water cooling vs heatsink cooling vs any other type????

[BonzaiDuck]

Aigomorla said: "WAIT! your using nvidia monitor to messure temps? "

That's only one program that I use. I check my BIOS monitor readings against SpeedFan, Asus Probe, Everest, MBM and other sources. When I can, I get a temperature monitor with tape-on thermal sensors, but of course, these temperatures are limited to the choice or opportunities to mount them.

I brought Nvidia Monitor into the discussion because that's what you show your temperatures with in this link you provided in an earlier post:

Your screenie capture

The case I'm building should have plenty of room for a pretty good water-cooling rig, and I could even "go there" eventually. In fact, the modding work I'm doing for this case includes that option. But I want to see what I can squeeze out of air -- in addition to the touted ThermalRight "Extreme" performance.

I'll be posting some things I do with the air-cooling in the next few weeks. The mods themselves can just as easily be applied to water-cooling blocks and water-cooling setups.

 

[aigomorla]

Originally posted by: BonzaiDuck

Aigomorla said: "WAIT! your using nvidia monitor to messure temps? "

That's only one program that I use. I check my BIOS monitor readings against SpeedFan, Asus Probe, Everest, MBM and other sources. When I can, I get a temperature monitor with tape-on thermal sensors, but of course, these temperatures are limited to the choice or opportunities to mount them.

I brought Nvidia Monitor into the discussion because that's what you show your temperatures with in this link you provided in an earlier post:

Your screenie capture

The case I'm building should have plenty of room for a pretty good water-cooling rig, and I could even "go there" eventually. In fact, the modding work I'm doing for this case includes that option. But I want to see what I can squeeze out of air -- in addition to the touted ThermalRight "Extreme" performance.

I'll be posting some things I do with the air-cooling in the next few weeks. The mods themselves can just as easily be applied to water-cooling blocks and water-cooling setups.

Theres currently only 2 ways to tweek the ultra120.

You either lap the base so you get rid of the bow. Or you find better mounts so you can compress that bow.

You pull off either one of those, you'll see very happy temps.

 

[bfdd]

goto newegg.com buy a Scythe Ninja Rev B Plus for 40 dollars and if you want to overclock it'll support your needs, unless you planned on going batsh!t crazy which you're going to need watercooling or better.

 

[BonzaiDuck]

AigoMorla

I wanted to keep the lid on this until I can see how well it works, but you've provoked me.

Yes, the mainstream ways of "modding" the U-120 or Extreme involve grinding away the ridge-line in TR's conscious design-decision to make a convex heatsink base. TR tech-reps "Bob" and "Hank Peng" told me they chose to do this because it was felt that the heatsink base only needed to make contact with the heat-spreader where the two processor cores resided. However, the concept of thermal resistance also includes a factor of areal-size, so instead of measuring it as just C-degrees / thermal-Watts, it is sometimes expressed as:

TR = C / (m^2 * W)

or in square centimeters, or square inches. Since processor sizes have shrunk, this has been an emerging issue in using TEC plates and in the effectiveness of TIM or thermal greases.

Since the heat would spread over the heatspreader-cap, even as it is being wicked away by the Ultra-120 convex base, it would seem that the thermal resistance of the heatspreader itself would be a factor in some of that heat continuing to spread over the heatspreader to warm it up a tad, so more comprehensive contact with the heatspreader may have some positive effect on temperatures and thermal resistance overall.

Because it might be prefered to build a jig to lap the base well and consistently (given its irregular shape to begin with) -- a jig that might be made from small lengths of hardwood -- and because I discovered that SVC is offering a "custom-lapped" U-120-Extreme for $30 more, I decided simply to "outsource" the lapping to them.

We discussed it over the phone. It took them a day or so to define for themselves how to build a jig so they could do it swiftly and accurately. They take off one millimeter at the ridge-line, or high-point on the base.

This of course leaves a mirror-polished copper surface. TR stipulates that their warranty is voided if you grind away the nickel-plate, which, they say, insures the "integrity" of the solder-joints between the heat-pipes and the base. I discussed this with the SVC tech-rep, and we mutually came to the conclusion that there is a chance -- eventually -- of corrosion around the edge of the nickel-plate where the copper is exposed, risking that the nickel might begin to peel away. But one could simply put a thin layer of thermal epoxy around that boundary to prevent such things from happening.

My friend across town bought the U-120-Extreme I had waiting in the retail box before I purchased a custom-lapped cooler from SVC, so no loss . . . .

But here is the basis of some improvements I intend to make:

January, '07 OverClocker's article

One reseller's current offering, about equal to AS5

[With 10% content, the Taiwanese company is hoping to minimize cost, capture more mainstream consumers, and take away Arctic Silver's market, since their formulation barely keeps up with AS5.]

One of several suppliers whose traditional customers work with cutting wheels and machinery

"Grit #1," their finest powder, is graded at a particulate-size of less than 2 microns, and the distribution of particle sizes ranges from 0 to 2 microns. If the particle sizes are normally or horizontally distributed, then figure an average particle-size of 1 micron -- or 0.00004 inches. By comparison, the size of silver particles in AS5 ranges between 0 and 0.49 microns. This is not a big difference in average particulate-size.

WHAT THAT MEANS, IN COMPARISON TO Arctic Silver's OWN BENCH-TESTS

AS rates their AS5 compound on a controlled application of the paste with a thickness of 0.001 inch. This would be equivalent to piling 25 1-micron "abrasive" particles on top of each other.

COST

$4 per carat; 5 carats per gram.
Minimum order-quantity = 5 grams.

I'll report on the coverage, but I have plans to split my order with one other person. If there is ample dust left over, then it will be split three ways -- participants already chosen.

SECOND-BEST ADDITIONS OF COMPLEXITY

http://www.overclockers.com/tips188/

Two months ago, I visited a handicapped friend to assist in setting up shared internet access so that his caretaker-brother could perform shopping and other tasks on the web without interrupting my friend's efforts at writing a comic-crime-novel using Dragon Naturally Speaking. I was rewarded with a brand-new 2007 US Mint-issue silver dollar -- 99.9% "fine silver." I emphatically stated that I wasn't a coin-collector, and that I was immensely grateful, intending to put the item to good use.

Remember that SVC grates off 1-millimeter of the U-120-Extreme's heatsink base. I share a view with others that adding 2.96 millimeters of soft metal to the bottom of the base is not going to add too much in pressure resulting from further compression of the mounting-screw springs, but consider this. The "Lady Liberty," Bald Eagle, raised-rim and raised-printing on the coin have a depth of about 0.35 mm. [I splurged to add a digital caliper/micrometer with depth-gauge to my tool-collection.] That means that lapping should result in removal of at least 0.70 mm, leaving a slug that is approximately 2.26 mm thick. Subtract the 1mm removed by SVC, and we're adding 1.26mm (or less) to the heatsink base.

But what good does this do? It adds, at most, 30 grams of weight to the heatsink base with virtually no increase in torque against the motherboard. This would mean that the resulting "composite base" now has an overall thermal resistance component that is lower. But the increase in weight is only about 3/80 or about 4%. Second, it introduces another thermal interface, where there was previously only one. It hardly seems worth it. But wait!

Of the two substances, copper and silver, they are both "malleable," but silver is the softer of the two (I believe). Suppose, then, that you either "roll" the slug into a 1mm thickness (providing enough of it to use on graphics cards, etc.) or just stick it between two, polished, flat hard-steel plates, put the sandwich on an anvil, and pound it a little, apply the digital caliper, pound some more . . . etc.? But then again, what good does THAT do?

Add a fraction of a gram of the abrasive powder to one of the steel plates; spread it around on an area equal to the area of the silver slug; lay the slug on top of the dust (carefully, so your $4/carat investment doesn't get "blown away"); add the other plate; put it on an anvil; give it a couple of light taps.

Then, the calculating risk-taker has two choices: load up some CK4800 paste with more abrasive powder, and assume that the pressure between the heatsink-base and the processor cap is sufficient to keep the slug in place (risk #1); or mix some more powder with one component of either ASTA thermal epoxy, or Arctic Alumina epoxy, and then mix the epoxy, spread it on the heatsink base thinly, get the slug on the paste before it's too late (risk #2), and clamp it with some very small clamps, thus making it impossible to return the Ultra-120-Extreme to its previous "retail-package" state if things don't "work out well" (risk #3).

Of course, one could always use some Gorilla Glue beads to secure the slug to the heatsink base after applying a 50%-abrasive (my code-word for . . . you know . . ) enhancement to CK4800 or AS3 or just regular ThermalRight thermal grease.

Here are the relative thermal conductivities of copper, silver . . . and, ah . . . . abrasive powder:

Cu = 400
Ag = 430
"crystal Carbon" = 1000 to 2000 (another code-word )

One would think that the third item would simply make "thermal interfaces" disappear in the relative sense of thermal resistance and thermal conductivity. Further, the Ag shim just becomes a vehicle for applying more "crystal dust," so only a 1mm shim would be necessary.

My local coin dealer sells 2 gram ingots of Silver for $1 each -- they're about a millimeter thick, and could be "rolled" or pounded into a thickness of 0.5 mm. So you don't even need to destroy a $35+ collector's item for its silver-content.

That leaves my "ducting" mod, and I'm going to cut and fit the panels over the next week or two. Which means, in turn, if you want to wait and see how this all turns out, you'll just have to wait . . .

Otherwise, for you risk-takers and spend-thrifts, don't e-mail Penn-Scientific -- call them. Ask for "Diane." They take credit-cards. She is also very pleasant to talk to.

For me, a vial of white-powder arrives by mail in the next few days.

And that's not something anyone would want to snort up their nose, either!

SUMMARY EMPHASIS

Recent news and white-papers in the industry point to the thermal interface between heatsink (or water-block -- Yeah, Aigomorla, Yeah!!) as critical to the future of the industry and to the future of us "enthusiasts." A lady professor at a prestigious university has shown that carbon-black can be made to surpass thermal resistance and conductivity properties of silver and other paste-components (with our "abrasive" exception), but the pressure needed to make it work is far too great for current heatsink applications (and for the life or operability of our expensive components).

Now -- I have some questions. I believe that economic reasoning and strategy have limited the offering by JetArt to the 10% formulation. Yet I have not seen anyone else attempting to produce a retail TIM with the easily-reached limit of 2.5 times the thermal conductivity of AS5. Citarella estimates the cost of a well-loaded TIM using the abrasive substance in question at about $50 for 10 applications.

What am I missing here? Or is this just something that has been "overlooked" in "the community?"

We're gonna find out soon.

 

[nealh]

Originally posted by: BonzaiDuck
AigoMorla

I wanted to keep the lid on this until I can see how well it works, but you've provoked me.

Yes, the mainstream ways of "modding" the U-120 or Extreme involve grinding away the ridge-line in TR's conscious design-decision to make a convex heatsink base. TR tech-reps "Bob" and "Hank Peng" told me they chose to do this because it was felt that the heatsink base only needed to make contact with the heat-spreader where the two processor cores resided. However, the concept of thermal resistance also includes a factor of areal-size, so instead of measuring it as just C-degrees / thermal-Watts, it is sometimes expressed as:

TR = C / (m^2 * W)

or in square centimeters, or square inches. Since processor sizes have shrunk, this has been an emerging issue in using TEC plates and in the effectiveness of TIM or thermal greases.

Since the heat would spread over the heatspreader-cap, even as it is being wicked away by the Ultra-120 convex base, it would seem that the thermal resistance of the heatspreader itself would be a factor in some of that heat continuing to spread over the heatspreader to warm it up a tad, so more comprehensive contact with the heatspreader may have some positive effect on temperatures and thermal resistance overall.

Because it might be prefered to build a jig to lap the base well and consistently (given its irregular shape to begin with) -- a jig that might be made from small lengths of hardwood -- and because I discovered that SVC is offering a "custom-lapped" U-120-Extreme for $30 more, I decided simply to "outsource" the lapping to them.

We discussed it over the phone. It took them a day or so to define for themselves how to build a jig so they could do it swiftly and accurately. They take off one millimeter at the ridge-line, or high-point on the base.

This of course leaves a mirror-polished copper surface. TR stipulates that their warranty is voided if you grind away the nickel-plate, which, they say, insures the "integrity" of the solder-joints between the heat-pipes and the base. I discussed this with the SVC tech-rep, and we mutually came to the conclusion that there is a chance -- eventually -- of corrosion around the edge of the nickel-plate where the copper is exposed, risking that the nickel might begin to peel away. But one could simply put a thin layer of thermal epoxy around that boundary to prevent such things from happening.

My friend across town bought the U-120-Extreme I had waiting in the retail box before I purchased a custom-lapped cooler from SVC, so no loss . . . .

But here is the basis of some improvements I intend to make:

January, '07 OverClocker's article

One reseller's current offering, about equal to AS5

[With 10% content, the Taiwanese company is hoping to minimize cost, capture more mainstream consumers, and take away Arctic Silver's market, since their formulation barely keeps up with AS5.]

One of several suppliers whose traditional customers work with cutting wheels and machinery

"Grit #1," their finest powder, is graded at a particulate-size of less than 2 microns, and the distribution of particle sizes ranges from 0 to 2 microns. If the particle sizes are normally or horizontally distributed, then figure an average particle-size of 1 micron -- or 0.00004 inches. By comparison, the size of silver particles in AS5 ranges between 0 and 0.49 microns. This is not a big difference in average particulate-size.

WHAT THAT MEANS, IN COMPARISON TO Arctic Silver's OWN BENCH-TESTS

AS rates their AS5 compound on a controlled application of the paste with a thickness of 0.001 inch. This would be equivalent to piling 25 1-micron "abrasive" particles on top of each other.

COST

$4 per carat; 5 carats per gram.
Minimum order-quantity = 5 grams.

I'll report on the coverage, but I have plans to split my order with one other person. If there is ample dust left over, then it will be split three ways -- participants already chosen.

SECOND-BEST ADDITIONS OF COMPLEXITY

http://www.overclockers.com/tips188/

Two months ago, I visited a handicapped friend to assist in setting up shared internet access so that his caretaker-brother could perform shopping and other tasks on the web without interrupting my friend's efforts at writing a comic-crime-novel using Dragon Naturally Speaking. I was rewarded with a brand-new 2007 US Mint-issue silver dollar -- 99.9% "fine silver." I emphatically stated that I wasn't a coin-collector, and that I was immensely grateful, intending to put the item to good use.

Remember that SVC grates off 1-millimeter of the U-120-Extreme's heatsink base. I share a view with others that adding 2.96 millimeters of soft metal to the bottom of the base is not going to add too much in pressure resulting from further compression of the mounting-screw springs, but consider this. The "Lady Liberty," Bald Eagle, raised-rim and raised-printing on the coin have a depth of about 0.35 mm. [I splurged to add a digital caliper/micrometer with depth-gauge to my tool-collection.] That means that lapping should result in removal of at least 0.70 mm, leaving a slug that is approximately 2.26 mm thick. Subtract the 1mm removed by SVC, and we're adding 1.26mm (or less) to the heatsink base.

But what good does this do? It adds, at most, 30 grams of weight to the heatsink base with virtually no increase in torque against the motherboard. This would mean that the resulting "composite base" now has an overall thermal resistance component that is lower. But the increase in weight is only about 3/80 or about 4%. Second, it introduces another thermal interface, where there was previously only one. It hardly seems worth it. But wait!

Of the two substances, copper and silver, they are both "malleable," but silver is the softer of the two (I believe). Suppose, then, that you either "roll" the slug into a 1mm thickness (providing enough of it to use on graphics cards, etc.) or just stick it between two, polished, flat hard-steel plates, put the sandwich on an anvil, and pound it a little, apply the digital caliper, pound some more . . . etc.? But then again, what good does THAT do?

Add a fraction of a gram of the abrasive powder to one of the steel plates; spread it around on an area equal to the area of the silver slug; lay the slug on top of the dust (carefully, so your $4/carat investment doesn't get "blown away"); add the other plate; put it on an anvil; give it a couple of light taps.

Then, the calculating risk-taker has two choices: load up some CK4800 paste with more abrasive powder, and assume that the pressure between the heatsink-base and the processor cap is sufficient to keep the slug in place (risk #1); or mix some more powder with one component of either ASTA thermal epoxy, or Arctic Alumina epoxy, and then mix the epoxy, spread it on the heatsink base thinly, get the slug on the paste before it's too late (risk #2), and clamp it with some very small clamps, thus making it impossible to return the Ultra-120-Extreme to its previous "retail-package" state if things don't "work out well" (risk #3).

Of course, one could always use some Gorilla Glue beads to secure the slug to the heatsink base after applying a 50%-abrasive (my code-word for . . . you know . . ) enhancement to CK4800 or AS3 or just regular ThermalRight thermal grease.

Here are the relative thermal conductivities of copper, silver . . . and, ah . . . . abrasive powder:

Cu = 400
Ag = 430
"crystal Carbon" = 1000 to 2000 (another code-word )

One would think that the third item would simply make "thermal interfaces" disappear in the relative sense of thermal resistance and thermal conductivity. Further, the Ag shim just becomes a vehicle for applying more "crystal dust," so only a 1mm shim would be necessary.

My local coin dealer sells 2 gram ingots of Silver for $1 each -- they're about a millimeter thick, and could be "rolled" or pounded into a thickness of 0.5 mm. So you don't even need to destroy a $35+ collector's item for its silver-content.

That leaves my "ducting" mod, and I'm going to cut and fit the panels over the next week or two. Which means, in turn, if you want to wait and see how this all turns out, you'll just have to wait . . .

Otherwise, for you risk-takers and spend-thrifts, don't e-mail Penn-Scientific -- call them. Ask for "Diane." They take credit-cards. She is also very pleasant to talk to.

For me, a vial of white-powder arrives by mail in the next few days.

And that's not something anyone would want to snort up their nose, either!

SUMMARY EMPHASIS

Recent news and white-papers in the industry point to the thermal interface between heatsink (or water-block -- Yeah, Aigomorla, Yeah!!) as critical to the future of the industry and to the future of us "enthusiasts." A lady professor at a prestigious university has shown that carbon-black can be made to surpass thermal resistance and conductivity properties of silver and other paste-components (with our "abrasive" exception), but the pressure needed to make it work is far too great for current heatsink applications (and for the life or operability of our expensive components).

Now -- I have some questions. I believe that economic reasoning and strategy have limited the offering by JetArt to the 10% formulation. Yet I have not seen anyone else attempting to produce a retail TIM with the easily-reached limit of 2.5 times the thermal conductivity of AS5. Citarella estimates the cost of a well-loaded TIM using the abrasive substance in question at about $50 for 10 applications.

What am I missing here? Or is this just something that has been "overlooked" in "the community?"

We're gonna find out soon.

ok..I am lost what are you suggesting..a custom lap job and adding silver to the HS base???

 

[ChefJoe]

Sounds like silver base plate and a silver/diamond version of arctic silver smear. Sounds very interesting.

 

[BonzaiDuck]

nealh --

True, I "write a lot." You're correct about the silver shim.

But look at the first and third links in my post.

Or -- put the print-side of your two thumbs together; then put the print-side of your two index fingers together.

What shape is that, Nealh?!!

 

[aigomorla]

*sigh*

I give up. At times like this i just keep quiet. If you want to live in your own little world where your air heat sink is better then my water, then so be it.

But if you ask around, you'll see very quickly, people who do a proper migration to water, rarely go back to air.

Thats all i need to say.

 

[BonzaiDuck]

Well, aigomorla, I've seen people try water and "go back" to air. I've seen others purchase an expensive Swiftech water-block, build an external reservoir from PVC pipe, put together a fan-and-radiator assembly -- and then just let it all sit on a shelf unused.

THG has a recent series of articles on over-clocking, and in the enthusiast community there seems to be a categorization of people as either "casual over-clockers" or "extreme over-clockers."

I don't think of myself as "casual," but I'm not "extreme." I've posted my philosophy earlier in regards to what we all do. I've got more of an interest in balancing hardware and defeating bottlenecks to get "good performance," but not extreme performance.

Heatpipe coolers are a "sort of" water-cooling -- but no reservoir or chiller. One might say that at the "cool" end, they're a sort of crude radiator. But I think the prevailing judgment is consistent with yours -- that water is better.

I think you overlooked my point in my last post. What I'm trying to do with a heapipe cooler and ducting is equally applicable to water-cooling rigs. So any improvement I get in the air-cooling category can quickly or easily be adopted to a water-cooling setup, and the water-coolers "win again."

Waterhoses would complicate my ducting strategy, but in taking extra trouble, the ducting strategy could enhance cooling of components that waterblocks miss.

But for water-cooling or air-cooling, I am still stymied by an absence of superior diamond based thermal paste. Why 10% in the JetArt product, when it could be 30 or 40 or considerably more? It must be cost, or the company's own strategy to minimize their own cost while marketing to a more "mainstream" consumer segment.

 

[nealh]

Originally posted by: BonzaiDuck
nealh --

True, I "write a lot." You're correct about the silver shim.

But look at the first and third links in my post.

Or -- put the print-side of your two thumbs together; then put the print-side of your two index fingers together.

What shape is that, Nealh?!!

HEHE..its a girl's best friend....

Looking forward to what you get

 

[BonzaiDuck]

I've noticed that with heatpipe coolers, they seem to "cycle" in proportion to the increase in temperatures between idle and load. There are perhaps a lot of indeterminate factors in how the sensors read temperatures, but I'd like to think my speculations about the phenomenon are in the ball park.

Unless the sensor-readings are due to momentary fluctuations in CPU usage, I see the temperatures spike and drop on a regular basis under SpeedFan. Certainly, there is also a sampling rate for the recording of the temperatures, but it "seems" as though the heatpipe coolant builds up for a moment at the hot end, and then vaporizes to release some heat. This could all be my imagination, for one might think it is a continuous process under some sort of equilibrium -- I can't be sure.

So with a serious reduction of thermal resistance at the interface between heat-spreader and heatsink base, you'd think (perhaps) that this "process" I imagine would just keep up with the more efficient removal of thermal energy from the cores and processor cap. Then, the ultimate limitation would be the aluminum fins and how increased fan-speed reaches some limit of effectiveness.

We are certainly going to find out, though . . .

 

[drakore]

Ok after reading all of this, it looks like Bonzai only refers to Swiftech kits. Now, to the original arguement between you and aigomorla saying that air can match water, this is not possible under loads... At the same ambient temperatures, the water cooling setup will win. Even with a massive heat sink the contact point is still fairly small, this being limited by the CPU slug size. As the heat dissapites, something called spread resistance occurs. Using water will limit this spread resistance because the heat doesn't have to travel as far through the waterblock to be swept away from the incoming cooled water.

Now, if you are going to compare water to air, you must take into account the item being cooled. Do you really think it is possible to match water with air on an overclocked quad core dissipating ~110W (Please correct me if that is the wrong amount of heat)? I noticed that the bulk of this arguement occurred when comparing an e6600, when the original arguement started in saying that can air match water? The e6600 doesn't dissipate that much heat so i don't think this is a fair platform for a comparison, however the water cooling heat delta should still be lower than air due to the whole spread resistance factor.

Now even with these crazy mods Bonzai plans to do (which are quite impressive i must say) I still don't think that could compare to a very top of the line watercooling system (ie a pump with good head pressure and flow rate with a thermochill PA120.3 and a good block Swiftech GTX or D-TEK Fuzion)

This link shows a direct comparison between two medium end watercooling and air cooling setups:

Air vs Water

Ok to wrap things up, water isn't for everyone, unless you have the time and money... for a high end setup you are looking at about 400$ minimum for the cpu alone. I don't mean to offend anyone, but I am just defending aigomorla in saying that with the best of the best air cannot match the best water cooling setup.

 

[BonzaiDuck]

I've had some additional thoughts about this today, and I think your observations are "right on."

The Swiftech kit in question -- a very good $240 kit -- shows the thermal resistance rating for the KIT (!!!):

H20 220-APEX ULTRA

People choose several water-cooling options. They opt for 1/2" tubing instead of 1/4" or 3/8"; they choose external reservoirs and radiators; they pick larger radiators and/or reservoirs -- all within the limitation that excludes TEC or evaporative enhancements. These things would tend to reduce further the "overall" thermal resistance rating -- of which the choice of materials for the waterblock and the thermal interface material are only one component.

Also, your remarks on "spread-resistance" seem reasonable.

On the matter of the E6600 processor, and quad-processors with stock TDP (I'm assuming your 110W quote is part of the intel spec for that processor), I'll need to go back and verify this again, but I recall somewhere a table of values or benchtest observations that an 65W (TDP) E6600, over-clocked by an additional 50%, would generate something close to 130W. I could be wrong, but perhaps we should look into that.

The quad processor -- if its TDP is 110W -- is about 7W "hotter" at stock settings than my old Prescott. But the comparison, I only think and speculate, would end there, in that there wouldn't be any correspondence in the over-clocked load-usage thermal wattage between the Kentsfield (apple) and the Prescott (orange.)

Prices are coming down this summer on both the C2D and the Q2D Kentsfield. I actually might consider coughing up the change for one, and as much as I don't like making extra trouble for myself, I might try it initially with the air-cooled rig, and if I'm dissatisfied, I might completely revise my cooling strategy.

Even so (hee, hee) remember these? . . . . :

No radiator-service pit-stops

. . . . a time when the connotations of "911" were positive . . .

 

[BonzaiDuck]

Footnote on the quad-core specs:

Checked Intel, and the spec sheet and updates for QX6800, QX6700, and the QX6600 (check those codes again, but these were all the spec sheets for variations under "Quad-Core.")

The two high-end processors have a TDP of 130W . . . !!!!!

The 6600 has a TDP of around 103W.

Imagine, then, what the over-clocked wattage could be.

 

[drakore]

Wow i didn't realize that an e6600 overclocked would generate so much heat. Sorry, about that 110W rating, it is correct for stock I believe, but I don't know what it is when overclocked by let's say 40%, because that was what I was referencing to in terms of air cooling being able to handle it at load.

Just one note on that H20-220 Apex Ultra kit, the Petra Tech's GPU elite kit with the MCR-320QP is a much better buy (in my opinion). You get the D-TEK Fuzion which is a non restrictive block and a pump with much better head pressure. It costs about 320$ US and the people are extremely helpful and friendly. I recommend the metal fittings for the rad, and you may want to add a resevoir. At the end of the day it is kind of pricey, but the performance will definitely best that of the H20-220 Ultra kit. At the end of the day the 80$ is worth it, and you will get a kit that will not fail your expectations of water cooling.

BonzaiDuck I would like to see pics (when it is ready) of your finalized TR-120, because it looks like some pretty cool modifications.

 

[BonzaiDuck]

drakore

I should beg pardon of members for my effusive verbal diarhea in recent days on this forum.

I'll try and remember to get some snaps of the heatsink part of the project, and post a couple. It may even occur before I can get the entire enchilada up and running -- migrating from the open-air "test-case" to the permanent one.

The last three weeks have been a setback for disappointments in painting the case with a poor paint-choice. I've had to strip off everything down to bare metal and start again with the etching primer. And I wouldn't have chosen to paint the interior, but for possibility of making this the real "bling-bling," perhaps for "contest-entry." So -- I went over the edge.

I spend a lot of time debating what tasks are sequential and which are independent, so I can keep things moving forward. I'm on pins-and-needles waiting for the package from Penn-Scientific, and I know as soon as it arrives, I'll start lapping that Silver Eagle.

Impatience can cause errors and mishaps, and I'm struggling a lot with it lately.

 

[imported_Husky55]

BonzaiDuck,

I feel your pain, man.

Keep up the faith. Technology can do that to you.

 

[aigomorla]

Originally posted by: drakore
Ok after reading all of this, it looks like Bonzai only refers to Swiftech kits. Now, to the original arguement between you and aigomorla saying that air can match water, this is not possible under loads... At the same ambient temperatures, the water cooling setup will win. Even with a massive heat sink the contact point is still fairly small, this being limited by the CPU slug size. As the heat dissapites, something called spread resistance occurs. Using water will limit this spread resistance because the heat doesn't have to travel as far through the waterblock to be swept away from the incoming cooled water.

Now, if you are going to compare water to air, you must take into account the item being cooled. Do you really think it is possible to match water with air on an overclocked quad core dissipating ~110W (Please correct me if that is the wrong amount of heat)? I noticed that the bulk of this arguement occurred when comparing an e6600, when the original arguement started in saying that can air match water? The e6600 doesn't dissipate that much heat so i don't think this is a fair platform for a comparison, however the water cooling heat delta should still be lower than air due to the whole spread resistance factor.

Now even with these crazy mods Bonzai plans to do (which are quite impressive i must say) I still don't think that could compare to a very top of the line watercooling system (ie a pump with good head pressure and flow rate with a thermochill PA120.3 and a good block Swiftech GTX or D-TEK Fuzion)

This link shows a direct comparison between two medium end watercooling and air cooling setups:

Air vs Water

Ok to wrap things up, water isn't for everyone, unless you have the time and money... for a high end setup you are looking at about 400$ minimum for the cpu alone. I don't mean to offend anyone, but I am just defending aigomorla in saying that with the best of the best air cannot match the best water cooling setup.

Already tried this arguement. So im just waiting for the crazy duck to do his own tests. If he's happy with them, great. I am not going to flame anyone for there own curiousity, unless its a stupid one, like lets throw my comp outside in the SNOW!.

So duckie, good luck on your tests, i am also waiting to see how you tweek that ultra120. My best friend who im giving my X3210 wont take it unless its aircooled.

And you guys all know how difficult it is to cool a quadcore @ 400+fsb.

But if anyone needs/requires water advice, you know you can always pm me.

Some TDP numbers for you guys:

Stock E6600 = 65W TDP. overclocked at 1.4Vcore = 121W
Stock Quad = 110W TDP Overclocked @ 1.45Vcore = ~215W
Stock 8800GTX = 210W TDP?? i know its 200+ and i have NO IDEA how much more a Vcore increase would cause.


So yeah, you guys see now why i recomend 120x3 radiators for quadcore + 8800GTX setups. :T


Mysetup:
Quad @ 1.47Vcore one primary loop with RD-30 + PA120.3 + apogeeGTX
7900GT x 2 <--- replacing them with 8900GTX's on 2 ek FC8800's + MCR220 + D5


Its extremely difficult to get nice temps with quadcore + sli 8800GTX's on the same loop. You can see from the math above.

 

[BonzaiDuck]

Yes, Aigomorla, I've been looking at your stickie on water-cooling, and it's fairly thorough.

Where the nVidia 8800 GTX comes in for its TDP value may be more of a "range" than an absolute value. Also, I noted earlier that nVidia shows a peak power-consumption value of some 171+ watts, but it would be worthwhile to go back and check to see if that's really the TDP value or the power-consumption value.

It's fairly certain that for power-consumption, the GTS consumes some 50+ watts less than the GTX, and that's a reason I've opted for the "second-best" card.

My case-mod (more like "synthetic case-build") project would easily accommodate large radiators and a big reservoir. In fact, while dreaming my way toward actually deciding what to do with it, I imagined a large copper reservoir with TEC coolers embedded in the copper jacket. At least the case -- and my modifications -- would accommodate some extreme ideas, even if they require additional minor mods to the case itself.

It looks as though the only really effective way to keep cool an overclocked quad would begin with water-cooling. But Intel may again attempt to mitigate the TDP issue in future models of the quad. It is difficult to keep up with the AMD/Intel duel, because the new developments seem to occur with milestones separated only by several months.

Last year's THG "Smithfield" build was similar to Aigomorla's figures for the quad, and I think OC'ing the low-end Smithfield to outrageous limits produced either a power-consumption value or a TDP well above the 200W range. Again, one would have to go back and check that article to see whether they were referring to electrical or thermal wattage.

 

[aigomorla]

Originally posted by: BonzaiDuck
Yes, Aigomorla, I've been looking at your stickie on water-cooling, and it's fairly thorough.

Where the nVidia 8800 GTX comes in for its TDP value may be more of a "range" than an absolute value. Also, I noted earlier that nVidia shows a peak power-consumption value of some 171+ watts, but it would be worthwhile to go back and check to see if that's really the TDP value or the power-consumption value.

It's fairly certain that for power-consumption, the GTS consumes some 50+ watts less than the GTX, and that's a reason I've opted for the "second-best" card.

My case-mod (more like "synthetic case-build") project would easily accommodate large radiators and a big reservoir. In fact, while dreaming my way toward actually deciding what to do with it, I imagined a large copper reservoir with TEC coolers embedded in the copper jacket. At least the case -- and my modifications -- would accommodate some extreme ideas, even if they require additional minor mods to the case itself.

It looks as though the only really effective way to keep cool an overclocked quad would begin with water-cooling. But Intel may again attempt to mitigate the TDP issue in future models of the quad. It is difficult to keep up with the AMD/Intel duel, because the new developments seem to occur with milestones separated only by several months.

Last year's THG "Smithfield" build was similar to Aigomorla's figures for the quad, and I think OC'ing the low-end Smithfield to outrageous limits produced either a power-consumption value or a TDP well above the 200W range. Again, one would have to go back and check that article to see whether they were referring to electrical or thermal wattage.

hmmmmm interesting.. yes i also too did hear about the 8800GTX's pulling around 171W.

But lets do some math.... A MCR220 can do about 250-260W. This is why i recomend a 320 for quadcores for the extra head room.

Now if you put 2 8800GTX's @ 171W = 340 watts of heat. A MCR320 should handle about 130W more then the 220. So thats quite a bit of heat. Hence why i said FAQ 8800's. Im waiting for the 8900's to come out.

Anyhow, about the tec idea, i already drafted it and was about to experiment with it. Then a legend named NOL, slaped me in the face and laughed at my concept. Dont mess with tec's unless u know absolutely what you are doing. The cooling requirements for something to pull that off is incredible.

You need to think of it in this aspect: Lets assume your running a E6600.

With tec on block, you need a 226W TEC to efficiently cool that sucker down. The tec will most likely draw about 150W downvolted + the 121W on your cpu.

Now you need to take care of the 271W of heat added just because you wanted those extra bit of low load temps. Also the hot side determins how cold your cold side is going to be. So you need a radiator with at least 2x overhead room.

So now your out looking for a 500W+ radiator, which there is only 2 i can think of, + your room will heat up expecially from this extra heat load, not to mention your electricity bill will make you pull your hair out.

You see why nol slapped me in the face when i said i want to TEC a quadcore?

 

[BonzaiDuck]

My idea was to get someone to weld together a copper reservoir with three or more TEC chillers welded into it (somehow) with the hot-plate facing outward, the cold-plate submerged in the coolant and facing the inside of the "jug."

I don't know what the power-consumption would be, assuming you'd go to "AllElectronics.com" and buy those surplus TECs-with-heatsinks that were originally manufactured for automotive food/beverage chillers/heaters.

It would have to be regulated, and the power-supply would probably need to be located outside the case -- defeating the purpose of trying to keep it "all together." Then there's the question of how you're going to dissipate the heat from the sinks on the hot side. One might be inclined to have those heatsinks extruding from the computer case, but what would one do about air-circulation? Otherwise, you'd have them in some sort of duct and pass air from an intake fan through that duct and -- zoom -- out the case rear.

I'm only guessing that the temperature-controlled regulation of something like this is also what you're suggesting as the main hurdle in all TEC setups.

I was just casually looking into TECs again, because a friend sent me a link to one or two of those hybrid TEC/heat-pipe coolers -- can't even remember the model names -- but the bill was something at least $120 or more.

My "take" on it was that these things appeared around the time the Conroe line became available, because only then would they be feasible given the TDP values and the areal size of the heatspreader cap. I had found some online "white-paper" some months earlier showing tests using contemporary processors that proved a simple TEC plate between processor and heatsink became unfeasible above 65 thermal watts.

So with the Conroe, my guess is that the company or companies designed the hybrid heatpipe-TEC-and-fan assembly for processors having those TDPs.

Even so -- and I don't believe the tests were run at anything but stock CPU settings -- there were some good reviews showing the device able to keep the idle-to-load spread within a few C degrees. But OC a Conroe or Q2D (which would already be above the feasibility threshold) -- and my guess is the contraption would be "in trouble."

 

[ChefJoe]

aigomorla:
According to the product literature, an MCR220-QP with 3/8" tubes and a reasonable flow rate handles 300-350 Watts of heat at 12 volts. Your figure is closer to the 7 volt statistic.

http://www.swiftnets.com/products/MCR220-QP.asp
SWIFTTECH

 

[aigomorla]

Originally posted by: ChefJoe
aigomorla:
According to the product literature, an MCR220-QP with 3/8" tubes and a reasonable flow rate handles 300-350 Watts of heat at 12 volts. Your figure is closer to the 7 volt statistic.

http://www.swiftnets.com/products/MCR220-QP.asp
SWIFTTECH

sorry, but about 70-80% of the people running watercooling tends to run there fans downvolted.

But your right, if you blast it at the full 12V at over 80-90+ CFM. Then yes that number will go higher.


As for me and most people, we watercool for silence as well as performance. :T


So i think my numbers are acutally near real world values.

 

[Spanki]

A sincere question for your WC folks... has anyone measured the temperature of the water exiting the radiator (entering the cpu block) ?

It seems to me that it will still be above 'room temperature' - how far above, is the question - and I'm sure it varies based on radiator size/efficiency, water flow rate and fan airflow, but isn't this one disadvantage to water cooling vs air ? There's lots of "fresh, room temerature air" available to air coolers, but I assume that water coolers "start out" cooler than they "end up" after running a bit (ramping up from room temperature to whatever the efficiency level of your rad setup is).

I'm not implying that this makes air coolers superior, just noting that you have to take that into consideration when you're speculating about how fast heat is being 'swept away' by the in-coming 'cool water'... obviously water is a much better conductor of heat than air is, but just how 'cool' is that water? (I seriously don't know)

 

[aigomorla]

Originally posted by: Spanki
A sincere question for your WC folks... has anyone measured the temperature of the water exiting the radiator (entering the cpu block) ?

It seems to me that it will still be above 'room temperature' - how far above, is the question - and I'm sure it varies based on radiator size/efficiency, water flow rate and fan airflow, but isn't this one disadvantage to water cooling vs air ? There's lots of "fresh, room temerature air" available to air coolers, but I assume that water coolers "start out" cooler than they "end up" after running a bit (ramping up from room temperature to whatever the efficiency level of your rad setup is).

I'm not implying that this makes air coolers superior, just noting that you have to take that into consideration when you're speculating about how fast heat is being 'swept away' by the in-coming 'cool water'... obviously water is a much better conductor of heat than air is, but just how 'cool' is that water? (I seriously don't know)

well i have temp probes in my reservoir. They come off after the cpu block.

Currently, at 100% load the ambients are around 74F

My coolant for cpu and gpu are: 27.5C and 28.1C respectively

im on dual loops also btw