Joe's Thermosyphon at overclockers.com

[gone_fishin]

Originally Posted by Cathar

It gets better:

http://www.overclockers.com/articles1257/

He puts a disclosure at the end of the review in smaller font.

His test results show the ambients steadily decreasing at each test.....don't see the logic in doing that.
Between tests 1 and 3 there is a 1.3C change in ambient and a 3C change in die temps for a 60% decrease in rpm's.

Then the 4th final test has only a .1C change in ambient which makes the raw numbers look like a big spike for only another drop of 37% rpm's between test 3 and test 4. If the same percentage of ambient drop would have continued, then the whopping 38C die temp would not have been observed (which he based his entire final conclusion of it being a poor choice at low rpm's)

Way to make a test enforce what you want to write, that's all I got to say.

[Hoot]

So, is this thread ever going to get back to discussing the technical aspects of thermosyphons?

Hoot

[ricecrispi]

I read Cathar's remarks about JoeC. Cathar is right about the intergrity issue. I mean there is a reason why they perform double blind experiments... Sometimes people bend information or slightly lean to a side to prove theories. Even subconsciously. It gets only worse when the variable $ gets added to the equation.

Anyways, I went to look at some of Joe's reviews.
The storm APEX kit has performance equal or even lower to astek micro extreme, sl-120, and his own thermosetup. I mean the Sl-120 heatpipe beats the Storm Apex kit hands down and across the board. The APEX kit got .15c/w in the P4 testbed but on a die sim gets .08-.1c/w. JoeC wrote "This is not exactly comparable to the P4 testing I've done but it's reasonably close, so P4 performance is outstanding." I think 7 C degrees is a big difference......

Joe's first article 1246 he calculates .14-.15c/w on the testbed. Running of the A64 case setup .19 to a lower .17c/w. A big SURPRISE, off a die sim he gets .22 c/w.

Article 1248 for the his own single tubed Thermosyphoon test he calulates a .12 c/w on the testbed and gets .18-.19 c/w in running setup. He claimed his small die simulator is out of service and doesn't post numbers..... Wondering why he didn't?

I think 7C is a big difference from Storm APEX kit underperform 7C and his 5C overperforming thermosyphoon. Combine the two and you get a huge margin.

[Cathar]

ricecrispi, I think you're getting confused over which testbed he's used for which results. He has 3 testbeds, and I believe you are comparing the results with one testbed (which gives quite high C/W results) with another testbed (which gives very low C/W results).

He has the old T'Bird die-sim Testbed.
The Acorp P4 testbed
...and his newer 1" squared copper slug testbed (large die sim).

[Etacovda]

i fail to see why he doesnt use one die sim for the lot personally.

Ive seen many arguments that watercooling is 'a waste of time because the temps are worse than with heatsinks' - and when asked where they're getting that data, they provide the p4 and the watercooling results from Joe. Course, trying to tell them that they're different sims and yeild quite different results, generally ends in them saying 'C/W is a measurement that doesnt change across platforms,' followed by some kind of insult, a general flame fest and a locked thread.

$10 says nothing gets done about it though.

[Ruiner]

Is a 1"square slug a real simulation of an IHS equipped core? I would highly doubt that there would be heat dispersal more than a couple of mm from the edges of the core itself, let alone homoegeneously throughout the 1" square.

IHS's are to protect the cores from poor installs, not to 'spread heat', despite their name.

I would think that the T-bird simulator is most accurate of the three he uses. Is that the consensus from the gurus here?

[Ls7corvete]

Is a 1"square slug a real simulation of an IHS equipped core? I would highly doubt that there would be heat dispersal more than a couple of mm from the edges of the core itself, let alone homoegeneously throughout the 1" square.

Some of the other threads are duscussing this right now, you may want to take a look.

i fail to see why he doesnt use one die sim for the lot personally.

Yea he should make that more clear in the articles, doesnt he have air cooling sorted by test bed? Should be easy enough to change the way data is organized.

[UNDERBYTE]

Originally Posted by Ruiner

Is a 1"square slug a real simulation of an IHS equipped core? I would highly doubt that there would be heat dispersal more than a couple of mm from the edges of the core itself, let alone homoegeneously throughout the 1" square.

IHS's are to protect the cores from poor installs, not to 'spread heat', despite their name.

I would think that the T-bird simulator is most accurate of the three he uses. Is that the consensus from the gurus here?

Testing is always relative and the data is always subject to how well it is interpreted as well as how it is aquired. A die is a die . A 1cm die vs a 1" in. will give you relative performace rankings with a pretty good degree of accuracy., which is what most people are interested in..

Maybe a 1' in. die gives an accurate represention of an Intel IHS, the #'s are the same. But should That be the Real focus of test results? Who Knows? I do not care enough to start my own test program

The XP90 C tests as well as some good water cooled systems. But that comparison is not fair in most case because it does not have the advantage of externally mounted radiators drawing in ambient air. Case air being 5,10,15 C higher will give an application C/W 2x or 3X higher.

Testing is always relative

The XP90 has several heat pipes spread across the base - My Guess is that at 100W+ small die it will start crapping out earlier than it would if tested at the higher loads on a 1" . I think OC falls short by not testing at higher W values. Limitaions of design become more apprant at higher load levels.

Testing is always relative


Design observation

The XP90 is a clever design. The heat conducting pipes are mostly located near the tips of the mounted fan. 75% of the Fan flow is in the last 25% of the outboard fan tip. The higher speed air flow at the tip is more effective at cooling than spreading them across the radiator and increases overall airflow through the sink by decreasing the pressure drop.

Might make a good water radiator design.

[Razor6]

Originally Posted by UNDERBYTE

Testing is always relative and the data is always subject to how well it is interpreted as well as how it is aquired. A die is a die . A 1cm die vs a 1" in. will give you relative performace rankings with a pretty good degree of accuracy., which is what most people are interested in.

That's not necessarily true, especially for water blocks. Due to design differences, blocks will favor larger die sizes to different degrees causing the relative rankings to change.

Predicted info from Les


The WW performs increasingly better compared to the MCW6000 as the die size increases.

[BillA]

nice data

[ricecrispi]

Thanks for clearing things up Cathar. I understand he uses different testbeds thats why I tried to list them. Looking back and forth it got confusing. I'm used to the old tbird sim and P4 testbed but all the other results from 2 different testbeds got the head spinning.

It just bothers me the APEX kit got such a seemingly lacking review.
The review was a page long and really lacking compared to old WCing reviews by JoeC.

Joe posted the large die sim results at .08C/W. He mentioned P4 test results but no post? I don't understand why he couldn't indulge when he covered the Kingwin aquastar 3x for 6 pages, expcially when it performs so bad.

He posted the old die sim results which are really bad, worse then TDX and many Wcing blocks. I'm guessing he used a different setup but with a small die sim the results was .15c/w, equal to mcw-5002?

[BillA]

aw crap
just noticed "Predicted info from Les", I thought it was the real thing

[UNDERBYTE]

Originally Posted by Razor6

That's not necessarily true, especially for water blocks. Due to design differences, blocks will favor larger die sizes to different degrees causing the relative rankings to change.


The WW performs increasingly better compared to the MCW6000 as the die size increases.

Point taken - the bigger the watt range, the more sampled, large sie, small die, in/out of case etc. the better.

Still my statement is true. Whether you are comparing large or small die indivdually the comparison is true for that particular configuration(assuming it's good data). All blocks on one particular die will establish a accurate range of values. Change the setup(die) and you change the values.

My point is you need relative data points for reference(How does it fit in multiple applications? What is the noise level compared to performance? etc.). one Data point on one die @ a single wattage when comparing dff. cooling solutions is a bit inadequate.

[jaydee]

Originally Posted by UNDERBYTE

Point taken - the bigger the watt range, the more sampled, large sie, small die, in/out of case etc. the better.

Still my statement is true. Whether you are comparing large or small die indivdually the comparison is true for that particular configuration(assuming it's good data). All blocks on one particular die will establish a accurate range of values. Change the setup(die) and you change the values.

My point is you need relative data points for reference(How does it fit in multiple applications? What is the noise level compared to performance? etc.). one Data point on one die @ a single wattage when comparing dff. cooling solutions is a bit inadequate.

There is no point at all in using a 1" die sim. There is no CPU on the market that is comparable to it. If your going to use a die sim at least use one that is in the mid range of all die sizes. Take the largest and the smallest and build the die size in the middle. Then the results should be somewhat comparable to a similarly wattaged system. I can understand the reasoning to doing die sim tests and system based tests but not a 3rd with a completely out of whack die size. Just makes things more complicated and confusing. Maybe thats is his goal though....

[UNDERBYTE]

Originally Posted by jaydee

There is no point at all in using a 1" die sim. There is no CPU on the market that is comparable to it. .

It looks like he got a pretty good approximation of the Intel IHS in his tests when you you compare the data.....Not so with the AMD IHS

If it is accurate data, then maybe a little more thought went the selection of the 1' die for Intel IHS than is apparent to the casual observer?

Anybody else here on the forum have a 1' in die they have tested on and can attest to the diff.? Otherwise idle speculation.

In any event I would like too see more data that validates the test method.

[jaydee]

A 1" die is not a 1" IHS. A IHS does not make heat. You have a die under the IHS, the IHS is not going to spread the heat from that die perfectly to all parts of the IHS. The heat will be concentrated in the area the die is under the IHS. If JoeC wants to mimic a CPU with a IHS he needs to make the die the same size as the CPU's die and then put a IHS on top of it. The extra TIM layer of the IHS adds another PITA to deal with.

It dosn't matter how accurate the data is if it applies to nothing.

[gone_fishin]

Originally Posted by jaydee

http://www.overclockers.com/articles1258/

Boy oh boy, he kept that case chilly if you compare the 2 result pages of the Asetek and his contraption.
He drew his conclusion for the reader, "This test resulted in a C/W of 0.14 - similar to the Acorp result." after showing a chart for a mobo testbed.....further he states it is a 200 watt solution but only tested it at 70watts.

Looks like painting a grey sky blue.
Nothing is known for certain except for what he says with no hard data to back it up.


That flexible tube puts the mass of the condenser and fan away from influencing the die mount, that is a big plus that I see in his design.

[ricecrispi]

Originally Posted by jaydee

There is no point at all in using a 1" die sim. There is no CPU on the market that is comparable to it. If your going to use a die sim at least use one that is in the mid range of all die sizes. Take the largest and the smallest and build the die size in the middle. Then the results should be somewhat comparable to a similarly wattaged system. I can understand the reasoning to doing die sim tests and system based tests but not a 3rd with a completely out of whack die size. Just makes things more complicated and confusing. Maybe thats is his goal though....

I'm with Cathar on JoeC recent change in test setups. They stayed pretty much the same for how many years?

Well having a die size too large is not good for JoeC because reading the link supplied by bobkoure
Exploration of the Theoretical Limit of Thermosyphon Cooling System

A huge determining factor was area (cross section) of the boiling area interfacing to the heat source. Larger the area, better results. Explains why he got such a bad C/W for small die sim and a low C/W for large die sim.

[gone_fishin]

Originally Posted by ricecrispi

I'm with Cathar on JoeC recent change in test setups. They stayed pretty much the same for how many years?

Well having a die size too large is not good for JoeC because reading the link supplied by bobkoure
Exploration of the Theoretical Limit of Thermosyphon Cooling System

A huge determining factor was area (cross section) of the boiling area interfacing to the heat source. Larger the area, better results. Explains why he got such a bad C/W for small die sim and a low C/W for large die sim.

So it boils down to surface area, pardon the pun.

[UNDERBYTE]

Originally Posted by jaydee

A 1" die is not a 1" IHS. A IHS does not make heat. You have a die under the IHS, the IHS is not going to spread the heat from that die perfectly to all parts of the IHS. The heat will be concentrated in the area the die is under the IHS. If JoeC wants to mimic a CPU with a IHS he needs to make the die the same size as the CPU's die and then put a IHS on top of it. The extra TIM layer of the IHS adds another PITA to deal with.

It dosn't matter how accurate the data is if it applies to nothing.

I really do not understand why people make it so complicated.

you have a thermal cascade with the die to grease to IHS and then from the package to IHS.

It is simple to measure that resistance, (case temp vs diode) then calculate how much copper area you need to approximate a case temp C/W = make and test it.
Die simulator/IHS simulator.......whats wrong with that? Just another relative way to look at data

You will have more delta temp variation in small die area changes then you will to large area dies. Going from 1' to 2 in. might yield a few 10th's of a degree. going from 1cm to 2cm would probably net you a couple of degrees C

1 cm and 1 in. dies are common in a number of papers I have read. So are .5 and .75 in. Choose your poison for quick mental HTC calcs

Looking only at die temps without an IHS is only part of the picture and will bias any test result if that is the only tool you use, a case C/W is just part of the total cascade as is the on die C/W.

[jaydee]

Note the black square in the middle represents the die and the entire dark blue area is the IHS...... Once again it is not a good idea to have a die sim so big.

Originally Posted by Groth

This is a little later than planned, but...

Given 32 mm square copper heatspreader 1 mm thick, heat it with a 10 mm square 100 watt uniform heat source, cool the opposite side with a heatsink with an h(eff) of 20,000 W/m^2, what's the heat flux density at the heatspreader/heatsink interface?

The whole thread: http://forums.procooling.com/vbb/sho...3&page=1&pp=25

[UNDERBYTE]

Originally Posted by ricecrispi

I'm with Cathar on JoeC recent change in test setups. They stayed pretty much the same for how many years?

Well having a die size too large is not good for JoeC because reading the link supplied by bobkoure
Exploration of the Theoretical Limit of Thermosyphon Cooling System

A huge determining factor was area (cross section) of the boiling area interfacing to the heat source. Larger the area, better results. Explains why he got such a bad C/W for small die sim and a low C/W for large die sim.

I am not sure it is all that bad. I estimate the condenser is probably .07-.08 C/W. (from comparable sized radiators tests)that would put the boiler in the .15 C/W range. Compared to OC waterblock tests on the same simulator it is up there in performance and it is still early in the development cycle. the best being .12 C/W

Just to note that while googling thermosyphons I came across a paper where the author had tabulated that there were over 2500 papers published worldwide on thermosyphons since 1998 - I am not sure I would want to sort through all that for the last word on the subject

[UNDERBYTE]

Originally Posted by jaydee

Note the black square in the middle represents the die and the entire dark blue area is the IHS...... Once again it is not a good idea to have a die sim so big.

The whole thread: http://forums.procooling.com/vbb/sho...3&page=1&pp=25

I am not sure I get the point. IHS as heat spreaders are marginal at best? You can send the graphic to Intel I think they would probably agree with that.

I thought we were discussing simulator test methods? This makes no connections or sense to me, Please explain

[UNDERBYTE]

Given 32 mm square copper heatspreader 1 mm thick, heat it with a 10 mm square 100 watt uniform heat source, cool the opposite side with a heatsink with an h(eff) of 20,000 W/m^2, what's the heat flux density at the heatspreader/heatsink interface?


jaydee

You made my point for me - What they are asking for is the CASE TEMP....And You can measure that or build a simulator

[jaydee]

I could care less about case temp. I was refering to posts 103-109 were die sims were being discussed. JoeC seems to want to confuse everyone with a mass of data that really dosn't pertain to anything except promoting his product. Maybe we were arguing 2 different points. If you are refering to the CPU done tests then I am wasting time. Testing on a P4 is pretty useless for real performance numbers.