It's finally up guys. More coming --- after a good long break from thermal grease and clips socket A heavy weights

What a fantastic survey.

These were exactly the heatsinks I wanted to see tested n' compared. And the criteria by which they were judged were precisely those most meaningful to us, the intrepid sailors of the boiling PC cauldron, searching for safe harbor in the form of quiet and effective cooling.

Job well done, Mike. Bravo!

Cheers.

Nigel

doesn't make me regret my swiftys !!!

does make me glad i passed on the zalman.

Thanks Mike! great work

Great stuff. I would have liked to have seen the Alpha 8045 included in the testing as Its also an old favorite and may have remarkable low airflow performance like the AX7 that isn't shown in other reviews. Other than that It looks like the AX7 is king if you're after really silent. I might have to get one.

good article... I recently ordered a swiftech mcx462-u and it'll hopefully arrive next week. It's good to see it performs well, even at 7 & 5 volts. Another great thing I like about the mcx462-u is that it'll work well on my amd xp (1600+ agoia y-stepping) and since I'm planning on going with Intel for my next mobo/cpu upgrade, I can still use the same swifty with a quiet, low-powered fan.

*hugs* his AX-7, it seems to be the best value heatsink out of that comparision u can get in australia, the zalman really turned me off when I saw that it had a single lug clip, now I know why

Great job on the article Mike. Especially with the tests @ 5V & 7V. Only recently has overclockers.com explored cooling performance with different airflow levels, but not to this extreme. Very useful information!

The flip-flop at 5V between the SLK800 and the AX7 is really neat too. Geek wisdom (maybe not so wise?) dictates that dense, thin fin designs like the SLK800 should perform better than widely spaced thick fin designs like the AX7 at higher airflows which leads us to believe that, as you turn down the airflow, heatsinks like the AX7 should perform better than the likes of SLK800s. I've always wondered how low can you go before a thin fin design will perform less than a thick fin design.

Your review helped to show that at a certain point the flip-flop does happen, but not until really-really low cfm's. It's good to know that dense thin fin designs can do well with little airflow - up to a certain point.

I know the SLK800 does not have the most dense fin design, maybe that's why it does well at low airflow like you predicted, but maybe something as dense as a Dynatron DC1206BM-O/CoolJag JAC102C-1A would behave differently?

Very nice Mike. I wonder if these results are applicable with a decoupled fan in a hood (like Ausone and my system). Probably so, because I tested both configurations and my temperatures were very similar.

Have you considered using BURNK7 (for Athlon) or BURNP6 (for Pentium4) instead of Prime95? These programs made Prime95 look like child's play on my Athlon and P4 systems. In particular, on my P4/2.26B @ 2.5ghz, BURNP6 achieved 66C temps, compared with only 61C achieved by Prime95. I am not sure if this will affect your rankings, but it would be interesting to know as it would push the limit of the CPU/cooling system.

Very nice! And hopefully the first of several such reviews.

I do have one point of contention. I wonder if the testing methodology was really fair to the Zalman. (And not just because I sunk $50 into one)
The Zalman is the only heatsink in the test that was specifically designed to be used with a 92mm fan, not an 80.

If you look at the sizes of the heatsinks they are all sized to be fully washed by the air flow from an 80mm fan, except the Zalman. Its larger size, 110mm long compared to 77mm for the rest of them, means that approximately 30% of it was hanging outside the airflow path of the fan.

I'm not sure what effect this would have had on the temps, but I'd like to see how it'd compare with a 92mm Panaflo mounted on it.

This was a concern I had as well, so I did try a PCPower&Cooling SIlencer 92mm, which is about the same noise level as the Panaflo and rated for 28 CFM. There was no appreciable difference.

Had both fans running, swap back and forth every 3 minutes and the HS temp stayed the same. I could feel the airflow from the fan coming all around the HS -- because of all the slots in the fins, the airflow does reach all over the place.

I think the softness of the clip is a problem on the Z. It is bent too easily, you can tell from the feel (and the sound when you knock against it) that the Thermalright clip is much springy / stiff. Probably if you use the Z. clip just a few times, it's fine, but I probably used it a hundred times. The ones I have probably got worn during my preps. I did spend months (off/on) working out this test procedure & in that time, did lots of HS experiments, which included a lot of mounting/unmounting.

One thing I noticed that if gentle even pressure is applied with the palm of the hand on my Z6000 set up on the CPU, the temp drops by a few degrees instantly. This didn't happen with any of the other HS, which tends to validate my suspcisions about the clip. If you go back and read my article about the test methodology, you'll see my POV is that mounting system is a hugely important part of the HS. An improved clip will probably make the 6000cu more effective, but it will not put it in the same class as the others despite its weight. Some conjectures:

1) The slot in the middle for the clip hurts performance
2) The fins are actually too long -- heat transfer becomes less efficient as you lengthen the fin. If it had shorter fins and more of them at the same weight, it would probably work better.
3) they should go a bit higher in mass, larger in fin #, lower & wider in fin size and use the 4 mobo holes for mounting. Then they would have something that might take on the Thermalrights.

I should have put this info in; the thing is so long though -- almost 8000 words already! I will add it later... ADDED -- to last page

BTW, in case it wasn't stated strongly enough in the review, none of these are recommended for use with the Panaflo at 5V (with any CPU at ~65w or higher) if you want a decent margin for error. The AX7 is *probably* safe if you have very good ventilation and/or don't push the system hard for long periods.

LeoV, I will check out BURNK7 & 6.

how about a p4 heatsink review just like that one? that would be great, no one has reviews with quiet fans on heatsinks.

If you read the methodology article you'll see I buit a P4 test platform... P4 coolers are coming.

oh cool, well i look forward to reading it.

It'd be great to throw in the Thermalright SK7 if you get a chance. At overclocker sites, it rates better than the AX-7 and just below the SLK800. It sells for as little as US$30, and it includes a fan (not that you'd use the fan it comes with).

I remember hearing some stuff about how (and I'm probably getting the terminology wrong) while copper's "thermal conductance" was higher than aluminum, it's "thermal permissivity" was lower. This apparently meant that while copper could suck up heat faster, it couldn't dissipate heat as fast as aluminum. Perhaps this also contributes to the low-airflow performance of the AX-7?

I have two AX-7s, both cooling 1Ghz CuMine PIIIs. One is an a "SlimPC" case (yeah, it's kind of crowded) using a 5V Panaflo. The other is cooled by a 5V stock 92mm fan from an older thermal-controlled Enermax PSU, oriented in suck mode. The first gets up to around 40C, the latter to maybe 50C.

Will see if they send me one...

That biz about copper somehow retaining heat is another one of those myths that seem to just go round & round. Two reasons you see copper/alum hybrids: weight & cost. Copper, if I recall, is more than double the cost, and harder to work. It weighs 3x as much. The best aluminum alloys have a heat coefficient of something like 130 "BTU per sq. ft. per hour per °F at 68°F"; copper gives you maybe 230. It passes the heat faster -- coming or going.

The weight is a serious issue for anyone who wants to make & sell them in large quantities to systems integrators & the like, because during shipping, heavy HSF have been known to break or come off in an assembled system. It's like having a captive wrecking ball inside there. By the time the package arrives, everything inside is smashed to smithereens. You don't see massive HSF in production PCs -- except for real customized expensive machines. This is why.

I purchased a CNPS-6000Cu for my XP 1800+, and this review just confirms what I found. This heatsink/fan/bracket kit sounds too good to be true because it is. It seems to have all of the right ingredients (copper construction, plenty of mass, etc...), but it just does a pretty poor job of actually cooling my processor. I'm running a fairly high airflow (read: not quiet enough) machine, and Prime95 puts me on the ragged edge. It hasn't been much different for the ZM-80HP that I bought. I have had to severely underclock my GF2 Ultra to get stable performance. I'm starting to feel like Zalman products just don't work as advertised.

A very welcome and well-done heatsink roundup as heatsinks are now compared to each other with really low airflow for the first time. I found one thing in the test setup however that gives the SLK-800 at least a theoretical advantage to the other heatsinks. The thing is that you didn't put those 4 pads on the bottom of the SLK-800, the consequence being that the whole compressing force of the clip went solely on the interface of the processor core and the bottom of the HS leading to a higher pressure than would have been the case with the pads on.

To give the heatsinks a fairer comparison, I think you should put the pads on the SLK-800 or alternatively take off the rubber pads of the processor. The former way would raise the temps of the SLK-800 by an amount that only testing will show and the latter way would lower the temps of the other heatsinks – and I bet that it would especially benefit the Zalman heatsink because it suffered from a less tight clip.

Old dude, those pads on the XP seem soft enough that the amount of upward force they contribute is not significant. The pads are an OPtion on the SLK800, not a requirement. It seems to me that if by "avoiding" the upward pressure of the pads, the SLK800 design manages to eke out another degree or two advantage, well that's part of the design, and they should be "rewarded" the extra cooling power thus gained.

In all honesty, I don't think you'd see any significant differences, having examined the pads supplied. HardOCP's actually took temp measurements w/ and w/o the pads and found no difference -- with a high airflow fan, of course.

Regarding the Zalman clip: that, too, is part of their design, a soft clip that seems easily bent -- more easily than most. If mounting system of the design hurts its performance, well, that's part and parcel of the product -- a point made clearly in the HS Test Methodology article. Perhaps I should have stated this more strongly: a softish 2-lug clip for a HS that weighs nearly half a kg is not very good design, in my opinion. Too little insurance against mishaps of all kinds...

Awesome. Any chance of including the CoolerMaster IHC-H71, even though it's only designed for 70mm fans?

I had a Zalman 6000-Cu for my old Pentium III. One definite problem with the weak clip was that, with the PC in the normal upright position, the heatsink's high weight and high center of gravity would cause the heatsink to actually be tilted with respect to the processor die, so that it didn't make good contact with the die. I alleviated this problem by installing rubber feet on the CPU package (a la AMD), which helped to prop the heatsink up on the lower side. But I was never comfortable with it.

Your test is drastically unfair to swiftech, you have used recent models for the other two companies yet use an old swiftech. there have been 2 revisions of the swiftech from this version and if you look at www.overclockers.com u can see there is a huge, so much that there are 4 heat sinks seperating the new and older models. Wouldn't it have been far more useful to use their newest product as well.

Sorry for the rant it's just impossible to get low airflow comparison results anywhere.

elandra -- it's only the first of many HS reviews to come. an MCX462+ is coming -- along with newer Thermalrights, Volcano, etc...

OK, Sorry I can be an impatient little bitch sometimes. Especially after a drunk night in the union.

Keep up the good work

i don't think it's really unfair to use an older swiftech. i'm going to go out on a limb here and make a a prediction when mike test the latest MCX462+: @5V you'll see about the same performace, maybe even worse would not be surprising. the airflow impedence of the MCX462+ is very high for a heatsink.

if you're taking requests, mike, the alpha would be nice in the next one. keep up the good work!

one suggestion: mentioning CFM numbers is a bit misleading as this is only quoted at 0 pressure. in any application, the real preformace of a fan is a combination of the P-Q curve, the airflow impedance, the density of the air (temperature and pressure), and the volumetric flow characteristics of the air flow. while it may seem at first to make sense to compare fans at free flow, as a comparitive basis, after more inspection, this could be very far from a apples to apples comparison as each fan may behave very differently in operating conditions. i'm not sure what i'm suggesting as a solution after this semi-rant (sorry for that), as modeling impedence of each HSF setup is not practical in this case, but i think at least fan P-Q curves and an estimated stab at impedence per setup might help. at least it might dispel this myth that CFM at free flow is even a decent metric. you may have already read these white papers- if not, i hope this sheds some light on what i'm talking about:

http://www.comairrotron.com/Engineering/airflow.htm
http://www.papstplc.com/index.php?page= ... les/art006

hyum, I agree about CFM of fans and delivery under load vs without. But hey, I had to start somewhere. Most fan makers don't provide anywhere close to the kind of info you seek -- why do you think I pushed to get that UBC fan project going in the first place?

I think the difference between the Panaflo's ability to push the air through a high impedance HS vs any other similar cfm/noise fan would be small. In other words, all similarly quiet/low 80mm fans will likely have the same result with that HS at that voltage.

If I am testing the HS with a fan rated for 24 cfm in free air at 12V, at least I wanted to provide some info about what it does in free air at the lower voltages. That's all that table was meant to be. You'll note it will be updated when the UBC project bears fruit.

The other thing is, regardless of airflow, most users will simply use a Panaflo fan or similar anyway. The results they get with a Panaflo or any other similar fan will be very dependent on other system variables. Regardless, the resutts ot the testing do show relative performance with low airflow. Something that I also suggested there: the results should noyt be considered in absolute but relative terms -- against the other competitors.

I can definitely confirm the lack of performance on the Z6000 when the clip is not applying enough pressure.

After only a month of casual moving the the PC around, the temps on my Athlon 1600+ started rising, to the point of nearly 70C. I started getting worried, thought maybe one of my fans had gone bad.

I reached in and applied some light pressure to the sink, temp went from 68C to 47 C instantly. No Joke. Immediately shutdown, pulled the sink, and bent the clip backwards so that it would hold the sink tightly on the die. Reapplied thermal spunk and put the sink back in. Temps are now consistently below 45C, even under load and running all day with low RPM fans and the 92mm at minimum.

The clip is by far the weakness of the Z6000, and I think I will be pulling that clip bi-monthly for periodic bending

MikeC, how did you get those diode readings as the mobo you used doesn't support AMD XP CPU internal thermal diode reading?

I guess you didn't read the HS Test Methodology article which I exhorted everyone to do?

I did but apparently not thoroughly enough. I'll reread it but if I don't find the answer to my concern I'll ask you again

Hey Mike, I think that you'll enjoy the newer MCX462+ much more than the original MC462 or the MCX462 non plus versions. I recently got a plus myself and its simply a better heatsink not just in performance but in design and ease of use. If you are familiar with the mounting holes of the MCX non plus, u'll see the plus holes are not as flimsy as the original MCX and the locked screws and stainless steel hardware are a welcome edition:). On another note, would u consider testing heatsinks passively!? Perhaps u would think the cpu would overheat, and it probably would, but maybe u can still do a "time until XX temperature is attained". I remember I read this once in a Zalman all copper flower P4 cooler review and I thought it was a neat idea. It would also probably reinforce the fact that in most cases meager airflow is much better than a purely passive cooling setup. I mean, I totally agree with you that a panaflo at 5v is inaudible in a closed case even at 2:30am as I type this and I've got very little other noise, the most noise is probably my cuda IV 80gb inside a sandwitch and suspended, infact its the only part of my system I can hear at all!

-Ken