Real Men Use Butterfly Heatsinks: Apack ZeroTherm BTF80&

Devonavar · Post by **Devonavar** » Wed Aug 02, 2006 11:04 pm

Real Men have Butterfly Heatsinks: The Apack ZeroTherm BTF80 & BTF90

andyb · Post by **andyb** » Sat Aug 05, 2006 5:26 pm

My first impression of these heatsinks was "you have got to be kidding.!!!".

The performance is quite astonishing though, only beaten by the Ninja at low noise levels.!!! Thats quite a boast.

Could the performance be partly due to the heatpipes being "upside down", most U shaped heatpipes are sealed at the top end of the heatsink, these seem to be upside down, so they must be sealed in the base.!!!

If that fan was easily replaceable it would be an instant hit, even if people couldnt bear to have a butterfly in their PC, they would be taken in by the performance and noise.

I cant say I disslike the Butterfly shape, and I am not so macho that I would refuse to buy one, however I already have a Ninja + Nexus, so I already have a good excuse

Good review BTW.

Andy

GamingGod · Post by **GamingGod** » Sat Aug 05, 2006 5:47 pm

Yea Im a little curious myself. Could someone just remove the stock fan and tie a nexus onto the side of that thing and retest to see if it could beat the ninja?!?

jaganath · Post by **jaganath** » Sat Aug 05, 2006 6:16 pm

The larger the fin surface area, the better the ability of the heatsink to transfer (by radiation) the heat from the CPU into the air.

Convective heat transfer is also enhanced by larger surface area. Radiative heat loss is probably quite low at these temperatures.

Tibors · Post by **Tibors** » Sat Aug 05, 2006 7:10 pm

jaganath wrote:Convective heat transfer is also enhanced by larger surface area.

Yes, it is per se, but not necessarily in a design like this, because the bigger fins increase the airflow resistance, which in turn hinders the convective heat transfer. Without decent modeling you can't tell which one wins.

Felger Carbon · Post by **Felger Carbon** » Sat Aug 05, 2006 10:32 pm

Aaagghh! No information on which way the fan points on a K8 system: at the PSU (bad) or the exhaust fan (good)! Marvelous review, otherwise.

SebRad · Post by **SebRad** » Sat Aug 05, 2006 11:33 pm

@Felger Carbon, if you look at the picture of the K8 bracket on bottom of page 1 it looks to me like they've thought of that and it can be screwed on either way! The s775 bracket has a rectangular hole in it (but screws to motherboard are symmetric so can be rotated 90Â°) where the hole in the K8 clip is "cross" shaped, like two rectangles rotated 90Â° to each other.
Of course I could be completely wrong!
Seb

jaganath · Post by **jaganath** » Sun Aug 06, 2006 2:12 am

Yes, it is per se, but not necessarily in a design like this, because the bigger fins increase the airflow resistance, which in turn hinders the convective heat transfer. Without decent modeling you can't tell which one wins.

My point was that convection should have been mentioned as well, as it also clearly benefits from increased surface area for heat exchange. I thought it was the closeness of the fins that impacted on airflow resistance, not the size of the wetted surface.

Bluefront · Post by **Bluefront** » Sun Aug 06, 2006 3:29 am

I've always wondered what the optimal distance between fins measured for a low airflow setup. Get the fins too close and airflow is restricted. Spread the fins too much and there is too little metal for the heat transfer. (This is true for any heatsink of a fixed overall size).

IMHO.....the Ninja strikes a perfect balance between the two. Don't know about this one.

BTW.....The children at Anandtech are having a good time laughing at this butterfly design, calling it a heatsink for girls or kids. The thermal results should prove otherwise.

sanse · Post by **sanse** » Sun Aug 06, 2006 6:20 am

andyb wrote:I cant say I disslike the Butterfly shape, and I am not so macho that I would refuse to buy one, however I already have a Ninja + Nexus, so I already have a good excuse

a big drawback of the ninja with a fan on it is the fact it blocks 2 memory positions. this butterfly doesn't.

IsaacKuo · Post by **IsaacKuo** » Sun Aug 06, 2006 7:13 am

I must concur with Jaganath's original assessment, but I'd go even further. The transfer of heat from the heatsink to air via radiation is insignificant compared to the conduction transfer.

Air is largely transparent to thermal infrared radiation--that's how come thermal cameras work. If air absorbed thermal radiation much, then using thermal cameras would be like looking through thick fog. In reality, thermal imaging can see THROUGH fog.

Electric radiators work by emitting thermal radiation directly at the user. Again, this only works because air is largely transparent to the radiation.

Regardless of fin spacing, the vast majority of heat radiation is going to pass right through the nigh transparent air and hit the next fin. The only significant thermal radiation losses are through the outer surfaces--and it hits other components inside the computer rather than getting absorbed by the transparent air in between.

The comment about radiation in page 2 is simply a mistake that should be corrected. The fin area affects how much heat is transfered by thermal conduction, not radiation. Fin spacing and total fin area has no significant effect on radiation losses--it's the total "outer" surface area which affects radiation losses (roughly, imagine a box within which the heatsink fits--the radiative losses are the same as for that solid box).

IsaacKuo · Post by **IsaacKuo** » Sun Aug 06, 2006 7:19 am

andyb wrote:Could the performance be partly due to the heatpipes being "upside down", most U shaped heatpipes are sealed at the top end of the heatsink, these seem to be upside down, so they must be sealed in the base.!!!

I think the performance may be due to the fact that the doubled up heat pipes give twice the transfer area between the pipes and the fins, without resorting to a doubled layer of heatpipes at the base (like the Ninja). If this concept were scaled up to the size of the Ninja, then I guess it could outperform it.

I really love this design, and the lack of fan swappability is of no concern to me since I'd remove the fan regardless. None of my CPU heatsinks have fans; the ones with forced airflow use ducts or partitions.

Tibors · Post by **Tibors** » Sun Aug 06, 2006 7:47 am

jaganath wrote:I thought it was the closeness of the fins that impacted on airflow resistance, not the size of the wetted surface.

This is where you go wrong. With bigger fins, the distance the air has to travel between them becomes longer. This increases the resistance.

If you have trouble picturing this, then take a look at a tube carrying a fluid. Make the tube narrower, then the resistance increases. Make the tube longer, then the resistance increases too.

Post by **MikeC** » Sun Aug 06, 2006 9:08 am

Imprecise editor's note -- "by radiation" -- removed.

Also, clips allow the HS to be mounted with the fan pointing in any of four directions -- with either K8 or 775. This is now clarified in the review text.

alfred · Post by **alfred** » Sun Aug 06, 2006 9:59 am

As innovative as the heatsink design may be, I can't figure how Apack engineers managed to conduct so much fan vibrations to the metal structure, resulting in countless frequency pikes in the 1000-4000 Hz range; this also gives us are a perfect example of how misleading dBA measurements can be when analyzing computer parts. Fortunately the article and the recordings both make it very clear. Thanks Devon for this excellent review!

Howard · Post by **Howard** » Sun Aug 06, 2006 10:31 am

I wonder how performance is affected by such small radius pipe bends... If they bent it less tightly, there would be more "hot" surface area too.

mattthemuppet · Post by **mattthemuppet** » Sun Aug 06, 2006 3:06 pm

Tibors - I think you meant to say jaganath was partly correct rather than wrong, no? Afterall, you confirmed what he said in your follow on comment, though you added distance travelled as well..

I think the effect of fin spacing + depth on impedence to airflow and heat transfer also has something to do with the boundary layer of dead air that covers each fin. In fluid dynamics (vaguely remembering from what my brother told me), whenever you move a fluid (air, water etc) over a surface there's always a thin layer of stationary fluid at the surface. [puts condescending hat on] If you have trouble picturing this, take a look at a tube carrying fluid. The fluid moves fastest at the centre and slowest (or not at all) at the outside. Make the tube narrower and you need a higher flow rate to overcome the drag of the slow/ stationary fluid at the boundary. Make the tube longer for the same diameter and you have the same result [/takes condescending hat off]. I'm no expert (I'm a biologist thankfully, I leave fluid dynamics to my brother), but I'm pretty sure it works something like that. In the context of fin spacing + depth, it's a trade off between surface area to dissipate heat and impedence of the fins, which in turn is balanced with the speed of the fan pushing the air through/ over the fins.

Howard · Post by **Howard** » Sun Aug 06, 2006 6:33 pm

I wonder if dimpling the fins would increase the cooling efficiency...

dhanson865 · Post by **dhanson865** » Sun Aug 06, 2006 7:47 pm

Heatsinks Compared at ~20 dBA@1m

Note the prices are based on buying in the US and assume the cheapest low speed 92mm or 120mm fan I could specify when searching prices and are then rounded to a $5 point. Very rough math that would change for those with spare fans laying around or those in other countries.

Even though the prices I filled in are based as specified above; the noise, voltage, Â°C/W MP are taken straight from the SPCR reviews.

Code: Select all

Heatsink/Fan                             Noise Fan V  Â°C Rise Â°C/W MP Weight
APack ZeroTherm BTF80             (   )  19    5      28      0.28    458	
Spire Verticool II SP601B3        ($35)  20    5      48      0.48    350?
Thermalright XP-90, Nexus 92      ($35)  20   12      22      0.29   ~460  (360+fan)
Thermalright XP-120, Nexus 120    ($45)  20    9      27      0.27    493* (370+fan)

APack ZeroTherm BTF90             (   )  19    5      26      0.26    678	
Scythe Ninja, Nexus 120           ($50)  20    9      22      0.22    788* (665+fan)
Thermalright Ultra-120, Nexus 120 ($55)  20    9      19      0.19    868* (745+fan)

* http://www.nexustek.nl/120mmcasefan.htm lists the Nexus 120mm fan at 123g
http://www.nexustek.nl/92mmcasefan.htm lists 98.5g
http://www.nexustek.nl/80mmcasefan.htm lists 85g

some recent Stock AMD heatsinks are

Code: Select all

Heatsink/Fan                             Noise Fan V  Â°C Rise Â°C/W MP Weight
Ajigo MF064-074 (70mm fan no pipes)(   ) ??    ?      ??      0.??    373
AVC Z7U7414001 (70mm 4 heat pipes) (   ) ??    ?      ??      0.??    440
AVC Z7U7414002 (70mm 4 heat pipes) (   ) ??    ?      ??      0.??    440?
Unknown (80mm 4 heat pipes,$20+fan)($25) ??    ?      ??      0.??    ???

The unknown stock cooler is $20 shipped from more than one source but you'd have to do a fan swap. Probably fairer to call it $25 or $30 when comparing to other complete/quieter solutions.

Also ran heatsinks that don't keep up. Close, but not directly comparable results.

Code: Select all

Heatsink/Fan                             Noise Fan V  Â°C Rise Â°C/W MP Weight
Scythe Mine with stock 100 fan     ($45) 19    7      28      0.35    560
Scythe Mine with Nexus 120         ($55) 19    7      27      0.34    >560?

Zalman CNPS-9500 LED, Stock Fan    ($60) 22    5      18      0.23    530

There are 2 or 3 concepts floating around in this post that I haven't fleshed out completely. Any help filling in the blanks would be appreciated...

Tibors · Post by **Tibors** » Sun Aug 06, 2006 11:18 pm

mattthemuppet wrote:Tibors - I think you meant to say jaganath was partly correct rather than wrong, no?

No I didn't mean he was partially correct. He was wrong, period. There are two effects which oppose each other. Without a more detailed moddeling, you can't draw easy conclusions.

jaganath · Post by **jaganath** » Mon Aug 07, 2006 1:05 am

Tibors is right about the fluid dynamics of the situation, the tube analogy is actually very helpful in thinking about the airflow through the fins, airflow will be fastest at the point exactly in the middle between two fins and slowest at the boundary layer. We can think about the closeness of the fins as a tube getting narrower and narrower, and hence requiring more static pressure to achieve the same mass airflow through the fins (as SPCR has found, closely-spaced fins work poorly with low [-pressure] airflow).

There are two effects which oppose each other. Without a more detailed modelling, you can't draw easy conclusions.

I didn't draw easy conclusions, I merely wanted it clarified that radiative heat loss is not the only mechanism that benefits from increased surface area. After all, for most of the fins even if they are radiating heat away it is only intercepted by other fins, so there is no net heat transfer from the heatsink to the environment.

Post by **MikeC** » Mon Aug 07, 2006 6:12 am

Postcript added to correct infor about fan removal: It simply sildes out. (Box at bottom of last page)

Aris · Post by **Aris** » Mon Aug 07, 2006 10:43 am

Just cause its got a propritary fan mounting mechanism, doesnt mean you cant swap the fan. We've seen it done on the old zalman 7000 heatsinks. I'm sure if you took it apart and thought about it a bit, you could figure out a way to attatch a nexus 92mm fan to this.

Post by **MikeC** » Mon Aug 07, 2006 11:08 am

Aris wrote:Just cause its got a propritary fan mounting mechanism, doesnt mean you cant swap the fan. We've seen it done on the old zalman 7000 heatsinks. I'm sure if you took it apart and thought about it a bit, you could figure out a way to attatch a nexus 92mm fan to this.

No one said it wasn't possible, just that it's not easy.

BTW, for comparison's sake, the surface area of the Zerotherm HS is given as 4,404cm2. The Zalman 9500 is 3,698cm2; the Zalman 7700 is 3,268cm2. I am sure the Ninja is bigger, tho it's not specified -- probably >5,000cm2.

Felger Carbon · Post by **Felger Carbon** » Mon Aug 07, 2006 11:16 am

MikeC wrote:BTW, for comparison's sake, the surface area of the Zerotherm HS is given as 4,404cm2. The Zalman 9500 is 3,698cm2; the Zalman 7700 is 3,268cm2. I am sure the Ninja is bigger, tho it's not specified -- probably >5,000cm2.

The Ninja is 11cm sq, so each lamella cannot have more than 121 sq cm. There are 23 lamellae, so the absolute maximum possible Ninja area is 2783 sq cm. Actually, it's a lot smaller. Sorry, Mike!

IsaacKuo · Post by **IsaacKuo** » Mon Aug 07, 2006 11:25 am

A square fin which is 11x11cm has a surface area of 242 sq cm, not 121 sq cm...

narrasuj · Post by **narrasuj** » Mon Aug 07, 2006 11:28 am

Felger Carbon wrote:
MikeC wrote:BTW, for comparison's sake, the surface area of the Zerotherm HS is given as 4,404cm2. The Zalman 9500 is 3,698cm2; the Zalman 7700 is 3,268cm2. I am sure the Ninja is bigger, tho it's not specified -- probably >5,000cm2.
The Ninja is 11cm sq, so each lamella cannot have more than 121 sq cm. There are 23 lamellae, so the absolute maximum possible Ninja area is 2783 sq cm. Actually, it's a lot smaller. Sorry, Mike!

Wouldn't they count both the surface and underside of the lamellae (I never knew they what exactly they were called, interesting)? So it should be 5,566 sq cm? Just what I thought of while reading the above post. Sothat means the Ninja still has the most. Any errors in my reasoning?

Post by **MikeC** » Mon Aug 07, 2006 11:36 am

narrasuj wrote:
Felger Carbon wrote:
MikeC wrote:BTW, for comparison's sake, the surface area of the Zerotherm HS is given as 4,404cm2. The Zalman 9500 is 3,698cm2; the Zalman 7700 is 3,268cm2. I am sure the Ninja is bigger, tho it's not specified -- probably >5,000cm2.
The Ninja is 11cm sq, so each lamella cannot have more than 121 sq cm. There are 23 lamellae, so the absolute maximum possible Ninja area is 2783 sq cm. Actually, it's a lot smaller. Sorry, Mike!
Wouldn't they count both the surface and underside of the lamellae (I never knew they what exactly they were called, interesting)? So it should be 5,566 sq cm? Just my reasoning, please let me know if I'm wrong.

Precisely how surface area is calculated in heatsinks. Both sides of a fin count, which is one reason why HS makers went a few years ago to extremnely thin fins. Before CPU heat reached >60~80W, every HS had fins at least 1mm thick, at least, often >2mm. Now, very thin fins (probably ~0.25mm) are the norm, which allows for much higher surface area for a given size & weight.

mattthemuppet · Post by **mattthemuppet** » Mon Aug 07, 2006 1:33 pm

Tibors, I wish I could live my life with the absolute certainty you present on here.

Lets see:

jaganath wrote:
I thought it was the closeness of the fins that impacted on airflow resistance

tibors wrote:
Make the tube narrower, then the resistance increases

are you both not saying the same thing, but in slightly different ways? If so, then if jaganath was wrong, period then you are also wrong, period. Or am I wrong, period? You are both talking about the same thing! Although you added a 2nd, equally important factor (distance travelled/ fin depth) doesn't make jaganath wrong, period IMO.

Kaizer · Post by **Kaizer** » Mon Aug 07, 2006 5:30 pm

Kudos to Devon C. for reviewing the Zerotherms. I was looking forward to a SPCR article on these heatsinks since my original post here http://forums.silentpcreview.com/viewtopic.php?t=32896

I do have a question for Devon, are the fans actually capable of running at 750rpm? If so, at what voltage or is it only possible via PWM?

On the topic of design, I would say that the Zerotherms are no where as feminine as the Vanessa L and S-Type heatsinks.

L-Type:
http://www.xbitlabs.com/articles/cooler ... omp_6.html

S-Type:
http://www.xbitlabs.com/articles/cooler ... omp_3.html

Unfortunately, the performances of Vanessa "sisters" look to be inferior to the Zerotherms'.

As for a fan swap, I have three suggestions.

1) There are holes on the rear end of the Zerotherms both on the top and the bottom-most fins. It might be possible rig some fan clips to mount a 90mm fan onto the rear. One can then use this heatsink in a:
a) Pull-push configuration with 2 fans
b) Pull-through configuration by removing the front fan
c) Push-through configuration by installing the heatsink so that holes now face the front of the motherboard.

2) Since the original fan can be removed without removing the central bracket, it's possible to secure two straight pylons (say using short, straight Steel Tec pieces, part number: S http://www.kzwp.com/steeltec/parts.htm) to the top by using the existing screws. Next, two pieces of 90 degree Steel Tec elbows can be secured onto the edge of the two pylons via nuts and bolts, thereby creating two holes for mounting a fan. I am not sure if this construction would worsen the vibration problem but I suppose one can always mount the fan with rubber grommets.

3) Drill your own fan clip holes on the tip of the fins (if you dare)

.