I think it would be a shame if SPCR started to become anti-fanless on principle. For certain classes of components it is relatively easy to run them fanless (cool hard drives such as Samsung Spinpoints, mobile/undervolted CPU's, etc). Power supplies are generally the hardest component to run without some kind of active cooling, but as you point out an essentially inaudible undervolted fan does the job admirably and at much the same dB level. But solid-state computing really is the holy grail of silent computing; it may be a couple of decades away, but there's no question that no moving parts = no noise.

There is a certain performance envelope which it is impossible to exceed at the moment if you are going to run your rig fanless (or with only one inaudible fan in the whole setup); for instance I should think that running an X2 4600 totally passively is completely out of the question. However that kind of computing power is really overkill for a lot of computer users, and a 2GHz rig with moderate graphics capabilities should be well within the grasp of most determined SPCRers. I acknowledge that blindly promulgating fanless operation regardless of whether it will actually improve the subjective acoustic experience of end users is folly personified, but equally we shouldn't discourage people from going down the fanless route if it is a viable proposition for their requirements.

MikeC wrote:


jaganath wrote:


How do you definte "totally fanless"? A fanless CPU heatsink? or no fans in the case at all?

My X2 4600 is running fine with a fanless Ninja in my P180 right now. Ambient is 24C and the CPU is currently at 31C. It will occasionally get up to ~43 while running benchmarking tests or imaging jobs. Does this seem unwise or risky? I based my decision based on the temperature data. The temperatures don't seem to be anywhere near putting the CPU as risk. Is there some other failure mechanism that bothers you?

I made a simple modification to the P180 which lowered the upper temperature about 5 degrees C, but it the system is mostly stock parts. The P180 is running both upper stock fans on low and CNQ is enabled. During the day, I use this workstation mostly for database work and programming (and occasionaly web surfing during lunch ) and some video/imaging work during my off hours. It's a tremendous improvement over the window-rattling Dell servers that I previously used.

I can understand the difficulty in defining useful benchmarks for heatsinks and fans. The apples and oranges need to be useful. Comparisons of 120mm fans at 7 volts isn't terribly useful if one fan pulls more CFMs. Comparing db at a constant CFM would be nice, but I'm not sure how easy it is to measure CFM.

I can also understand's Mike's concern about recommending that a particular heat sink can be run fanless. A CPU HS's ability to cool the CPU would be affected by: CPU power, PSU choice (how much does the fan run), PSU location (P180 vs standard case), CPU location (determined by the motherboard; is it close to the case fans?), video card power, etc... Creating a reference system may be useful, but it may be more work that you want to deal with.

Is there a set of criteria in whcih you would consider it is safe to run a CPU heatsink fanless?

Jason

...which is hardly fanless.


I was referring more to a truly fanless system -- one that has no fans. If there is a fan within a few inches of the CPU HS, and it creates airflow through/around that HS, then it really is difficult to describe it as fanless.

We have three fanless systems under examination right now -- two using Zalman TNN cases & one the mCubed HFX. These are designed and manufactured specifically for fanless operation, and cost many times more than ordinary quiet cases + fanned HS & PSU -- yet are not without flaws, and I have my doubts about whether components in these fanless cases will actually have the same reliability/longevity.

You'll see/read more in the reviews...

might be worth adding that the SI-97A is compatible with K8 sockets due to a modification to the base. Only the SI-97 is just Skt A. Mighty fine HS too, plus I can always keep it if/ when I upgrade to 939

and it's Katana, not Kanata - you must have been tired by the time you finished this!

Me and justblair were discussing this only yesterday. The popularity of "semi-fanless" (or "pseudo-fanless" as JB refers to them) setups has rocketed in recent years, especially on SPCR (unsurprisingly). If one really wanted to be hardcore about it you would insist on zero fans in the PC or even near the PC. However having to dissipate all your heat passively/through convection is really tough above ~50W total system heat (ideally the CPU should emit less than 25W of heat). Still, I am very impressed that you are able to run an X2 4600 with no direct active cooling.



Those are great temps. Damn, I have trouble getting those temps with a much, much less powerful CPU!



I personally think any dissipation under 25W is a perfect candidate for fanless operation with the right heatsink and case design.




FWIW I consider the Zalman TNN cases massive engineering overkill. They've approached the problem like "This computer dissipates X watts of heat; how can we safely dissipate that without fans?" whereas a simpler (and cheaper) approach is to ask "I can safely dissipate X watts without fans; how can I reduce my computer's heat output to that level?".

All the time that modern PC's put out 100, sometimes 200W of heat, unaided convection simply isn't going to cut it to remove case heat. At 50W it's feasible, at 25W it's a stroll in the park.

I think that is an interesting point. Though I think both approaches have their own merit.

My current project (Working title "the beast") works on the second principle, turn down the watts till you get to totally passive. When i finally get finished i will post in the gallery. I am trying to organise all the resources that I used to help me so that I can post a mini guide dedicated to reducing clock cycles... If I can reduce the research required for the next person on this quest I will get a happy glow. Anyone who needs some help in the meanwhile pm me and I will send you some links.

To link my experiance of your figures on wattages.

My passive rig, has the processor outputing about 29w of heat that with a typhoon cooling it keeps it between 50c to 60c running MCE. I will tweak it further as I reckon I can knock down the temp just a bit more. 25w would be as you say a stroll in the park.

The approach that you consider overkill though I would not discount straight away. My system would not be everyones cup of tea. i wouldn't like to challenge it with a game for instance, 100% processor load for a prolonged period of time would likely create a situation where I achieve thermal runaway. Ok the mobo would hopefully protect the CPU but I'm loathed to test it in the same way I wouldn't test the air bags of my car by pointing it at a lampost.

Looking to the future I may wish to run Vista MCE if it offers a significant improvement over the current OS (Though what that improvement might be I can only guess) and the cases you mention might be a solution.

From now on I also propose a new lingo when referring to CPU coolers.

Pseudo Passive/fanless (PP)= The cooler relies on another components fan to provide cooling ie a case fan or CPU fan.

Semi Passive (SP)= The cooler runs passive at idle but engages its own fan (or a shared fan) when reaching a set temperature or when launching a particular app.

Purely because I came up with the pseudo passive phrase yesterday and like it! Its also better english I believe.

While I am in a dictorial mood, I also demand to be reffered to as I'l Presidente from this day forward... Or I'll have your goat confiscated by my militia.

great to see a re-done list! no more recommending ppl read an out of date list

Any chance of getting the 2 heat pipe and 4 heat pipe AMD stock coolers rated as a baseline so we can have an idea whether or not it is worth upgrading from these?

Also considering you can buy these relatively easily some users with smaller heatsinks might consider upgrading to this heatsink.

http://www.heatsinkoutlet.com/AMD-B.htm


http://www.madshrimps.be/?action=getarticle&articID=399

says it is a Delta fan that is thermally regulated. Not sure if you could keep the fan from ramping up or if a fan swap would be required

Actually, socket F replaced socket 940 (not counting 130 nm FX series).
However, I'm not sure what kind of mounting socket F uses. This motherboard have only 2 holes for each heatsink. Looks like the old K8 hole pattern, but I'm not sure.

Correction made, along with a number of other small tweaks to Aug 21/06 update of this article.

I was wondering why the Arctic Cooling Accelero S1 hasn't been added to the list yet, but I guess it's too new. Its review sounds like it's going to top the passive VGA cooler list, though. And it's cheap too!

Long needed update of this list, finally posted. Thanks for the reminder, mcv.

Thanks for the update, you know I recommend your work.


A few requests for some day when you feel like you aren't overwhelmed:

1. It would be nice if HS reviews made an effort to hit the ~20 dba mark for 9v Nexus 120 but also hit the ~23 dba mark for 12v Nexus 120. I know you don't recommend coolers that need the higher airflow but it makes it easier to translate from review to review (between sites, or even between reviews of products with lower fan sizes such as the Ninja Mini)

2. For units with integrated fans an accurate weight would be nice as some vendors and ODMs don't specify weight with fan. Some don't specify weight without fan. I'd be interested in accurate numbers for both in cases where there is an included fan.

3. Zerotherm Nirvana NV 120. Looks like something I'd like to see an SPCR review on.

AT did a new revision of their test bench and got their ambient idle down to the limit of their SP meter (and you know how inaccurate it is). It's frustrating to read a dba chart where half the results or more are bumping into the bottom limit of the SP meter.

4. Scythe Ninja Mini. I've said this before but I'll politely bug you about it one more time. An addendum to the review with data for a Nexus 92 at 12v and another 92mm or 100mm fan at about ~22/23 dba would be appreciated. It'd be two more fans to test but hopefully you can wrap it into a 92mm fan review/roundup to gain the extra reward for the effort.

5. Noctua NH-U12F and Tuniq Tower. I'd likely never use these but if you can swing a minimal test just to get the ~20 and ~23 noise levels it'd help again with the cross site review comparisons and help confirm or dissuade purchase decisions for those willing to use your data.

i found error

the review link for Arctic Cooling Accelero S1 is missing a hyphen, so it says "article not found".
bad url: ...com/article793page1.html
good url: ...com/article793-page1.html

im actually reading the article now out of genuine interest. im sure others will want to read it too now that its officially #1 according to mike... and its half the price of the next best option!

Thanks for the long-due update, love it!

(I just thought out loud about the update in http://www.silentpcreview.com/forums/vi ... 824#389824)

what about the Ultima 90? I thought that it was a really well performing heatsink and quiet if with the right fan

I second this question. Where's the Ultima 90? Some reviews compare its performance to the Ultra 120, which is nice considering its lower weight and size.

Hello Mike,

It might be worth mentioning in the Xigmatek HDT-1283 that it is the lightest weight tower? (Ditto for the lightest blow-down heatsink?)

Also, I'm a little confused by the order of the list: the Ninja Rev B is below the Ninja Mini and the Thermalright XP-120, and all three are below the Thermalright SI-128? Is that correct, in terms of their performance? Or, is this just a massive multi-way tie?

OK


The latter. There are at least a dozen rated at 7.5 -- the most common rating. The position on the list is obviously not significant if the Q rating is the same. It'd be nice to show this more graphically, but it's hard to do.

With regard for performance, I don't understand how the Ninja Mini with 80/92mm fan can cool as well as the Ninja Plus Rev B with 120mm fan, or both run passively.

It seems unlikely, but the Ninja Mini with 92mm reference Nexus performs about the same as the rev B w/ Nexus 120 at similar noise levels. This is on our standard Pentium D950 platform. Just examine the data. Now, if you use the TR springs to bolt the rev B, the performance is much better -- but we haven't tested what happens to the Mini if you do the same.

Neither can really run passively, btw, except with other fans creating airflow close by.

I really meant passively as in no fans on the heat sinks, not passively as in no fans in the case.

Considering the difference in size between them, I am inclined to believe the Ninja Rev B would cool more effectively when comparing with no fans on heat sinks. Has your testing proved otherwise?

I agree. But it still doesn't change the relative cool/quiet rating of the two because fanless operation is not a big portion of the "score". Not when there is a steadily growing number fans that run inaudibly while still moving a useful amount of air. There's just not that much to be gained from running the HS w/o a fan unless the fan is within an inch or two, in which case, it's almost the same as being on the HS anyway.

I checked prices on the 4 best top down coolers in the recommended list and rounded to the nearest $5 increment. Given the Triton doesn't come with a fan it isn't much of a bargain item.

It seems to me from the reviews and the tables that the Scythe Andy is worth a nod above the 3 nearest competitors. I'd bump it to an 8 on the recommended scale or get serious about bumping some of the others down to 7.

http://www.silentpcreview.com/article839-page5.html Makes the best case for the Andy where it beats the Triton 75 and Zipang. The Zipang was 2 dbA louder and 2°C hotter at 12v versus the Andy's stock fan or Nexus at the same voltage. At lower speeds the lead is similar for the Andy. To see the Andy beating the Xigmatek HDT-D1284 take a look at http://www.silentpcreview.com/article825-page8.html

Same pricing drill on the smallest tower heatsinks. I don't have an issue with the recommendation score on this pair in relation to each other. But wherever the Xigmatek HDT-D1284 ends up on the rating list the winner of this pair or both deserve(s) to be along side of it.

http://www.silentpcreview.com/article818-page6.html gives you a comparison but for some strange reason Lawrence Lee didn't test the Ninja mini with a 92mm reference fan like Mike Chin did on the Xigmatek. In the stock configuration the Xigmatek wins on performance with a larger fan. If you used the same fan on both which would win?

One has a height advantage, one has a weight advantage. I don't care which one wins or even if they call it a draw. I just want the same fan on both heatsinks so I can compare apples to apples on CFM and dBA as they relate to °C Rise on these two heatsinks.

Somebody get both heatsinks in the same room with the same CPU and a 92mm reference fan and give us the head to head match.

Heck use it as an excuse to redo your sound measurements for the 92mm reference fan.

On the Ninja, Ninja Rev B, Ninja Copper recommendation scores



Based on the old test data the original Ninja was very good. Easily an 8 or 8.5. I understand pulling the recommendation for the original to avoid confusion but I wonder if the Rev. B shouldn't be a notched down to 7 instead of 7.5.

The Ninja Copper is reviewed but not on the recommended list? By temps I call it a 7.5 but if you do you give it a 7.5 you clearly need to keep the Rev B a notch below it.

With the High cost and weight of the Copper with middling performance it surely doesn't deserve an 8.

dhanson865 --

You make some v. good points about relative positioning on the heatsink list, but it's hard to get much finer granulation. My argument for keeping all those 7.5s where they are is that that is more or less where they belong. They are all pretty good for most people looking for quiet cooling. The score is not based on just on cooling and noise performance (both of which have a certain margin of error) -- there's other factors like ease of use, mounting system, weight, and to a small degree, price.

Still, there are many reasons to redo ALL the sound recordings and measurements soon. 2 huge reasons:

1) The new mic/SLM/RTA system appears to have more linear dynamics than the old B&K2203s. Or that's what I would like to believe. The readings are not the same, in any case -- usually ~2 dBA lower at 20 dBA or lower, and somewhat higher as the level rises (say above 25 dBA). I need to get a NIST level 1 reference SLM to confirm the levels once and for all.

But...

2) There's no point doing this -- or making any new recordings or measurements -- until after the anechoic chamber is built. Why? That sound isolation chamber will make a difference in all the sound measurements and all the recordings. So why redo anything now if it all has to be done again very soon?

Been out of the desktop loop for about a year and was looking at the new heatsinks and not too many sites review for SHHH.

THANKS great update!!!

So by inference you must mean the Scythe Andy is noticeably harder to install or you dislike it's mounting system since it has a low price compared to the other 7.5s, a low/medium weight vs the other 7.5s, and better performance in most low/medium airflow cases.

This is borne out by the comments on page 2 and 3 of the review about issues with the clips and the mounting process. Not the worst in the world but you did have more than one comment about it.

Fair enough, it is your ranking system. If you feel it belongs in the big virtual tie you probably have good reason to put it there. I just wanted to make sure it wasn't a casual mistake.



Hey, I'm not saying when you should do it, I'm just asking that when you do retest it would be appreciated if you retest like sized heatsinks with fans that are now considered highly ranked and do so in a consistent fashion.

My contention is that if anyone swaps fans on a Ninja Mini they are likely to use a 92mm fan. According to the review 92mm Fan Clips are included in the retail package. Given that, it makes sense to test it with the 92mm reference fan instead of the 80mm reference fan.

All in all I appreciate the work you and your fellow reviewers do. I just want to ask certain questions from time to time to learn more from your efforts and I think it is healthy for the site in general for me to share my questions so the effort you spend replying to me is distributed to all SPCR readers and may prevent someone else from having to repeat the same question.

Thanks for your time and responses.

Oh and in case I didn't make it clear. You do realize that the Ninja Copper isn't on the recommended heat sink page don't you?

http://www.silentpcreview.com/article30-page1.html search for Ninja Copper and the only place you'll see it mentioned is in the comment block for the Scythe Ninja+ Rev B.

It's totally up to you if you don't want it to be a recommended heat sink. I'm just asking if that was intentional or not?