10-Way heatsink roundup

[jmke]

Got around to finishing a roundup with these coolers:

# AMD Stock Cooling (from 3200+)
# Arctic Cooling Freezer 64 Pro *
# Noctua NH-U 12 *
# Noctua NH-U 9 *
# Sharkoon Red Shock *
# Silverstone NT02 *
# Scythe Samurai Z *
# Arctic Cooling Freezer 64
# Scythe Shogun
# Thermalright XP-120
# Zalman CNPS9500CU

for those interesting in silent & powerful computing it'll surely be interesting.. I hope

http://www.madshrimps.be/gotoartik.php?articID=384

 

[zest]

Saw this too...Don't know why they used crappy acousti DP fans in testing...

 

[cw42]

so who wins? i don't like how their charts are setup, and don't want to read through it.

 

[jmke]

@Zest: it's used for a casefan, it's silent (the whole point of it) why would it be "crappy"?

@cw42: what's wrong with the graphs? :-(

 

[cw42]

Well... i'd rather see comments of the final graphs. There is ALOT of data on them, and people are inherently lazy... so i don't want to read through it all to find out for myself which is the "best". I would have liked to see the reviewers own written opinion/recommendation of which HSF is best suited for what type of application (performance vs silence), instead of just numbers.

It would also be nice to have a drop-down menu on the bottom of each page because i get bored reading, and would like to skip around... and it's a pain to scroll all the way up to skip pages

sorry, i'm just really picky heh.

 

[jmke]

good comment cw42, thank you, I've added the Drop Down menu at the bottom now too;

about the global conclusion, it's impossible to name "the best" as different factors play a role on deciding what heatsink would fit you best.

 

[GalvanizedYankee]

jmke, are you setting this up for madshrimps?

If you are, I really like madshrimps. They do some decent PSU reviews..:thumbsup:
Yayyy Zippy/Emacs

madshrimps being on The Continent does take some getting use too.

I'm in total agreement with having drop-dowm menus. Many review sites put two paragraphs
per page and many slow loading adds. When a site does that I don't even bother and close the page.

spcr is the god around here for silent reviews=tough competition.


...Galvanized

 

[IlllI]

i think the graphs are ok as far as the info goes. but for me, the main problem seems to be that they are too small and a little difficult to read. if possible, i would suggest making the graphs a little larger so they could be read more easily

 

[lifeguard1999]

I agree with cw42, they get an "F" for graphs, and here is why. The graphs are bad because you put temperatures from two different points in the case on the same axis (CPU and PWM). It would make sense if you put idle and load for one component on a axis. It makes no sense to have two components on the same axis. Then you add noise, which is based on dBA, not degrees C, on the same axis.

Don't get me wrong: the review is very good. It is just the graphs that need to be improved.

HardOCP has a similar problem. They sometime use a curve fit for their graphics cards review. The problem is that the points that they are curve fitting are points that are based on different games. How can you judge the frames per second of a point halfway between Quake 4 and Far Cry? It makes no sense.

Again, this is not a flame or a personal attack: the review is very good. madshrimps, HardOCP, AT and other sites all have this problem of how to convey information to the reader.

 

[RampantAndroid]

Originally posted by: GalvanizedYankee
jmke, are you setting this up for madshrimps?

If you are, I really like madshrimps. They do some decent PSU reviews..:thumbsup:
Yayyy Zippy/Emacs

madshrimps being on The Continent does take some getting use too.

I'm in total agreement with having drop-dowm menus. Many review sites put two paragraphs
per page and many slow loading adds. When a site does that I don't even bother and close the page.

spcr is the god around here for silent reviews=tough competition.


...Galvanized

Someone made a comment about people not reading...and its just been proven true. jmke is listed as the author on the first page at the top....

 

[jmke]

Originally posted by: lifeguard1999
I agree with cw42, they get an "F" for graphs, and here is why. The graphs are bad because you put temperatures from two different points in the case on the same axis (CPU and PWM). It would make sense if you put idle and load for one component on a axis. It makes no sense to have two components on the same axis. Then you add noise, which is based on dBA, not degrees C, on the same axis. .

idle values for heatsinks have no use; who cares what temperature your CPU is when it's idle, what counts is temperature under load, because that's what the heatsink is there for, remove heat under CPU LOAD

about the graph setup, by adding noise/cpu/pwm in one graph you get a quick and accurate idea of how loud the HSF is, how hot the CPU runs, and how hot the area around the CPU socket gets. I can't see why this graph setup would be confusing?

in what other way can I relate noise & CPU temp in one graph? if I split them up, I'm stuck with 6 pages of long graphs, and you'll have to skip from on to the other to know if the Samurai Z at 46°C is actually producing 71dBA of noise...

all comments are welcome and appreciated, and I thank you all for the time you take to look at the site, I try to improve where possible (I add the quick page jump at the bottom ) and if I can make the graphs better I'm open for suggestions; but just don't quite see how I can make them compact, easy to read and good for reference beside the method I'm using now.

 

[0roo0roo]

the graphs need different groupings or something, else its just a pain to look at. the idle/full load etc need to have their separate groupings so its easy to compare the difference. the main heatsink should be highlighted.

 

[jmke]

uhm, there are no "idle" values, what do you by "mean heatsink" ?

 

[0roo0roo]

well i mean the heatsink being focused on in the comparison should be bolded/highlighted in a way that makes it stand out from the reference coolers.

 

[GalvanizedYankee]

n/m

 

[lifeguard1999]

Originally posted by: jmke
about the graph setup, by adding noise/cpu/pwm in one graph you get a quick and accurate idea of how loud the HSF is, how hot the CPU runs, and how hot the area around the CPU socket gets. I can't see why this graph setup would be confusing?

in what other way can I relate noise & CPU temp in one graph? if I split them up, I'm stuck with 6 pages of long graphs, and you'll have to skip from on to the other to know if the Samurai Z at 46°C is actually producing 71dBA of noise...

And therein lies the problem. One of the reasons I dislike Tom's Hardware is that there are tens of pages of graphs. I like AT and Xbitlabs as they have a "print page" function that puts it all on one page for me. So that is one possible solution: multiple graphs with a print page function so people can scroll back and forth. Works great for broadband users like me, not so good for dialup users.

Here is a second possible solution. On the lower axis, have it be degrees C in red. On the upper axis, have it be dBA in green. On the bars for the values of temp and noise for each heatsink, color it red if it is in C, and green if it is in dBA.

I am sorry; I should have offerred possible solutions last night in my other post. It is bad form to critque something and then not offer ways to do it better.

 

[jmke]

we have a print page function, it's under the "author & editor text"
I will look into your suggestion for the graph

afaik AT & TomsHW don't have HSF roundups with comparable data, last HSF review at AT is.. ?

@0roo0roo: I should highlight all the heatsinks then? because they are all the focused ones

edit: I get what you mean, AT puts the product tested in one color, and the others in another color; although it is a good way to approach graph when testing one (or two) new products, when almost ALL products in the graph are "NEW" that method won't work :/

 

[cw42]

i have to agree with the HardOCP graphs sucking, i mean the FPS ones because it's completely pointless to look at. I like their testing methods, just not the way some of it is organized heh.

jmke: one idea i had is to be able to sort each graph. What I mean is that for each graph, you'll be able to click on CPU, PWN, or Noise... and it will recreate the list in the order of lowest to highest. I didn't like having the browse up/down the list looking for which performed the best in each catagory.

after looking at the graphs for a longer time, they don't seem THAT bad. The problem is that you have so many different test setups that takes up space. I really don't think it's neccesary to compare that many diff test setups anyway. Just pick a setup that has good airflow, at very little noise so u can get a good judgement of CPU noise. You're testing CPU HSFs, not case airflow.

I don't really mind a long article even if it is like 30 pages. As long as it's organized, and easy to read... that's all that matters to me.

 

[jmke]

cw42 that graph system would be super but would that require flash? or javascript? I'm not that good with those (no programming/scripting background) and having to enter that many data first in excel and than again in another app would be.. well.. mind-numbing ;-)

but it would be cool indeed

 

[tw33ter]

no scythe ninja?

 

[Diasper]

Wow that must have been a crazy amount of work involved to do that.

The only thing is the data needs to be presented in a much more easily understadable way - because currently it's pretty inpenetrable.

I'd suggest sorting the heatsink according to noise. I'd suggest 3 levels of noise.

- Nigh silent level
- Pretty quiet but allowing performance
- At the 100% value of the heatsink

The first two given the noise db would be the same for all CPUs would make the data MUCH more easily understandable - ie performance/noise which is what most people are concerned about. After that people can then compare it against cost.

From those graphs it seems the Artic Cooling 64 Pro is doing pretty well for itself - its predecessor proved a very good heatsink especially from a performance/noise/cost point of view.

Perhaps another cooler to consider adding is the Asetek Sythe Ninja as its a heatsink which (at least when I looked a few months back) was several under reviewed especially comparing against other heatsinks. The other problem was the mounting mechanism is some reviews didn't benefit it and so temps were worse than could be. If you did a test I'd recommend using the mounting method recommended for DFI boards - here.

 

[jmke]

the Scythe Ninja and Asetek VapoChill Micro have already been reviewed at the site, would be hard to include all heatsinks I tested in the past

the graph performance/noise is sorted by noise also, of course, showing you the coolest and most silent HSF near the bottom of the list.


(Ninja: http://www.madshrimps.be/?action=getarticle&articID=224
Vapo: http://www.madshrimps.be/?action=getarticle&articID=348)

 

[themusgrat]

I think the graphs are good. Thanks for the review; we don't get enough heatsink reviews around here.

 

[Diasper]

Hmm, I think one way to make the results more easily understandable and an easy thing for you to do is to perhaps offer large scatter graphs of noise vs cpu load with all the heatsink results on it. This way you can incorporate all the data from all the heatsinks including with different speed fans in an easily comparable table. In fact by interpolation you can probably guess the performance of any heatsink at different fan speeds/noise outputs.

This would be a very easy way to summarise data - all you would need is to have as many graphs for as many different setups that you have eg one for each cpu mhz & voltage setting that you test or active PSU etc.

The other reason why this graph would work well is because you can then easily section it up into noise levels. eg for those looking for nigh silent computing they can look for whatever heatsink provides the best temperatures at say sub 35dB. For those looking for a bit more performance and don't mind the noise so much they can look for the best performer at under 40dB etc. I'm not sure on what dB figures would represent what in your graphs but you could easily add a comment above or under each graph.

The fact that you take more than one speed setting result means using some interpolation we can guess what results for any fan might be at say 40db if say results at 100% and 50% fan speed only offer data for 45dB and 35dB respectively. Of course, you could also just provide data for how the heatsink performs at 75% to get an even better idea of how the performance delta scales with fan speed vs performance (and thus sound too).

Does that sound like a viable idea? I think if data could be summarised quickly and easily into graphs like that that would keep it a short simple review that would be much more accessible (which would hopefully mean more readers in the longer term) and therefore be much better. I think the danger of tech websites is that so there can be a tendency for being too techy and detailed rather than offering all the salient data in as simple and comprehensive a format as possible. Of course, when some websites try and make themselves too accessible they sometimes forget to undertake the technical aspects diligently, thoroughly and from all angles.

The only other thing I think that could be added is comments on the nature of the sound emitted from the different heatsinks as there will be differences depending on the fan used or the mounting mechanism (eg introducing more rattle etc). I think silentpcreview do something like that although I'm not sure if they comment on the quality of the sound emitted at different fan speeds.

Outside of that at some point one day I reckon it'd be very interesting to do a comparison of some directional heatsinks say in a case like the P180 or otherwise with ducts on them to see how they perform against the more traditional blow-down heatsinks given how heat can be directly exhausted out. With the the Asetek Vapochill Micro Ultra Low Noise in a rig I recently built is getting some great results as the heat gets directly blow out.

One more thing - on your Vapochill review what mounting mechanism did you use - did you use the one they say or the one recommended for DFI? I'm asking because the DFI one gives much better contact (poor contact is a problem frequently reported) and can lower temps massively (as much as 10C). Also did you use Artic Silver 5 after removing the previous epoxy (without leaving residue etc) to see if temperatures got even better. I suspect temperatures for the Asetek Vapochill would be even better then and could equal/better than those like the XP90/120...

Just to re-iterate the second poster - if I was doing the tests I might perhaps do them differently ie without sound dapening materials and probably with a case where the PSU is at the bottom so as to take it out as a factor helping cooling or adding to noise as a passive PSU could easily be included.

 

[jmke]

Hey there! Sorry for the delay, here are my answers

Hmm, I think one way to make the results more easily understandable and an easy thing for you to do is to perhaps offer large scatter graphs of noise vs cpu load with all the heatsink results on it. This way you can incorporate all the data from all the heatsinks including with different speed fans in an easily comparable table. In fact by interpolation you can probably guess the performance of any heatsink at different fan speeds/noise outputs.

I?ve been trying to get a scatter graph to work, much like in this review: http://www.matbe.com/articles/lire/267/comparatif-de-20-ventirads/page18.php
However I have no clue on how he got his dots ?connected?; because with 10+ heatsinks (he?s got 20 in there) it becomes a rather hectic affair and hard to see how each heatsink performed. Those who are colorblind are sh*t out of luck too ?

The other reason why this graph would work well is because you can then easily section it up into noise levels. eg for those looking for nigh silent computing they can look for whatever heatsink provides the best temperatures at say sub 35dB. For those looking for a bit more performance and don't mind the noise so much they can look for the best performer at under 40dB etc. I'm not sure on what dB figures would represent what in your graphs but you could easily add a comment above or under each graph.

Yes that would be cool, but the dots system might not work out ideally for easiness of reading.

Does that sound like a viable idea? I think if data could be summarised quickly and easily into graphs like that that would keep it a short simple review that would be much more accessible (which would hopefully mean more readers in the longer term) and therefore be much better. I think the danger of tech websites is that so there can be a tendency for being too techy and detailed rather than offering all the salient data in as simple and comprehensive a format as possible. Of course, when some websites try and make themselves too accessible they sometimes forget to undertake the technical aspects diligently, thoroughly and from all angles.

I?m really confused as to why the graphs I currently use are HARD to use? They clearly show what heatsink offers you the best noise/performance ratio? Or am I missing something?

The only other thing I think that could be added is comments on the nature of the sound emitted from the different heatsinks as there will be differences depending on the fan used or the mounting mechanism (eg introducing more rattle etc). I think silentpcreview do something like that although I'm not sure if they comment on the quality of the sound emitted at different fan speeds.

In an open testbed, like SilentPCReview uses, yes you can have a ?nature of the sound? however when inside a case all these ?typical? sounds are muffled and you are left with very similar sounding.. sounds. So describing them would be a futile challenge?

One more thing - on your Vapochill review what mounting mechanism did you use - did you use the one they say or the one recommended for DFI? I'm asking because the DFI one gives much better contact (poor contact is a problem frequently reported) and can lower temps massively (as much as 10C).

A fellow reviewer at the site reviewed the VapoChill Micro, it was not tested on the DFI, so mounting was no issue; see http://www.madshrimps.be/?action=getarticle&articID=348

Also did you use Artic Silver 5 after removing the previous epoxy (without leaving residue etc) to see if temperatures got even better. I suspect temperatures for the Asetek Vapochill would be even better then and could equal/better than those like the XP90/120...

I used Arctic Alumina on ALL heatsinks, and previous paste was removed with ActiClean
http://www.arcticsilver.com/arctic_alumina.htm


Just to re-iterate the second poster - if I was doing the tests I might perhaps do them differently ie without sound dapening materials and probably with a case where the PSU is at the bottom so as to take it out as a factor helping cooling or adding to noise as a passive PSU could easily be included.

The majority of PC cases out there have PSU at the top, so it will be more preventative to test in a case similar to the most used out there?