More efficient waterblocks lead to higher water temperatures?

Just throwing this one out there for discussion.

Now for a perfectly insulated heat die who's only source of heat dispersal is through the waterblock, then the opening statement is going to be false because watts in equals watts out through the waterblock, so no matter what, the water will always be receiving the same amount of heat and therefore will always be the same temperature given a water cooling source of fixed performance and the system reaching equilibrium.

What the opening question relates to is for CPU systems. I believe it has been fairly conclusively established that the CPU actually dissipates a not insignificant amount of heat through the CPU pins and into the area of the motherboard surrounding the socket. This is most easily noted when watercooling a CPU and watching the on-die thermistor. When the case is well ventilated and a fan is introduced to blow over the socket area substantially lower on-die temperatures are seen (ranging from 2-5C for modern CPU's). This is observed when the water-cooling radiator is distantly removed from the case where it cannot be impacted in any significant way by the socket air-flow.

So clearly there are two cooling paths here, that being the primary cooling path of the waterblock, and the secondary cooling path of the motherboard.

Therefore in a computer system we now have a tied relationship between the waterblock cooling performance, the motherboard cooling effect, and the resultant water temperature.

Let's say, for example, that the motherboard has a C/W of 1, and the waterblock has a C/W of 0.25, being figures that I believe are not too distant from reality. This would mean that the cooling system has a total C/W of 0.20. Now I could be wrong on this but I'm treating the thermal resistance as electrical resistance, using using Rt = (R1 X R2) / (R1 + R2) to arrive at that figure.

Let's simplistically assume that the water is a constant ambient temperature for now.

If the CPU is emitting 60W of heat then the CPU will reach 12C above ambient, for which 48W of that heat is being dissipated by the waterblock, and 12W by the motherboard secondary heat paths. ie. 48W of heat is entering the waterblock, and therefore the water.

Now let's assume that we plug in a block with a C/W of 0.20. The total thermal resistance of the system is now 0.1666 C/W. (1 x 0.2) / (1 + 0.2)

For the same 60W CPU, the CPU now reaches 10C above ambient, and now the waterblock is dissipating 50W of that 60W heat total. (10 / 0.2)

ie. 2W more heat is entering the water with the more efficient water-block.

Is there something wrong with my logic? 2W is not a lot, and it's certainly not an amount large enough for most anyone to pick up on by watching their water temps (re: HardOCP comments about more efficient waterblocks overloading a radiator), but if the theory is sound, then the effect is still there. It's not zero, and it's not large, but it is real. Yes?

no difficulty with the logic, but the measurement is impossible
the increment is too small, and would in any case be dwarfed by the mounting variation
(run the calcs, at even low flow rates)

the other 'problem' is that if the dimensions of the 2ed wb are not the same as the first, the changed airflow will also result in a change in the % secondary cooling

look at the 462-U bp
ever wonder why they did that ?
(so they will perform better than my testing will indicate as I address only the WCing path)

be cool

Agreed on the "dwarfing". I was more just throwing out some possible values and trying to stick a figure on it, and then attempt to credit/discredit the comment made by Steve Lynch that a more efficient waterblock could possibly overload the radiator and cause worse temperatures. It would take a radiator whose C/W is worse than that of the combined block/motherboard system for this to happen, so Steve's point is pretty much junk unless the radiator happens to be a 2in x 2in fin area heater-core (just a pie-in-the-sky guess).

Regarding the mobo cooling effect and block shapes, cooling the back of the motherboard where all the soldered socket pins poke through and therefore has about 5x the surface area very close to the CPU than anything on the front of the motherboard means that by far the more effective way to take advantage of the mobo cooling effect is to get a small amount of airflow happening behind the motherboard. Indeed, blowing air behind the motherboard yields far better "motherboard" cooling effects with far less necessary air-flow than cooling the front of the board.

I do agree about the block shape having an affect though for the front of the board, however in my experience with playing with this, this is only important when there is near zero air-flow about the socket area. When there is some forced convection angled at about 45 degrees striking the socket area then from I've been able to tell the effect is largely unconcerned with the size/shape of the block being used.

I find it very difficult to think that the heat generated by a CPU could overload a heatercore. Having much more knowledge in the automotive field than I do computers, I know from experience that many cars, when overheating, can be cooled down sufficiently by turning the heater on "High." It shold be noted, however, there is a much larger Delta T in an automotive application than exist in computer applications. But the fact that a standard heatercore with good convection can dissipate the heat generated by automotive combustion leads me to believe that the 50-150 watts generated by a cpu would be an easy load for a heatercore to dissipate.

I honestly don't know how many watts of energy a heater dissipates in a standard car, but it must be more than 100. A heater on full blast can warm up the average car in less than 5 minutes, which would lead me to believe that it is somewhere in the KiloWatt range.

I believe that we should all take from this the fact that for those of us trying to get that extra MHz, or 1 degree C, should look into cooling the backside of the motherboard better. The next time I'm in the shop I'll make a new motherboard mounting plate out of reinforced aluminum and cut out as much as possible other than where mounts need to go to test this out.

Nice topic Cathar.

Yes, a car has to dissapate alot of heat through the radiator.
Yes, it's ALOT of kW. A combustion engine is maybe around 30-40% effective at creating motional energy out of fuel, most of the rest is heat.... So taking an average car with 150 horsies we also get 70-80 kW of heat!. As for your analogy with heating up a car it's excellent. It goes quickly to heat with the cars heatercore. As I live in Sweden it's also common to use an electric car heaters so we don't have to waste time to get the windows de - iced. If we have around -10C outside it takes around 2-3 hours @ 1,5 kW to get them ice free.
I don't think there is a "point" where the heatercore is "overloaded" (unless the darn thing is melting!). It's just a matter of getting enough delta T to provide enough cooling. The more heat the block can shove to the water the better.

This is a tricky issue. First of all a car's heater core is putting
out more than 1 KW when the engine is up to temp. It also
has a great delta T to work with to drive the system.

As for secondary heat paths in CPU cooling there are two
basic issues. One, Can the readings even be counted upon,
and two, what is the relationship between the WB and this
secondary heat path. Clearly, the more efficiently the WB
system removes heat, the less will be dissipated thru
secondary paths. (Again the delta T driving force will be less
for the secondary paths.)

Could an efficient WB overload a heater core? Am I allowed
to laugh? This is not reasonable. The answer is: No!! Think
about the temps involved and then look as the heat transfer
capabilities for the expected flow rates.

As far as WB design, the most efficient blocks would be those
which transfer the most heat and "generate" the least amount
of heat from the flow direction changes and turbulence where
it does no good. From my other thread you can see that this
can amount to a few watts at good flow rates.

Cathar, your block does everything correctly. It is my favorite
block from a design POV. To even think that the extra heat
removed by your great block overloads a heater core is simply
silly. Where is the link to the guy who dreamed this up?

A more important question is this: Can we see a change in
equilibrium temps of the input water indicating that the
radiator is no longer "keeping up" with the increased load
of a more efficient WB?

BillA, you could add more tests to your busy day.

Tecumseh said:

Where is the link to the guy who dreamed this up?

Click to expand...

It's right there in the [H]ardOCP waterblock review, and is repeated numerous times in the infamous thread in the [H]ardforums by Steve Lynch in which about half of the forum readership got banned for even daring to suggest improvements. I'm sure BillA remembers it well.

I think the other half got banned after the P4 debacle. I'm not sure, but I think that new forum memberships are also still suspended.

Totally off-topic though.

The thread was not so much focussed on my block design but more on the "myth" and how it may apply to any block design.

Cathar said:

It's right there in the [H]ardOCP waterblock review, and is repeated numerous times in the infamous thread in the [H]ardforums by Steve Lynch in which about half of the forum readership got banned for even daring to suggest improvements. I'm sure BillA remembers it well.

I think the other half got banned after the P4 debacle. I'm not sure, but I think that new forum memberships are also still suspended.

Totally off-topic though.

The thread was not so much focussed on my block design but more on the "myth" and how it may apply to any block design.

Click to expand...

You will have to do better than that. Nobody in their right mind is
going to wade thru all that crap over there. I started and gave
up. Direct link please. OR just blow it off. The guy is dead wrong.

Hey, I couldn't even be bothered to find it again myself.

But since you pushed me, check out the second post of this thread:

http://www.hardforum.com/showthread.php?threadid=505078

Well, I'm not defending the guys, just wanna give you a lil perspective of where they are coming from. Here's a news blurb from Kyle the other day (dec 14, 1rst edition):
"Swifty Kit:
I have to admit I have seen a lot of water cooler kits as of late and this one really gets my attention. It looks as though Swiftech has put together a kit that will be worthy of my business and possibly yours. We need to get Steve beating on this immediately.


The H20-8500™ liquid cooling kits are designed to bring the efficiency, and quiet operations of liquid cooling to confined environments with little or no case modifications required. The kits are ideally suited for mid-tower cases, featuring at least one 80mm fan exhaust opening."

If you have a look at the kits, it's got a BI Micro Rad. So if you define "overwhelming the rad" as the system reaching an equalibrium point above what the cpu can take then I guess it could be possible. Maybe while Steve is "beating on" it he will manage to overload it! hehe.

peace.
unloaded

Unloaded, I think the point in dispute was not so much a terribly undersized radiator, which is almost as bad as sticking a passive heatsink atop a CPU, but whether it's possible for a waterblock to be more efficient than another waterblock, so much so that it pushes the radiator "over the edge".

As my opening maths shows, we're talking maybe 2W difference, or even if a really shoddy waterblock was used, maybe 5W. There's no way that an extra 5W into the water will make or break a system with an adequately provisioned radiator.

I'm curious to see though if they report the water temps with the BI Micro. I mean it says it's rated for 275W, but heck, I'd sure like to see what the water temps are when 275W of heat is pushed into it. If it's less than 20C above ambient, I'd be surprised. Hope they plug that 121W peltier in...

Cathar, if you look at the "Toward a real-world "C/W" correction..."
thread also in this "Technical Discussion" area you will see
that just circulating water thru a block can dissipate more
than 2 watts. We can "see" that this must be true, but
like Bill said, measuring it is next to impossible.

To even think that a more efficient block can push a radiator
"over the edge" is laughable.

I usually don't concern myself with dilettantes like you
know who, but he has done harm to those who don't
have a science or engineering background.

A far bigger issue concerns using peltiers for heaters. Yes,
it can be done. It has been done. There are a whole lot
of issues that must be worked out before you just bolt
one on and let it rip. I have reviewed several papers on this
very issue. The chance of dilettantes ever getting it right is
just about zero.

You have to recognize that when Steve says a kit "looks good" he means exactly that: it is aesthetically pleasing. He has said before that water temps are not relevant.

2-5C drop by pointing a fan at socket? I saw this with my Asus A7V-133's probe (the surface mount thermistor) but have never noticed this with actual diode reader. I will check this out shortly; I have my MAX6655evsys back out and on an Epox board now.

Im just gonna throw this out, I dont know If this is correct or not, and isnt totally on subject,

Different waterblocks have diff equlibriums rite? like in the water temperature?
I saw a 2c water drop when I dropped the direct die block on my cpu vs a regular copper block, but the cpu temp also went up 2c, that means it was 4c worse.

The only variable I don't think has been properly discussed here is the fact that if two different waterblocks have a signifigantly different pressure drop across them, the work done on the water overcoming this pressure drop will impart additional heat energy to the water. If you have the pump just upstream of the waterblock you are costing yourself a little bit of efficency, because the pump will heat the water some finite small amount. If the pump is downstream of the waterblock but upstream of the radiator, you could gain some fraction of a degree. Extremely high flow pumps would obviously be more suceptible to this problem.

No, I don't have the exact numbers. I forgot most of what I learned in my Fluids and Thermo courses way back when.

Gliderguy

the 'coolant heat' has been discussed in this thread
http://forum.oc-forums.com/vb/showthread.php?s=&threadid=142956

and since then Tecumseh and I have kinda hammered it into the ground by e-mail;
of slight interest wrt the cooling system (if the rad/airflow is marginal), effect on wb efficacy about zilch

be cool

you know im no genius. i possess no degrees of any source and i am a machinist and welder. i get all my math from the machinist handbook and one of my favorite measurement terms is called "the eyeball". the problem i have with that thread is this.

If a block dumps so much heat into the system that it overwhelms the cooling capability of the radiator

Click to expand...

well if the block was that outstandingly awesome that the radiator just stopped working.

1)the solder on the radiator would become bendable and flexible and the system would leak. i mean seiously it probably would considering the heatercore is meant to take in and get rid of a lot more heat than my computer can push. i don't know if this would really happen but ive never seen a waterblock that could whoop a heatercore.
2) why would that make it a bad waterblock? i mean if it could get rid of heat so well id buy two radiators and then 3 if need be. i mean damn thats a nice block. i mean the whole idea that a waterblock can do its job too well is ridiculous.
3)regular water would be boiling at that point. i know using water wetter raises the boiling temperature or something but there is no way that it could get that hot but i mean look at the reason they use water wetter in the car. to stop boilovers and whatnot. and that must mean the water is close to the boiling point hence the point that his processor would be dead. i mean if a heatercore could fail. but i think it would take nothing less than a flame or soldering iron. which i think solder melts at 300 or 900 degrees.

my 3 cents. don't bother correcting me if im wrong cause i probably am.. BTW cathar my copycat block works great// thanks for the idea