...which basically means that acustically, I see a high risk that the net system level's noise level will be worse, not better. About the only major saving grace will be for those larger enterprise organizations who have the luxury of high performance remote data storage - - which simply means that the noise pollution is moved from the office cubicle down the hall to the Server Room, where its the IT Tech's workplace problem.


-hh

 

[Rich.Cohen]

Generally, the least reliable components in any piece of electronic equipment are the moving mechanical parts - disk drives and fans in particular. The more separate fans a piece of equipment has, the more likely it is that one of those components will fail and require shutdown and repair. So the fewer the number of fans the better, assuming one has done a careful thermal design that allows that one fan to cool all the heat loads.

Everything up to the last sentence is correct. Conclusion in the last sentence depends on more than the effectiveness of the fan design. Multiple fans can allow continued degraded operation after a single failure depending on overall design. The aircraft industry has a lot of experience with this issue with regard to the number of engines. Of course the penalties associated with engine failure in an aircraft are somewhat larger than those associated with fan failure in a computer.

 

[wonderspark]

I would have preferred the tube were designed larger, with an equally larger (thus even more quiet) fan, so that there was more space for things like CPUs, storage, and maybe even a PCIe slot or two. Perhaps 8 or 9 inches in diameter over 6.6". It would still be much smaller than the prior design, still be a mind-blowing tube, yet be more versatile and less likely to be knocked over so easily.

 

[OdinEng]

The aircraft industry has a lot of experience with this issue with regard to the number of engines.

This analogy is not typically applicable to consumer or commercial computers.

An aircraft with multiple engines is designed so that it can continue to fly in the case of a single engine failure, sometimes even multiple engine failures.

Most computers that have several fans have each fan dedicated to a single area of cooling: a CPU fan, a video card fan, a power supply fan, a case ventilation fan. If any one of these quit, that area overheats and it's game over. The reliability of a system like this is reduced by multiple fans because each one is a potential single point system failure.

 

[Gav Mack]

I am wondering if it's gonna get a buildup of cat hair... (I have to clean the grill on the front of my oMP about once every 2-3 weeks...)

Perhaps you could cut a Dyson HEPA filter to fit underneath

In seriousness a new design of desktop with this core and airflow coming primarily from the bottom perhaps fitting a removable filter will be a good idea to adopt, to avoid blowing particulates that can block the elements in the PSU, heat-sink and furthermore spouting out of the 'blow hole'.

 

[wildmac]

Perhaps you could cut a Dyson HEPA filter to fit underneath

In seriousness a new design of desktop with this core and airflow coming primarily from the bottom perhaps fitting a removable filter will be a good idea to adopt, to avoid blowing particulates that can block the elements in the PSU, heat-sink and furthermore spouting out of the 'blow hole'.

That really might need to be done.. one cat is a shedder.. and they've discovered that the shelf above the computer is warm.

And I can't be the only one that will have these sorts of problems with dust and other stuff.

 

[costabunny]

Perhaps you could cut a Dyson HEPA filter to fit underneath

In seriousness a new design of desktop with this core and airflow coming primarily from the bottom perhaps fitting a removable filter will be a good idea to adopt, to avoid blowing particulates that can block the elements in the PSU, heat-sink and furthermore spouting out of the 'blow hole'.

hmmm that give me an idea......


(rummages in cupboards for some colour changing leds and a smoke machine to mount the new MP on....)

 

[wildmac]

hmmm that give me an idea......


(rummages in cupboards for some colour changing leds and a smoke machine to mount the new MP on....)

What have I done...

 

[gr8tfly]

As I recall laminar flow is highly dependent of the cleanliness of the surface. I wonder what will happen as dust accumulates. The laminar flow will probably retard dust accumulation, but it can't stop it unless the flow is at lease occasionally at a considerable speed. I wonder if the nMP will come with instructions for regular cleaning?

True. In the case of a wing, an airfoil that's designed to be laminar does need to be smooth. But, though it's counterintuitive, some "roughness" designs actually do work to improve flow. On wings, it works by turning laminar flow into a turbulent boundary layer, delaying flow detachment to further aft (which has much higher drag than the turbulent boundary layer). I'm not sure how this applies to the nMBP's heatsink surface, but it could have an affect in areas that transition from one shape to another.

Wish I could remember where I noticed the mention of surface design in relation to the nMBP. But, I'd wager Apple's engineers looked very closely at optimal flow designs for heatsinks, plenums, intake ports, and exhaust (I'm thinking of this as the area where the heatsink shape would likely change before the fan). Speaking of which, there would have been extensive aerodynamic studies done for the custom fan blades - and since these are airfoils, I'd also bet the topic of laminar, boundary layer and BL separation were all considered (and many more aspects, I'm sure). I'm not a mechanical/aerodynamic engineer, but a commercial instrument pilot with SEL and glider ratings, many, many hours of competitive aerobatics, given talks on spins and spin recovery, and taught aerobatics. Between that, and my natural curiosity, I've learned a bit about aerodynamics. (Sorry, didn't mean to start a resumé. )

It'll be very interesting to see some third party analysis of the cooling system once the nMBP ships.

 

[Rich.Cohen]

True. In the case of a wing, an airfoil that's designed to be laminar does need to be smooth. But, though it's counterintuitive, some "roughness" designs actually do work to improve flow. On wings, it works by turning laminar flow into a turbulent boundary layer, delaying flow detachment to further aft (which has much higher drag than the turbulent boundary layer). I'm not sure how this applies to the nMBP's heatsink surface, but it could have an affect in areas that transition from one shape to another.

Wish I could remember where I noticed the mention of surface design in relation to the nMBP. But, I'd wager Apple's engineers looked very closely at optimal flow designs for heatsinks, plenums, intake ports, and exhaust (I'm thinking of this as the area where the heatsink shape would likely change before the fan). Speaking of which, there would have been extensive aerodynamic studies done for the custom fan blades - and since these are airfoils, I'd also bet the topic of laminar, boundary layer and BL separation were all considered (and many more aspects, I'm sure). I'm not a mechanical/aerodynamic engineer, but a commercial instrument pilot with SEL and glider ratings, many, many hours of competitive aerobatics, given talks on spins and spin recovery, and taught aerobatics. Between that, and my natural curiosity, I've learned a bit about aerodynamics. (Sorry, didn't mean to start a resumé. )

It'll be very interesting to see some third party analysis of the cooling system once the nMBP ships.

I'm also a pilot, commercial instrument rating SEL. The only glider I've ever flow was when I lost an oil line. Many airline jets have turbulence generators mounted on strategic locations on their wings. I recall that laminar flow wings are much more sensitive to ice build-up than other designs. (That may be a faulty memory.) If that's the case dirt & dust build-up could be an issue. Hopefully Apple's testing included a dirty room. One test the Army used to put rifles through was to dump them in a mud slurry and see if they still worked. Anyway a dirty environment could be an issue for the nMP.

 

[ActionableMango]

Less energy in = less heat to dissipate

I imagine the thermodynamics of the system are helped immensely by the fact that the nMP has a wimpy 450W power supply.

 

[LongSticks]

I would have preferred the tube were designed larger, with an equally larger (thus even more quiet) fan, so that there was more space for things like CPUs, storage, and maybe even a PCIe slot or two. Perhaps 8 or 9 inches in diameter over 6.6". It would still be much smaller than the prior design, still be a mind-blowing tube, yet be more versatile and less likely to be knocked over so easily.

Another first for the nMP haters......haven't seen the CofG for the nMP criticised. Good work!

----------

what have i done...

pmsl :d

 

[DigitalVT]

I've been waiting for this thread to appear, only because there was no threads on flight casing the nMP. So as you can imagine, the "trash can" in the normal upright position doesn't easily lend itself to 19" rack space!! Place it horizontally, then two side by side (allowing access to ports), would be more conducive to the space. But would this mess up the "thermodynamics", and create heat build up on the higher top edge of the horizontal "can"! Which convinces me more and more to case them vertically as presented, but to somehow utilise the spare space behind!? That's where the old MP towers were so handy, they didn't mind being on their side or vertically housed!!

 

[Gav Mack]

I've been waiting for this thread to appear, only because there was no threads on flight casing the nMP. So as you can imagine, the "trash can" in the normal upright position doesn't easily lend itself to 19" rack space!! Place it horizontally, then two side by side (allowing access to ports), would be more conducive to the space. But would this mess up the "thermodynamics", and create heat build up on the higher top edge of the horizontal "can"! Which convinces me more and more to case them vertically as presented, but to somehow utilise the spare space behind!? That's where the old MP towers were so handy, they didn't mind being on their side or vertically housed!!

It's going to need a total rethink of how thermal cores are stored in a rack for sure. I know a business outside of the IT industry who also use rack mount kit which generate enormous amounts of heat who are buying a nMP simply to tear it down and see how they've done it and adapt this design to fit their enclosures. Currently they are thinking about meshed shelves to host their units and using fan shelves at the bottom of the rack.

 

[haravikk]

the 'trashcan' is simply not tall enough to have a good enough temperature delta to get the convection rise - if one really wanted to rely on this, it would become a floorstanding PC that's 1.5m tall.

That's if they were relying entirely on convection, I think FrankHahn's point is still valid as the fan only has to make up for what the natural heat rising can't do.

In building my storage enclosure I've tried to mimic a similar principle by keeping the PSU at the bottom and making sure smaller components and cables are right to the sides of the case, leaving all of the empty(-ish) space towards the top of the case. Heat seems to rise into it just fine, at which point the fan just needs to spin lazily to help it to rise out of the case, and to let cool air come in through vents at the bottom of the case and between the hard-drives. Obviously my use case isn't quite the same as hard drives don't generate quite as much heat as a pair of GPUs and a CPU will, but the principle works really well; in fact I could probably run the fan slower, but it seems to cut off below about 400rpm (it's only quoted at 500-1000rpm so it's not that surprising).


My main wish from the Mac Pro is that we see more daughterboard-based CPU systems. Imagine a CPU on a card similar to a GPU, you just slot the CPU into a motherboard at one side, and the GPU at the opposite side, both facing inwards, forming a box. Just slap a nice big fan on the front, and another at the back and a single flow of air would cool both of those hot-running components. It would require the CPU card and GPU to both have heatsinks (or options) suitable for passive cooling to really get the most out of it, but for a compact system would be fantastic, especially if the ports that are traditionally soldered straight onto the motherboard were instead located on a separate board, as one of biggest problems I have with PC cooling is that thanks to the heaps of ports on a motherboard you just can't fit the same size fan at the rear of a case that you can at the front; this means the separate I/O board would also need to support the GPU's output being somehow routed to it.

It'd be a hard departure to handle though, as it just requires too many components to support it. Maybe if GPUs started supporting some kind of standard I/O header that ordinarily their PCI bracket (with built in ports) is connected to, simply remove this and connect it to your own I/O board and you can suddenly locate your connections anywhere you like.


It's certainly food for thought, as I'm a big fan of small form factor computing and like I say, I just don't like how relatively small rear exhaust fans always end up being on PC cases, as I always want the biggest fans that will possibly fit.

 

[wonderspark]

Another first for the nMP haters......haven't seen the CofG for the nMP criticised. Good work!

The world is too small and life is too short to hate. You could say I don't appreciate the new design as much as you do might, but calling me a hater is like calling Nelson Mandela or Siddhartha Gautama a hater.

 

[flat five]

I've had a quick search and not seen a thread yet.

there's some talk sprinkled throughout various different threads over the past few months..

one of the things which should definitely be brought into focus if the topic is thermodynamics of the nmp is the secondary piece of material in between the cpu/gpu processors and the heat sink.. there's more going on than "oh, it's just a normal heatsink with a different shape"..

-----------


here's a some stuff i wrote on it in this thread: (in the middle of a thermal paste conversation but still relates)
https://forums.macrumors.com/threads/1681183/




""
it's not just taking a wild stab at it and saying 'i bet no thermal paste because apple are wizards' (or whatever)

there's obviously a secondary plate of material in between the gpu/cpu and the thermal core.. you can either ignore it and claim 'there's no evidence of the nmp etetc' or do a little research to try to figure out what it is..

it appears to be a graphite heat spreader.. google it


there are lots of pdfs and research papers etc available through that search as well as evidence of large corps who already have graphite spreaders and/or films in their products.. such as panosonic:
http://semiaccurate.com/2013/08/15/panasonic-make-a-10-micron-thick-carbon-phone-heatsink/

intel research:
http://qats.com/cms/2010/05/03/inte...n-ats-thermal-management-technology-analysis/


i won't post all related links but here's a telling video:""

http://www.camnano.com

 

[gr8tfly]

... One test the Army used to put rifles through was to dump them in a mud slurry and see if they still worked. Anyway a dirty environment could be an issue for the nMP.

Mud slurry + nMP = "Oh The Humanity!"

We have a natural source of dust: our beloved Cockatoo. You should see the HVAC filter after a month! Planning on getting some pentalobe drivers for the rMBP and new MBA so I can periodically clean the fans, etc.

Glad to meet another pilot. "Dirty side down" as they say. (Although, we busted that one with our (my wife's a pilot too) aerobatics.). At least our Super Decathlon had an oil line vent routed out by the tail wheel - that made keeping the "dirty side" clean much easier.

 

[ScottishCaptain]

there's some talk sprinkled throughout various different threads over the past few months..

one of the things which should definitely be brought into focus if the topic is thermodynamics of the nmp is the secondary piece of material in between the cpu/gpu processors and the heat sink.. there's more going on than "oh, it's just a normal heatsink with a different shape"..

Look familiar?

That's a vapour chamber heatsink. High end GPUs have been using them for a while, it's basically just a giant flat heat pipe that distributes the heat from a central source across a larger surface area. Each side of the Mac Pro heatsink is essentially a giant vapour chamber. The only difference is that Apple has elected to paint the majority of theirs black.

There's no "graphite" spreaders or anything like that involved. It's just a vapour chamber.

-SC

 

[VirtualRain]

there's some talk sprinkled throughout various different threads over the past few months..

one of the things which should definitely be brought into focus if the topic is thermodynamics of the nmp is the secondary piece of material in between the cpu/gpu processors and the heat sink.. there's more going on than "oh, it's just a normal heatsink with a different shape"...]

Image

Look familiar?

That's a vapour chamber heatsink. High end GPUs have been using them for a while, it's basically just a giant flat heat pipe that distributes the heat from a central source across a larger surface area. Each side of the Mac Pro heatsink is essentially a giant vapour chamber. The only difference is that Apple has elected to paint the majority of theirs black.

There's no "graphite" spreaders or anything like that involved. It's just a vapour chamber.

-SC

Interesting... thanks for pointing this out. I was wondering if there were heat pipes buried in that HSF somewhere... this explains that away completely.

 

[flat five]

Image

Look familiar?

That's a vapour chamber heatsink. High end GPUs have been using them for a while, it's basically just a giant flat heat pipe that distributes the heat from a central source across a larger surface area. Each side of the Mac Pro heatsink is essentially a giant vapour chamber. The only difference is that Apple has elected to paint the majority of theirs black.

There's no "graphite" spreaders or anything like that involved. It's just a vapour chamber.

-SC

whatever it is, it's still worth discussing. but yeah, that looks similar to the thing in the nmp. do you have a picture which shows it assembled? I'm curious to see what type of bracket corresponds with those four holes

 

[wallysb01]

The only difference is that Apple has elected to paint the majority of theirs black.

Black does radiate heat better than any other color. Maybe they are counting the extra radiative heat loss and we should expect these things to be hot to the touch under load, especially at the top.

 

[flat five]

Interesting... thanks for pointing this out. I was wondering if there were heat pipes buried in that HSF somewhere... this explains that away completely.

fwiw, i do think it's a separate plate and not part of the core itself..

the core has a recess to accept this thing.. it's like this on all 3 sides

(blue arrow- ignore the green circles in the pic)

 

[flat five]

do you have a picture which shows it assembled? I'm curious to see what type of bracket corresponds with those four holes

the reason i was asking to see the bracket was because i noticed something peculiar with the gpu & cpu brackets which makes me believe they have a spring like action/tension control.. in the same thread linked to earlier, i drew up this:


"... in that the central ring will sit lower than the outer four flanges if sitting by itself on a flat surface.. like so:"

(exaggerated side view for clarity)

my first guess on that was to allow a person using a screwdriver to achieve proper tensioning without the need for a torque driver.. subsequent guesses were to allow the processor to actually be smashed into the plate as well as move with swelling/contraction.. (there's another guess which does those things as well as eliminates the need for thermal paste but i'll leave that out for now )

anyway, searched some images for vapor chamber heatsinks and saw this:

sort of reinforces the idea that the nmp cpu/gpu brackets are spring loaded..