Monitoring, loop tuning and expansion

[evolvedpc]

... snip ... From there, we can indirectly control the amount of heat dissipated into the air by having better thermal transfer into the block (aka, a better mount). This does, however, mean that more total heatload from the cpu will enter the block, and if you follow the chain, more will enter the water. This is theoretical, I cannot say if it will work exactly like this irl, but in terms of theory, better mount = more energy dissipated into the water, which means higher water temperatures, and the increase in rad performance at that higher temp means that the water's equilibrium will only increase slightly. This warmer water is not reflected in cpu temps and actually will be reflected in LOWER cpu temps because the cpu is able to disperse more heat into the water.

Hmmm, who is right here probably doesn't really matter, but the discussion is interesting, so I will continue.

Will water temps increase if the block is mounted better? I continue to think they won't - at least not measurably. I will state, for the record, that you are undoubtedly correct about the water temp changing somewhat - I just don't think you will be able to measure it with the instruments we have available to us in the water-cooling world, most of which seem to work to an accuracy of about 0.5C.

I do agree with most of what you have said, but I think our two arguments rest on the delta between how much heat is transferred into the air/mobo/etc rather than into the water (via the block) for a poorly-mounted vs a well-mounted block. i.e. When the thermal path to the block/water is more resistant, is there significantly more heat transferred to places other than the water?

For you to be correct, the answer would have to be that yes, significant additional heat is radiated elsewhere, if the thermal path to the block is poor. For me to be correct, the difference in heat radiated elsewhere would have to be minimal. the answer, of course, comes down to how much more resistant the path to the water has become.

I think we can agree that conservation of energy dictates that the total amount of heat radiated will equal the total amount of heat produced by the chip. We know it's going somewhere, or the chip would shutdown/fry.

It just comes down to whether it goes via the water, or via the mobo/air.

What I reckon is happening, is that the chip can't dissipate heat to the air quickly enough to stop it from frying, so it heats up until it has enough of a temperature delta to overcome the increased thermal resistance of the path through the block.

With a poorly mounted block, let's say we see a cwdT increase of 15C over a well-mounted block. That means that the chip will be an additional 15C hotter than the surrounding air. Can 15C more make a significant difference to the amount of heat transferred to the air? This probably represents about a 50% increase in the chip/air delta, giving us a corresponding 50% increase in the amount of heat transferred via the air. (I think the relationship is roughly linear.)

So, we started (with a good mount) with Hw watts of heat going via the water and Ha watts of heat travelling via the air. We now have Ha * 1.5 travelling through the air and Hw - Ha*0.5 travelling through the water.

Let's say, for argument's sake that Ha is 1/24th of the amount travelling through the water. That is the difference between the thermal conductivity of water vs. air and assumes that the surface area of the water interface is the same as that of the air interface. (Which is strongly skewed in favour of the air interface - it won't really be anywhere near that much transferred to the air.)

Running with these figures, we start with 4.2% of our heat going via the air, for a well mounted block. Once we have cultivated a little algae, we end up with 6.3% via air. That (extremely generous) 2.1% of our heat that was diverted represents about 4.2W. (Assuming your CPU is chomping through 200W of power, which it might be, overclocked.)

The heat capacity of water is 4.179 at 30C. 4.2W = 15.12KJ/h. 15.12KJ will increase the temperature of 500ml of water by 7.23C. So, if we have 500ml of water in our loop, that 4.2W (if not dissipated by our rad at all) will increase our loop temp by 7.23C, over the course of an hour.

However, the story does not stop there. Looking at the skinnee labs review of the smallest radiator I can find on their site, we see that an increase in awdT of even 1C will increase the amount of heat radiated by about 15W, so 4.2W will cause an awdT increase of about 0.28C, as the rest will be radiated.

So, if we expect the thermal interface to the air to have the same area as that to the block, and if we manage a 15C improvement in block mounting, we just might see the water temps notch up past a half-degree increment.

Not sure if I made all of that entirely clear, but it seems that (in this extreme example) you might just be able to measure the difference in water temps, if you have a very accurate sensor, but probably not.

 

[m0r7if3r]

I couldn't agree more, I think where our difference is is what end of the spectrum we're on, I'm thinkin of this in terms of ideal or near ideal conditions whereas you are thinking in terms of real world conditions...if only we could get those ihs's better mounted. I think the actual change in water temp calculations are interesting...cause there's what, 3-5L in a loop...this, naturally, means that quality of mount will be affecting it well below the equipment we have access to's threshold, thus, I think (should we want to experiment) we would have to do something totally wrong and totally right, to draw out the differences. Maybe something along the lines of using cheese or mayo as a thermal paste vs some indigo xtreme and monitor water temps (heck, if someone's willing to risk the gear, you could mount it dry...), that might get it within the range of the equipment that we have at our disposal.

 

[Bobnova]

Quite personally, i wouldn't worry about the mobo dissipated heat. With a stock cooler (read: lousy, hot) actively cooling the back side of the mobo(via thermalright's mobo cooler) lowers core temps maybe 4*c to 6*c.
With a good air cooler, actively cooling the back side of the mobo results in .5*c or less change.
Given that a water cooling loop worth building will equal a good air cooler, the mobo isn't getting rid of much at all, not with little to no airflow and no contact with a heatsink.

I would measure the waterblock temp rather then the water temp, personally. Maybe drill a hole like the extreme benchers do.
The water temp doesn't seem relevant unless you're wondering about the efficiency of individual loop parts, rather then the loop as a whole.



As far as the OP goes, that is easy.
Run the pump at full blast, and turn the fans down.
Or for quietness: Turn the fans off (at idle!) and then drop the pump speed till you can't hear it, then turn the fans so you can just barely hear them, and then adjust your OC to fit that amount of cooling.

 

[m0r7if3r]

Quite personally, i wouldn't worry about the mobo dissipated heat. With a stock cooler (read: lousy, hot) actively cooling the back side of the mobo(via thermalright's mobo cooler) lowers core temps maybe 4*c to 6*c.
With a good air cooler, actively cooling the back side of the mobo results in .5*c or less change.
Given that a water cooling loop worth building will equal a good air cooler, the mobo isn't getting rid of much at all, not with little to no airflow and no contact with a heatsink.

I would measure the waterblock temp rather then the water temp, personally. Maybe drill a hole like the extreme benchers do.
The water temp doesn't seem relevant unless you're wondering about the efficiency of individual loop parts, rather then the loop as a whole.



As far as the OP goes, that is easy.
Run the pump at full blast, and turn the fans down.
Or for quietness: Turn the fans off (at idle!) and then drop the pump speed till you can't hear it, then turn the fans so you can just barely hear them, and then adjust your OC to fit that amount of cooling.

While that's all very practical and well and good, this thread is more (at least now) about chewing the fat over potential ways to increase effectiveness of cooling and the reasons behind that.

oh, and I listed heat going into the mobo just for completeness, realisitically, the crap (and I do mean crap, stuff is junk) that they make the boards out of is pretty thermally insulative, if i had to guess, i'd say more so than air, but regardless, its part of the system, and thus heat is transferred into it.

as to the waterblock vs water temp, you present an interesting point, maybe archer could do some measurements at different points on the water block, maybe top of the block vs cpu surface, see how much heat makes it to the top of the ihs...and then how much heat is going into the block, it'd probably be a statement to the effectiveness of the thermal paste.

 

[evolvedpc]

I couldn't agree more, I think where our difference is is what end of the spectrum we're on, I'm thinkin of this in terms of ideal or near ideal conditions whereas you are thinking in terms of real world conditions...if only we could get those ihs's better mounted. I think the actual change in water temp calculations are interesting...cause there's what, 3-5L in a loop...this, naturally, means that quality of mount will be affecting it well below the equipment we have access to's threshold, thus, I think (should we want to experiment) we would have to do something totally wrong and totally right, to draw out the differences. Maybe something along the lines of using cheese or mayo as a thermal paste vs some indigo xtreme and monitor water temps (heck, if someone's willing to risk the gear, you could mount it dry...), that might get it within the range of the equipment that we have at our disposal.

Er, that's not quite what the figures show. The quality of the mount is very significant to chip temps. (With a dry mount - which I effectively did last week, by accident, to my GPU - you get the chip's auto-cut-off kicking in as soon as you put any load through it.) It's just that the path from the chip direct to the air is so thermally resistant that you'll have the chip overheating and cutting out before any real heat goes that way. The (minimal) extra wattage that is diverted into the water with a better mount is radiated away before it can raise water temps measurably.

So, you are dead right, in that some heat is diverted. And I am right in that it is not going to raise water temps, measurably. But cwdT does remain a significant factor in loop design, as it does affect chip temps - just not water temps.

If you look at Skinnee Labs' thermal paste comparisons, you will see that different pastes can make several degrees of difference to chip temps - about a 3C spread between the best and worst pastes. You will also see that his water block reviews typically use figures from 5 different mounts of the same block. He lists figures that tend to vary by a degree or so, from mount to mount.

 

[m0r7if3r]

Hah, I was reading skinnee's tim reviews for the 3rd time just now...probably as you wrote it...I'm curious as to how you could get the chip to more effectively transfer heat into the water...naturally, more surface area...maybe if you made the block taller with tiny micropins or something like that...maybe zig-zagged micropins or if you could do grooves on the micropins...i'm just thinking out loud here, jump in with ideas. Anyone know if remounting the IHS is plausible? I'm interested as to how it's attached, I'm gonna see if i can get my hands on a dead processor and rip it apart.

 

[Radeon28]

Hah, I was reading skinnee's tim reviews for the 3rd time just now...probably as you wrote it...I'm curious as to how you could get the chip to more effectively transfer heat into the water...naturally, more surface area...maybe if you made the block taller with tiny micropins or something like that...maybe zig-zagged micropins or if you could do grooves on the micropins...i'm just thinking out loud here, jump in with ideas. Anyone know if remounting the IHS is plausible? I'm interested as to how it's attached, I'm gonna see if i can get my hands on a dead processor and rip it apart.

Remounting the IHS is indeed plausible. My friend took his IHS completely off when he lapped it, and then remounted it. He also took the IHS off of one of his CPUs and put the WB directly onto the cores (which I am temped to do cause one of my cores runs at 56C under load... with high end watercooling on it. I'm sure if someone really wanted to they could integrate the IHS and WB... but that would require talent and some expensive tools.

 

[m0r7if3r]

wonder if intel could do something equivalent to eVGA's FTW edition where they do that schindig for you...what were the results from the direct mount, do you know? I'm tempted to do some testing once i get my new board in, cause i've got a 775 processor i'm never gonna need again and a stock aircooler to go with it that i'm not at all concerned to waste...

 

[Radeon28]

wonder if intel could do something equivalent to eVGA's FTW edition where they do that schindig for you...what were the results from the direct mount, do you know? I'm tempted to do some testing once i get my new board in, cause i've got a 775 processor i'm never gonna need again and a stock aircooler to go with it that i'm not at all concerned to waste...

He said the direct contact dropped his temps 10C, and that his CPU was only 15C more than ambient temp. (lets take that with a grain of salt, cause he may have boasted a little there, though it is plausible). I have a socket 939 processor which is running a little hot, and I'm tempted to give it a shot (it should increase the heat transfer tremendously... though I'm worried about electrical current coming off of the cores and going into the WB... this would MEGA accelerate galvanic rust... even if the slightest amount of aluminum was in the loop, the galvanization would happen 100s of times faster with electrical current introduced... the surge could also hurt the pump.

Can anyone clarify if a metal object directly touches a core conduction will occur???

Thermal paste would be a whole other part to this equation, since thermal paste is conductive.. is it safe to put it on the cores, or are we going to have to go direct metal to metal mounting?

 

[Bobnova]

Non-soldered intel cpus drop anywhere from 5*c to 25*c with the lid removed.
Soldered intels drop 3*c to 6*c.

So if your friend pulled the lid off a non-soldered one, 15*c is totally believable.

 

[m0r7if3r]

i have found this for the i7 920. Guy's reporting 30c loaded at 4ghz, ht enabled. I don't do processor design (sorry ) but I know a guy, I'll get back to you guys on this.

 

[evolvedpc]

He said the direct contact dropped his temps 10C, and that his CPU was only 15C more than ambient temp. (lets take that with a grain of salt, cause he may have boasted a little there, though it is plausible). I have a socket 939 processor which is running a little hot, and I'm tempted to give it a shot (it should increase the heat transfer tremendously... though I'm worried about electrical current coming off of the cores and going into the WB... this would MEGA accelerate galvanic rust... even if the slightest amount of aluminum was in the loop, the galvanization would happen 100s of times faster with electrical current introduced... the surge could also hurt the pump.

Can anyone clarify if a metal object directly touches a core conduction will occur???

Thermal paste would be a whole other part to this equation, since thermal paste is conductive.. is it safe to put it on the cores, or are we going to have to go direct metal to metal mounting?

Use Arctic MX-3, if you are worried about conductivity - it hasn't got any. It also performs pretty well - although not quite as well as Shin Etsu or Indigo Xtreme. I use it and I'm happy with it - but you do need to watch out for dud tubes; I have had 2 tubes where the paste was solidified. (Both bought from the same place, on the same order.)

 

[evolvedpc]

i have found this for the i7 920. Guy's reporting 30c loaded at 4ghz, ht enabled. I don't do processor design (sorry ) but I know a guy, I'll get back to you guys on this.

That's a pretty cool thread - I wonder if it could be made to work with a standard WB?

 

[m0r7if3r]

disclaimer:not responsible for processor damage!

So, I think its time to move away from ihs removal and see what else we can think of to bring down dt_cw

 

[QuietIce]

I have a socket 939 processor which is running a little hot, and I'm tempted to give it a shot (it should increase the heat transfer tremendously... though I'm worried about electrical current coming off of the cores and going into the WB... this would MEGA accelerate galvanic rust... even if the slightest amount of aluminum was in the loop, the galvanization would happen 100s of times faster with electrical current introduced... the surge could also hurt the pump.

Can anyone clarify if a metal object directly touches a core conduction will occur???

Thermal paste would be a whole other part to this equation, since thermal paste is conductive.. is it safe to put it on the cores, or are we going to have to go direct metal to metal mounting?

I've de-lidded a couple of Opty's with no problems at all. The core itself is not an issue but the electronic components around the core (once the lid is off) is a concern so if you use AS5 or other conductive TIM be careful with how much you use. It only takes a tinny bit of TIM for an Opty core. You should see temp drops from 3-7°C depending on your cooling solution. I've got two Opty's under Storms (see sig), one is lidded, the other is not. The lidded Opty runs ~5°C hotter than it's twin with the same vCore ...

 

[Owenator]

I don't think de-lidding is the answer here. Solder is a much better TIM than any paste.

This has some good graphics and an explanation of the thermo behind water cooling.
http://www.electronics-cooling.com/...ures-in-a-water-to-air-hybrid-cooling-system/

 

[QuietIce]

I don't think de-lidding is the answer here. Solder is a much better TIM than any paste.

This has some good graphics and an explanation of the thermo behind water cooling.
http://www.electronics-cooling.com/...ures-in-a-water-to-air-hybrid-cooling-system/

Socket 939 CPUs are not soldered nor would I recommend de-lidding a soldered CPU.
(I would never try to de-lid my Phenom II. )

s939's just have an internal paste between the aluminum IHS and the chip. The IHS is relatively easy to remove with a razor blade and patience, you just have to cut (not too deeply!) through the rubbery adhesive they use on it. Or maybe the attachment method includes a very thin rubber gasket around the edges - it's hard to tell when it's been shredded by a razor blade ...

 

[Owenator]

Socket 939 CPUs are not soldered nor would I recommend de-lidding a soldered CPU.
(I would never try to de-lid my Phenom II. )

s939's just have an internal paste between the aluminum IHS and the chip. The IHS is relatively easy to remove with a razor blade and patience, you just have to cut (not too deeply!) through the rubbery adhesive they use on it. Or maybe the attachment method includes a very thin rubber gasket around the edges - it's hard to tell when it's been shredded by a razor blade ...

I agree totally! I was referring to the previous link where the person did actually de lid a soldered on CPU. Seems counter productive to me.

What we really need is a way to increase the spread of the heat from the CPU to the waterblock. I read about a company that was making what looked like a flat copper plate but was actually hollow and full of fluid. The idea beign that the heat spread very fast from athe cpu die to the entire surface making it easier to cool. It works because water vaporizing removes heat very fast. Then it moves around to spread the heat. Quite a "cool" idea. Not sure if they sell them yet. http://www.novelconceptsinc.com/heat-spreaders.htm

 

[QuietIce]

I agree totally! I was referring to the previous link where the person did actually de lid a soldered on CPU. Seems counter productive to me.
No doubt!

What we really need is a way to increase the spread of the heat from the CPU to the waterblock. I read about a company that was making what looked like a flat copper plate but was actually hollow and full of fluid. The idea beign that the heat spread very fast from athe cpu die to the entire surface making it easier to cool. It works because water vaporizing removes heat very fast. Then it moves around to spread the heat. Quite a "cool" idea. Not sure if they sell them yet. http://www.novelconceptsinc.com/heat-spreaders.htm

Great idea! - but they'll need to upsize it to work on CPUs.

The planar capillary in the IsoSkin heat spreader has such high liquid transport capacity, that additional internal microstructures (wick) are not required for applications having power densities less than 20 watts per square centimeter.

I think our CPUs exceed that limit ...

 

[Owenator]

Great idea! - but they'll need to upsize it to work on CPUs. I think our CPUs exceed that limit ...

I thought there was some with internal wicking structure but I'm not sure. IIRC the time I emailed with an engineer from there they were saying that they were working on ones more inline with our CPU power densities.

The problem a lot of times is that we want cooler to overclock whereas a CPU thermal engineer sets the heat load then selects the cooling. We are going from it a little differently where we want to increase the cooling so we can increaset he power/voltage. And the vendors make what engineers ask for becasue they have the $$. Maybe if we pooled our money we could get them to develep one?