ZEROtherm Nirvana CPU Cooler

Want to talk about one of the articles in SPCR? Here's the forum for you.
MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

ZEROtherm Nirvana CPU Cooler

Post by MikeC » Tue Oct 27, 2009 1:46 pm


RoGuE
Posts: 540
Joined: Mon Jul 06, 2009 9:11 am
Location: Massachusetts, USA
Contact:

Post by RoGuE » Tue Oct 27, 2009 4:35 pm

Thanks for the review. I skimed over it, and the results didn't really surprise me. It's refreshing to read reviews that are "straight up" ..not beating around the bush because you got some free samples.

Thanks again.

Hypernova
Posts: 124
Joined: Sat May 31, 2008 2:33 am
Location: Australia

Post by Hypernova » Wed Oct 28, 2009 2:11 am

Boy what a stupid deign on the fan placement. How do they expect the fins to get air with it so far away. Most of the air flow would take the easy way which is straight out of the huge gap instead of the heat sink.

Jordan
Posts: 557
Joined: Wed Apr 28, 2004 8:21 pm
Location: Scotland, UK

Post by Jordan » Wed Oct 28, 2009 3:57 am

When do proprietary fans ever prove a good option these days? I kinda wish manufacturers would just ditch the idea and let us choose which fan we want to use. :?

Sunrise
Posts: 40
Joined: Mon Jan 26, 2009 10:30 am
Location: Finland

Post by Sunrise » Wed Oct 28, 2009 5:03 am

Does delta t / watt stay roughly constant with varying wattage? I haven't really given it much thought before, but the real question when considering a cooler is what processor I can keep at a 30-40C temperature rise over ambient at a given SPL, not how cool I can keep the processor. I realize you'd need additional equipment to test what wattage it can cool adequately, but is this the only reason no site provides numbers for wattage at a given temperature?

Assuming my guess about a constant delta t / watt is correct, here are the wattages for the table on the last page @ 16dB and 30C:

Prolimatech Megahalems 234
Thermalright U120E 195
Thermalright HR-01+ 180
Xigmatek HDT-S1283 180
Scythe Kabuto 180
Noctua NH-U12P 167,14
Zalman CNPS10X Extreme 167,14
Zerotherm Zen 156
Scythe Mugen-2 156
Thermalright U120 156
Noctua NH-C12P 146,25
Scythe Ninja 2 137,65
Thermolab Baram 130

burebista
Posts: 402
Joined: Fri Sep 02, 2005 12:05 am
Location: Romania

Post by burebista » Wed Oct 28, 2009 5:08 am

I'm stumped. I ignore that we are almost in 2010 and test platform is with a Presler but I can't ignore this
Since the Nirvana fan is not easily replaceable, it cannot be easily compared to other 120mm fan heatsinks where our reference quiet fan was used. However, we made a calculated extrapolation based on fan RPM.
Extrapolation in a cooler review? :shock:
Nirvana with Gentle Typhoon
Image
Nirvana with Noctua
Image
Here you have a modern platform and Nirvana at stock speed and here are my friend tests (with Scythe SlipStream SY1225SL12LM) and no way it's possible to say something like this:
Poor performance:size ratio
Eventually you can say Poor performance:size ratio based on extrapolation.
Or to paraphrase last sentence from review:
Sorry, SPCR, better luck next time!

burebista
Posts: 402
Joined: Fri Sep 02, 2005 12:05 am
Location: Romania

Post by burebista » Wed Oct 28, 2009 7:58 am

MikeC wrote:What are you blathering on about?? The meaning of your post is utterly opaque.
I want to say just one thing Mike: if you want to do something do it right or don't do it.
MikeC wrote:There's no way to replace the stock fan with the reference fan, so if you want a noise/airflow/cooling relevant comparison, extrapolation is necessary.
Two of my friends replaced stock fan with Gentle Typhoon, Noctua and Slipstream for their testing. I'm really sorry if you don't find a way and they did.
MikeC wrote:If you think the extrapolation is inaccurate, say so and why.If you think the extrapolation is inaccurate, say so and why.
All I want to say is that with same fan @800 RPM Nirvana is practically on par with top coolers and far away from your extrapolated conclusion.
Of course on a modern CPU.

Image

LE: Ops, I'm speaking with myself now. Where did your post go? :shock:

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Wed Oct 28, 2009 8:28 am

burebista wrote:LE: Ops, I'm speaking with myself now. Where did your post go?
I decided my post was too knee-jerk... but it seems to have evoked a more complete critique from you, so that's good.

Now you say your friends use a "modern" CPU, suggesting that it automatically makes it better. Perhaps you're right that we should use a hotter CPU, but the exact type of CPU doesn't matter, only that it runs hotter and uses a current socket. But the 78W we get at the 2x12V socket at full load is still representative of a wide range of CPUs favored by silencers. Also keep in mind that our particular CPU has a rated TDP of 95W.

So let's talk about the TDP of your friends' "modern" CPU. Well, there are 2 versions of the Intel Q6600 used by your friends. One is rated for 105W TDP and the other is rated for 95W TDP. They mention that theirs is a "G0" stepping -- this is the 95W version: Their processor has exactly the same TDP as ours. They're using prime95, just like we are. Do you really think the actual power measured at the 2x12V socket will be any different than the 78W we measured on ours?

Finally, I disagree that replacing the stock fan on this heatsink is such a wonderful thing to do. There is such a wide range & number of good coolers on the market that provide excellent quiet cooling without any modding or fan swapping at similar or lower cost. Why bother?

burebista
Posts: 402
Joined: Fri Sep 02, 2005 12:05 am
Location: Romania

Post by burebista » Wed Oct 28, 2009 9:17 am

MikeC wrote:Perhaps you're right that we should use a hotter CPU, but the exact type of CPU doesn't matter, only that it runs hotter and uses a current socket.
I hope you're joking, right? It's not only about TDP but IHS size and cores number and HT for Nehalem for example. Put a HDT heatsink with 4-5 heatpipes on a CPU with a small IHS and put it again on Nehalem and Phenom II which have a larger IHS. On which CPU they'll perform better?
MikaC wrote:Their processor has exactly the same TDP as ours. They're using prime95, just like we are. Do you really think the actual power measured at the 2x12V socket will be any different than the 78W we measured on ours?
Do you believe that a Pentium D 830 put same heat as i7-920 with HT on only because both have TDP 130?
And if you look carefully at that graph you'll see a Q6600 G0 @3.3GHz and 1.344V Vcore. You know that heat grows liniar with frequency and at square with Vcore increase so you can't compare your Presler with a Q600 overclocked. You simply can't.
MikeC wrote:Finally, I disagree that replacing the stock fan on this heatsink is such a wonderful thing to do.There is such a wide range & number of good coolers on the market that provide excellent quiet cooling without any modding or fan swapping at similar or lower cost.
Yep, here I agree with you.
MikeC wrote:Why bother?

Because you're a reviewer and you must respect your readers and show them real data not extrapolated one collected on ancient hardware.

RoGuE
Posts: 540
Joined: Mon Jul 06, 2009 9:11 am
Location: Massachusetts, USA
Contact:

Post by RoGuE » Wed Oct 28, 2009 4:30 pm

burebista wrote: I hope you're joking, right? It's not only about TDP but IHS size and cores number and HT for Nehalem for example. Put a HDT heatsink with 4-5 heatpipes on a CPU with a small IHS and put it again on Nehalem and Phenom II which have a larger IHS. On which CPU they'll perform better?
Whoa, whoa, whoa. First of all, Bureb...chill. Mike and his writers didn't HAVE to write this review for you, so don't act like they owe you something. Cause they don't. If anything, you owe him a little respect.

Second of all, your accusation regarding their lack of a "modern" cpu is ridiculous, and actually made me laugh.

Have you ever taken a heat transfer class? I have. Recently actually. Half the class deals with heat sinks (much like the ones we strap to our processors). Power, is heat, is power, is heat. I don't give a shit how many cores a processor has, all heat sinks big and small are designed to whisk that heat away as quickly as possible. What this means is, the base (usually copper) is at a uniform temperature. Sure there are hotspots, but essentially, that base is uniform. What Mike and Lawrence test is the ability of the heatsink to remove this uniform heat, and at what sound levels. If X cpu has 8 cores, and Y cpu has 1 core, and if they are both rated for 95W, you can bet your ass that any heatsink will perform the same on both processors. I understand your point about different size sockets, but if anything, more cores distribute heat more EASILY, so your comment about the i7 actually hurts your argument. Yeah, it might put out more heat, but with more cores, there is less hot-spoting going on, and heat is more easily transferred to the heat sink base.

BUT I DIGRESS, say I'm wrong, and i don't know shit about heat transfer. Even if I don't, it's not like you are taking the numbers he gives you and calculating out exactly how much heat you are removing from your CPU in Joules. SPCR tries to keep the test base a constant, so that old test results can be compared to NEW test results. It's not about accuracy, its about consistency. Clearly you missed this point all together.

As for the whole testing different fans thing, I think Mike handled your comment appropriately, I as a buyer would never even think twice about taking a fan out of a hea tsink like this. It's really not something I would do unless I was given the heat sink. So including that in the review is irrelevant IMO.

Get ur attitude in check kid.

tehfire
Posts: 530
Joined: Mon Jan 01, 2007 9:57 am
Location: US

Post by tehfire » Wed Oct 28, 2009 6:07 pm

RoGuE wrote:
burebista wrote: I hope you're joking, right? It's not only about TDP but IHS size and cores number and HT for Nehalem for example. Put a HDT heatsink with 4-5 heatpipes on a CPU with a small IHS and put it again on Nehalem and Phenom II which have a larger IHS. On which CPU they'll perform better?
Whoa, whoa, whoa. First of all, Bureb...chill. Mike and his writers didn't HAVE to write this review for you, so don't act like they owe you something. Cause they don't. If anything, you owe him a little respect.

Second of all, your accusation regarding their lack of a "modern" cpu is ridiculous, and actually made me laugh.

Have you ever taken a heat transfer class? I have. Recently actually. Half the class deals with heat sinks (much like the ones we strap to our processors). Power, is heat, is power, is heat. I don't give a shit how many cores a processor has, all heat sinks big and small are designed to whisk that heat away as quickly as possible. What this means is, the base (usually copper) is at a uniform temperature. Sure there are hotspots, but essentially, that base is uniform. What Mike and Lawrence test is the ability of the heatsink to remove this uniform heat, and at what sound levels. If X cpu has 8 cores, and Y cpu has 1 core, and if they are both rated for 95W, you can bet your ass that any heatsink will perform the same on both processors. I understand your point about different size sockets, but if anything, more cores distribute heat more EASILY, so your comment about the i7 actually hurts your argument. Yeah, it might put out more heat, but with more cores, there is less hot-spoting going on, and heat is more easily transferred to the heat sink base.

BUT I DIGRESS, say I'm wrong, and i don't know shit about heat transfer. Even if I don't, it's not like you are taking the numbers he gives you and calculating out exactly how much heat you are removing from your CPU in Joules. SPCR tries to keep the test base a constant, so that old test results can be compared to NEW test results. It's not about accuracy, its about consistency. Clearly you missed this point all together.

As for the whole testing different fans thing, I think Mike handled your comment appropriately, I as a buyer would never even think twice about taking a fan out of a hea tsink like this. It's really not something I would do unless I was given the heat sink. So including that in the review is irrelevant IMO.

Get ur attitude in check kid.
+2

RoGuE
Posts: 540
Joined: Mon Jul 06, 2009 9:11 am
Location: Massachusetts, USA
Contact:

Post by RoGuE » Wed Oct 28, 2009 6:39 pm

burebista wrote: You know that heat grows liniar with frequency and at square with Vcore increase so you can't compare your Presler with a Q600 overclocked. You simply can't.
lmao..did you just pull this out of your ass? Heat grows with POWER OUTPUT. The more power the cpu consumes, the more heat is generated. Amount of heat has nothing to do with frequency..it could be running at 100hZ and still put out 95W if it were designed that way. Yes, power consumption goes up when frequency and vCore go up, but the end deciding factor is the power consumption. So yes, he can compare his presler to a q600.

dev
Posts: 43
Joined: Sun May 10, 2009 12:47 am
Location: Here

Post by dev » Wed Oct 28, 2009 10:23 pm

I've always found your cooler reviews very htpc friendly and less relevant for actual power users.

It's quite clear that the fan that equips the Nirvana is pretty poor compared to your reference fan and to use extrapolation only makes matters worse. But to come up with that sort of conclusion is just poor testing.

Let me ask you this: if you're not goning to test a product right, why bother?

burebista
Posts: 402
Joined: Fri Sep 02, 2005 12:05 am
Location: Romania

Post by burebista » Wed Oct 28, 2009 11:14 pm

I usually don't talk with half-learned guys but for you I'll made an exception because you enlighten my day.
RoGuE wrote:Heat grows with POWER OUTPUT. The more power the cpu consumes, the more heat is generated. Amount of heat has nothing to do with frequency..it could be running at 100hZ and still put out 95W if it were designed that way. Yes, power consumption goes up when frequency and vCore go up, but the end deciding factor is the power consumption.
I hope that you read twice your statement before you hit that Submit button. :D
Let's see if I understand correct what you're saying after your heat transfer class.
More power=more heat. This is good
Power goes up with frequency and Vcore. Looks OK, it means that heat goes up with frequency and Vcore. Ops, not, because you said that Amount of heat has nothing to do with frequency.
I hope that you pass that heat transfer class.
RoGuE wrote: So yes, he can compare his presler to a q600.
Of course, you can compare anything with anything (like my example with P4 and Nehalem because they have TDP 130) but the result is meaningless.
RoGuE wrote:SPCR tries to keep the test base a constant, so that old test results can be compared to NEW test results. It's not about accuracy, its about consistency.
Then I'm thankful that they don't start with a PII or something, just for consistency sake. Who cares about accuracy, we all want consistency. :roll:
RoGuE wrote:Get ur attitude in check kid.
.......
lmao..did you just pull this out of your ass?
Thanks for your kind words, when you'll have 42 years and 17 of them in IT field maybe I'll take them into account but until then go back at your heat transfer class.
Don't bother to reply.

All I want to say in my posts is that you can't bash a heatsink on extrapolated conclusions because someone is lazy to do a fan swap or to test it on modern hardware.
You have above a graph with Nirvana and a Slipstream @800 RPM put on a quad oc-ed. It's not extrapolated. Now after looking at that graph can you conclude Poor performance:size ratio?

Sunrise
Posts: 40
Joined: Mon Jan 26, 2009 10:30 am
Location: Finland

Post by Sunrise » Thu Oct 29, 2009 3:53 am

So much wrong with this thread... Heat does rise with the square of the voltage and linearly with frequency for the same processor core. Although the stock q6600 g0 might put out roughly the same as the SPCR standard presler, the one tested was overclocked and overvolted giving significantly more heat.

The age of the processor has absolutely nothing to do with cooling performance. You can test heatsinks on any flat surface with a constant heat output.

Extrapolation of results: The measured temp rises were 22C @ 15dBa and 27C @ 14dBa. The extrapolated ones were 21C @ 16dBa and 32C @ 13dBa which is far from unreasonable. The bad performance comes from using the loud stock fan, not from inaccuracies in the extrapolation.

Comparing the Nirvana with stock fan to others with a quiet replacement fan might be unfair but it reflects reality if replacing the fan on the Nirvana is too much work to be worth the trouble. This point is obviously debatable and seems to be the only valid critique of the review.

dev
Posts: 43
Joined: Sun May 10, 2009 12:47 am
Location: Here

Post by dev » Thu Oct 29, 2009 6:41 am

Removing the fan from Nirvana is difficult if you are 3 years old or you are missing your thumbs.

The stock fan is not loud but just not up to silent standards. If you find this fan loud then the ones equipping the Zalmans are unbearable. Problem is its not efficient and certainly not built for low rpm.

If I managed to test quite a few fan on this heatsink I'm pretty sure MikeC himself will have no problem.

PS: A little bit of history: Nirvana PWM is an updated version of the Nirvana Premium or if you wish a rev. B version with a pwm fan and 1366 mounting, and has nothing to do with the FZ120 which was launched after the Nirvana Premium.

shleepy
Friend of SPCR
Posts: 454
Joined: Tue Aug 02, 2005 3:32 am
Location: SF Bay Area, California

Post by shleepy » Thu Oct 29, 2009 12:34 pm

I think that the core of the matter is this:
The fan in this heatsink isn't MEANT to be switched out. Unlike, for example, Thermalright heatsinks that don't come with "integrated" fans, and often ship without fans at all, switching out the fan in a heatsink like this one essentially constitutes "modding."

You don't review a crappy case and say, "well, if we drill more fan holes and replace the HDD mounting mechanism with a custom one of our own, this would be a GREAT case," and give it a good overall review. Why would you do an analogous thing with a heatsink? Just like fan positioning and HDD mounting are essential features of a case, the ability of a heatsink to easily mount a non-standard fan is part of what makes a heatsink good or bad. If the fan isn't meant to be replaced, then I say that it's fair to review it as is, and not go beyond what the manufacturer meant for the heatsink to be.

As for the other arguments that burebista and dev (same person?) are making - they are entirely illogical. 'Nuff said. Just don't listen to them.
Last edited by shleepy on Thu Oct 29, 2009 1:50 pm, edited 1 time in total.

RoGuE
Posts: 540
Joined: Mon Jul 06, 2009 9:11 am
Location: Massachusetts, USA
Contact:

Post by RoGuE » Thu Oct 29, 2009 1:16 pm

burebista wrote:I usually don't talk with half-learned guys but for you I'll made an exception because you enlighten my day.
RoGuE wrote:Heat grows with POWER OUTPUT. The more power the cpu consumes, the more heat is generated. Amount of heat has nothing to do with frequency..it could be running at 100hZ and still put out 95W if it were designed that way. Yes, power consumption goes up when frequency and vCore go up, but the end deciding factor is the power consumption.
I hope that you read twice your statement before you hit that Submit button. :D
Let's see if I understand correct what you're saying after your heat transfer class.
More power=more heat. This is good
Power goes up with frequency and Vcore. Looks OK, it means that heat goes up with frequency and Vcore. Ops, not, because you said that Amount of heat has nothing to do with frequency.
I hope that you pass that heat transfer class.
You missed my whole point. And that is, the frequency of the CPU doesn't matter if they are both generating the same power output! Yes, the power goes up with frequency, but if they are both consuming 95W, and the frequencies are different (clearly) than power and frequency are unrelated.

I admire your experience in the IT field, but conisder this: "modern" processors haven't been around that long, so my experience is just as valid as yours (despite the fact that I'm 21).

You seem like a smart enough guy, but you really are missing the whole point I was trying to make. You also continue to miss the point that SPCR reviews of heatsinks are meant to be consistant. NOT accurate to your specific CPU. In other words, if I hooked this heat sink up to an i7 920 overclocked, I wouldn't expect to see the same temps and noise levels!!! However, I would be able to look at the SPCR data, and make an informed decision on which heat sink to buy, based on their CONSISTANT test data.

I'm not really sure how I can make that any clearer...you seem to be so stuck in your ways that there is no convincing you otherwise.

mascotzel
Posts: 21
Joined: Sun Sep 17, 2006 2:35 am
Location: Romania

Post by mascotzel » Thu Oct 29, 2009 4:40 pm

RoGuE wrote: You missed my whole point. And that is, the frequency of the CPU doesn't matter if they are both generating the same power output! Yes, the power goes up with frequency, but if they are both consuming 95W, and the frequencies are different (clearly) than power and frequency are unrelated.
and for this you are using manufacturer's stated TDP? (i doubt that if you actually measure power draw it will be the same)
so 2 completely different CPU's,but with the same TDP, will rhave the same thermal behaviour ? (that's what I understand form your post)

I read the review and came to this page for :
1. Nirvana is older than Zen
2. extrapolated results have no place in a decent review (make it as a sidenote if you must)

RoGuE
Posts: 540
Joined: Mon Jul 06, 2009 9:11 am
Location: Massachusetts, USA
Contact:

Post by RoGuE » Thu Oct 29, 2009 5:14 pm

mascotzel wrote: and for this you are using manufacturer's stated TDP? (i doubt that if you actually measure power draw it will be the same)
so 2 completely different CPU's,but with the same TDP, will rhave the same thermal behaviour ? (that's what I understand form your post)
Essentially, yes. Power is heat per unit time (Watts = Joules/seconds) So, if 2 processors are both rated for around the same power draw (95W for example), than they both put out the same amount of heat in Joules per unit time! What does this mean? It means Mike could strap the heatsink to a 95W heating pad the size of the heatsink base, and it would perform roughly the same as if it were mounted to a "modern" cpu (given constant thermal interface conditions).

I think a lot of this confusion about power, and different CPUs and whatnot, is really just a common lack of understanding of how power relates to heat generation.

I'm not trying to get in anyone's face, I'm merely defending the SPCR heat sink test platform from an engineering/experimental standpoint.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Thu Oct 29, 2009 6:22 pm

Let me get this straight --

dev -- you're one of the reviewers of the review that burebista linked?
and mascotzel -- are you linked to that review or the product in any way? Let's clear the air here, it's getting hot.

First, no more insults from anyone to anyone. I see one more snide or rude remark here, the poster's account is going to be wiped out.

mascotzel -- your declaration that "extrapolated results have no place in a decent review" has no basis in reality. Extrapolation is an essential tool in science and engineering. Don't make sweeping generalizations, and be straight. I think you don't like how bad our test results make the nirvana look. I'm not sure why.

What was the extrapolation for? To make an educated guess at what the cooling would be if the stock fan was set to make 16 dBA and 13 dBA of noise at 1m. This was based on the assumption that most readers would not buy this heatsink to swap the fan. So assume the fan is fixed, that the fan and heatsink are one, and compare it in some fair way to others at the top of our recommendations.

Why extrapolate? Because often it's when all the testing has already been done, the heatsink packed away, and we're working on the writing that fine details crop up. Larry wanted to add this info, but didn't want to set up the whole test again to measure, because it was trivial, the data points needed.

He had a measurement of 22C rise at 15dBA and one of 17C rise at 20 dBA; Larry's guess of 21C at 16 dBA is perfectly reasonable. We also measured 27C rise at 14 dBA -- that's a 5C rise for a 1 dBA drop (from 15 dBA) which means the airflow is falling below the ideal bell curve of the heatsink. Larry used that 5C for the next 1 dBA, which is how he obtained 32C temp rise at 13 dBA. Again, this is perfectly reasonable.

I sincerely believe actual measurements would be within a degree of Larry's estimates. But they're not even necessary to do. There's just no denying the simple fact that the fan on the Zerotherm, mounted the way it is, sounds terrible at every speed.

You can argue that other heatsinks have nasty fans too but that's not the point. Those heatsinks get thumbs down here too.

You can also argue that it's unfair to compare the Nirvana with the others when the others all have the advantage of the quiet reference fan. That's because all the others allow the reference fan to be mounted w/o any issue or have equally quiet fans (the Scythes and the Noctuas) or leave the fan choice up to the user. That stock fan and mounting is part and parcel of the Nirvana, and if they leave the Nirvana disadvantaged, that's not our problem. (I'm reiterating shleepy's point here)

The real point is whether the review is "sound" -- has scientific validity, repeatability, and most of all, usefulness to the audience. (That's what the criticisms have either implied or stated.)

The answer is yes -- on all counts. It's the same test method that's been used for years, the results are usually repeatable ad nauseum (ie we'll get the same result within a degree if we run the test a bunch of times). the results are useful to our readers. Certainly, afaik, no one can touch our acoustic analysis. Now if you run >120W CPUs at full tilt all the time, a test database with a hotter CPU (or equivalent heat source) might be more useful... but most readers don;t do that. Not even gamers can push the CPU like prime95 continuously for a hour or more.

You can also argue that if the CPU was running a lot hotter, and if Nirvana and all the other heatsinks were tested on that CPU, that the Nirvana would come out looking better.

Maybe. Maybe not. It did in lab501's testing. Is lab501's test more scientific? I'm not sure. There's certainly no attempt to measure the CPU heat, which is at the crux of burebista's argument that it's superior. In contrast, we know ours is pulling 78W because we measured it using reliable tools.

An aside: One thing that's a bit dubious about the lab501 result is how well the nirvana did with low airflow. That dense honeycomb structure is smack in the middle of the fans, and it's what gives it high fin density, which should translate to better cooling with higher airflow and hotter CPUs. But its position increases impedance to airflow just where the fan has least airflow -- in the center where the hub is.

This is the opposite approach taken by the Prolimatech, where the fin density was increased along the outer edges of the fins and reduced in the center -- which more accurately reflects the airflow pattern of the fan, and hence maximizes airflow evenly through the surface area. And we know that the Prolimatech has been ranked tops by just about everyone, with lower or higher airflow, cooler or hotter CPUs.
Last edited by MikeC on Fri Oct 30, 2009 11:21 pm, edited 1 time in total.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Thu Oct 29, 2009 7:20 pm

burebista wrote: I want to say just one thing Mike: if you want to do something do it right or don't do it.
Burebista great screen name but I think this was rude.
Also your experience in IT has nothing to do with this subject.

Heat is complex and I'm sure I will omit or be wrong in some parts I'm not an expert.

Heat is direct proportional with power consumption but this is not all.
And yes power consumption is about directly proportional with frequency and square(V).
Heat is generated by the transistors inside the CPU.
There are multiple layers of heat insulator and conductor between the transistors and heatsink on the "new" Intel and AMD Desktop CPU's.
There are also a metal plate glued on top of the desktop chip's so you will need to consider also the glue thermal resistance.
This is probably one of the reason the mobile CPU have the die exposed (better thermal interface).
The size of the chip die is another important factor in heat dissipation.

What I want to say is that two different CPU with the same power consumption will have different die temperature with the same heatsink applied the same way.
The only way to test heatsink's is to use the same CPU then the comparison between different heat sinks will be valid.

dev
Posts: 43
Joined: Sun May 10, 2009 12:47 am
Location: Here

Post by dev » Thu Oct 29, 2009 10:46 pm

Mike all that reply and no valid argument. Fact of the matter is you guys chose to speculate instead of testing. That from my point of view is a no-go for a reputable source like spcr.

And to answer your question I have owned a Nirvana and while its a few degrees worse than a Megahalems at every rpm it does a better job low rpm compared to the Fz120 for example. And if poparamiro confirmed my findings I'm going to consider his results valid simply because he didn't extrapolate.

burebista
Posts: 402
Joined: Fri Sep 02, 2005 12:05 am
Location: Romania

Post by burebista » Thu Oct 29, 2009 11:32 pm

shleepy wrote:As for the other arguments that burebista and dev (same person?) [...]
Thank you. Beside I'm moderator/user on a couple of our forums you can find me here on SPCR but on ocforums, hardwarecanucks, jonnyguru, guru3d, techpowerup and XS too.
shleepy wrote:The fan in this heatsink isn't MEANT to be switched out.
OK, then don't include extrapolate results, use just what you've tested.
electrodacus wrote:Burebista great screen name but I think this was rude.
Sorry, I apologize to all if it sound rude it wasn't my intention. All I say here I say because I care about SPCR and I'm VERY grateful to Mike&reviewers and forum crowd for teaching me a lot about silence.
I've criticize the guys on HWC too on their heatsink testing methodology after I saw how CM Hyper 212+ was a miracle of heatsink.
They listen and changed it and in next round-up results are normal now.
To be honest I want to spoke about SPCR testing bed a long time ago but I was lazy. This Nirvana review filled my glass so I spoke now.
electrodacus wrote:The only way to test heatsink's is to use the same CPU then the comparison between different heat sinks will be valid.
Absolutely agree. But from time to time you need to move on on next CPU generation for new heatsinks. You can't be stuck with an older CPU just for consistency sake. I as a reader I don't care how Prolimatech/10X/whatever is doing on a P4 because any modern decent heatsink can cool that CPU and we are close to 2010 not to 2006.
Now for consistency sake in 2014 when I'll come here and read a heatsink review I'll see same Presler tested?
electrodacus wrote:Also your experience in IT has nothing to do with this subject.
Agree with that too. :)
But when one of my best forum friends has a hobby to test rigorously any heatsink that he can buy/borrow/receive from manufacturer then I can speak about heatsinks without fear.

Mike I really appreciate your explanations, theory is very nice but practice kills us. :)
I saw here some ideas that you can use a thermal pad/whatever heated uniform like a CPU IHS and test heatsinks because they'll perform the same like in real life. It is a site who's doing that (frostytech) and all their results are false because we don't live in a perfect world. If you take a look at their Top 10 Intel heatsinks you'll see that is full with HDT heatsinks. Unfortunately in real life no actual HTD heatsink can touch a Prolimatech/X10/TRUE at high TDP and airflow because their exposed heatpipes are off of the IHS. More heatpipes more waste.
Of course for SPCR crowd a undervolted/underclocked CPU can be cooled with cheapest HDT heatsink without problems and they have a great performance/price ratio.

Bottom line, I'll always came on SPCR with pleasure but if heatsink methodology is kept the same I'll listen only fans sound measurements because I know that extrapolated data on an ancient CPU is useless for me.

Thanks for reading and trying to understand my points.

Cistron
Posts: 618
Joined: Fri Mar 14, 2008 5:18 am
Location: London, UK

Post by Cistron » Fri Oct 30, 2009 3:14 am

Heat is heat. The only substaintial difference to a new CPU test-bed I can see, would be the way of mounting and the contact surface between heatsink and heatspreader. Based on these assumptions I doubt the ranking of recommended heatsinks would change much and I suppose most people wouldn't buy from the bottom of the list anyway.
Mike all that reply and no valid argument. Fact of the matter is you guys chose to speculate instead of testing. That from my point of view is a no-go for a reputable source like spcr.
And who really gives a damn? The mounted fan is apparently not sound and most people wanting a quiet PC will buy another heatsink for a probably cheaper price. You guys are making a mountain out of a molehill.

mascotzel
Posts: 21
Joined: Sun Sep 17, 2006 2:35 am
Location: Romania

Post by mascotzel » Fri Oct 30, 2009 5:14 am

I had nothing to do with that review,I went here right after I finished the review (there's that linked "discuss in out forums" at the end of each your reviews)
I said in my post those things

and testing cooling hardware is not exact like "1+1=2 math"...
that's why extrapolating doesn't just do it,in my opinion
fans are not the same all over the world also

anyway,as far as CPU cooler testing methodology goes for,SPCR is great for the silence related part...and as for cooling performance,an interested user must read much more

and note this,I'm NOT saying that Nirvana is better than you extrapolating results,but it could be,as it could be worse...the thing it didn't have the Nexus fan on it to see the actual result

jhhoffma
Posts: 2131
Joined: Mon Apr 25, 2005 10:00 am
Location: Grand Rapids, MI

Post by jhhoffma » Fri Oct 30, 2009 9:35 am

I have to be honest, I don't understand all the animosity or disdain toward the review.

The purpose of an SPCR review (and just about any other review) is to gather as much data as the review can and present it to the reader in an impartial fashion so that the reader can better make up his/her mind regarding a purchase.

As Mike said, extrapolation is a VALID scientific/statistical tool, providing there adequate data to make the extrapolation meaningful. The fact that the data was extrapolated and not derived directly, is clearly spelled out in the review. If anyone was looking closely enough to that section to read the numbers, it could not be missed.

All SPCR reviews have some slight subjectivity (opinion) to them. This is important because a lot of newer readers don't fully understand enough about thermodynamics to really comprehend all the data. If MikeC says "Based on this feature, we prefer ProductX to ProductY" it doesn't mean that all users will feel the same. In fact, that's why the forums are such a great place to come directly after reading the review (and why a link to the thread is at the end of each article). Other viewpoints are definitely welcome, but really need to done with a respectful tone (which was not done here). No one here is attempting to mislead the reader, just provide additional information.

Most newer readers aren't going to go to the extreme to modify the heatsink to fit a stock fan. Those that are, usually are experienced enough to know they can go beyond the parameters of the review.

(edit)One more point. The key point to comparative testing is keeping the conditions the same for all test samples. Changing the setup to a "more modern system" would be counter-productive as it would invalidate all previous testing, just as changing microphones would. One heatsink review (MadShrimps?) uses a simple resistive heat source (like a small hot plate) to conduct thermal testing. Does that make his reviews/testing invalid?

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Fri Oct 30, 2009 12:31 pm

jhhoffma wrote:I have to be honest, I don't understand all the animosity or disdain toward the review.
Me neither. The only thing that might explain it is that our results conflict with those other results, which somehow threatens them? I dunno.

The inability or refusal of those folks to understand the extrapolation we used is truly amazing and appalling. Here's a last attempt to force illumination.

Image

If you can't understand or accept what was done after this, there's nothing more to be said and this discussion might as well be locked.

And stop confusing the issues. Extrapolation is one issue, and how hot or cool our CPU is another. They are separate issues. Keep them separate!!
Last edited by MikeC on Fri Oct 30, 2009 11:24 pm, edited 1 time in total.

RoGuE
Posts: 540
Joined: Mon Jul 06, 2009 9:11 am
Location: Massachusetts, USA
Contact:

Post by RoGuE » Fri Oct 30, 2009 1:45 pm

This thread has become some kind of absurd, subjective pile of nonsense that contains very little scientific understanding of basic testing principals.

Bottom line, there is nothing wrong with reasonable extrapolation, and using an "old" platform does not affect the quality of the review in anyway, shape or form. See my above posts for detailed explanations.

It really should be locked...

nomoon
Posts: 210
Joined: Thu Apr 14, 2005 8:35 pm
Location: Allen, TX US
Contact:

Post by nomoon » Fri Oct 30, 2009 6:33 pm

Thanks for the review! It told me everything that I needed to know.

For what it’s worth, a large part of my doctoral dissertation covered thermal modeling. I took my first heat and mass transport class as an undergraduate at MIT. I whole heartedly endorse Mike Chin’s methodology for testing heat sinks. Mike knows what he's doing.

The only way to accurately compare heat sinks is to use the same CPU and motherboard, and only change the heat sink, which is exactly what Mike does. Mike changes as few variables as possible.

It doesn’t matter if my system uses a CPU that is a few years newer and hotter (or colder) than Mike’s standard system. The heat sink that performs best on Mike’s system will almost most certainly be the best performing heat sink for any other system, regardless of the CPU TDP. Mike’s methodology measures the relative performance.

95 watts is 95 watts. It doesn’t matter whether a chip is running at a high frequency and low voltage, or a low frequency and high voltage. It doesn’t even matter if we know whether the chip is running at 50 watts or 100 watts, as long as it's the same for each test. Relative performance is all that matters if you want to compare performance of various CPU heat sinks.

If someone is controlling their fan speed by choosing discrete voltages for their fan, then Mike’s comparative performance chart showing temperature rise versus voltage is perfect. One might argue that it might useful to see graphs of db versus temperature rise for people who are limiting their fan speed based on temperature rise. However, even that point is debatable. Mike’s comparative performance chart is still extremely useful regardless.

Once again, we have a well done report. We got more than our money's worth :D. If this particular cooler was a contender for first place, then I might see more interest in seeing non-extrapolated data points for low voltage operation. It’s not even close, so I’m not interested.

Jason Lewis

Post Reply