|
|
General Liquid/Water Cooling Discussion For discussion about Full Cooling System kits, or general cooling topics. Keep specific cooling items like pumps, radiators, etc... in their specific forums. |
Thread Tools |
10-30-2002, 05:44 AM | #151 | |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Quote:
6mm is WAY too thick. Don't stop with whole millimetres either. The devil truly is in the fine details here. Last edited by Cathar; 10-30-2002 at 05:55 AM. |
|
10-30-2002, 07:10 AM | #152 | |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Quote:
As a rough guide I still rely on the Christmas tree. I find it difficult to envisage Heat Transfer Coeffs (to the bp) from 1mm spaced fins exceeding ~ 100kW/m*m**k unless velocties exceed ~10m/s. Eg . For my own purposes I will probably use any thinner piece of copper that falls in my hands. I am only interested in "look-see" and have no developement objective. |
|
10-30-2002, 01:01 PM | #153 | |
Cooling Savant
Join Date: Oct 2002
Location: Boston
Posts: 238
|
Quote:
Also, referring to an earlier post, pressure drop is proportional to velocity squared only when comparing points on either side of a pump, for example. It doesn't tell you anything about form friction - that is, friction or pressure drop caused by fittings. Pressure drop caused by fittings in turbulent flow can best be described through an empirical correlation - there's no direct mathematical relationship between pressure drop and velocity. In laminar flow, pressure drop due to fittings is proportional to velocity. People keep tossing the word "thermodynamics" out there. This is entirely chemical engineering transport - heat transfer and fluid flow. I haven't seen any thermodynamics issues relating to this stuff. Alchemy |
|
10-31-2002, 01:39 PM | #154 |
Cooling Savant
Join Date: Oct 2000
Location: Rockledge,FL,US
Posts: 731
|
__________________
My old and retired watercooling setup. Watercooled K6-2 450 at 600 Also Retired - Watercooling an XP1800@1782MHz |
11-03-2002, 07:23 PM | #155 | |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Quote:
|
|
11-03-2002, 07:50 PM | #156 | |
Cooling Savant
Join Date: Oct 2000
Location: Rockledge,FL,US
Posts: 731
|
Quote:
__________________
My old and retired watercooling setup. Watercooled K6-2 450 at 600 Also Retired - Watercooling an XP1800@1782MHz |
|
11-03-2002, 08:01 PM | #157 | |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Quote:
The micro-channel concept is truly ancient by modern computing standards. This stuff has been around for 20+ years. Where my block differs to many of the older designs is that it combines jet impingement with micro-channel and a specific range of base-plate thicknesses, while using the micro-channel walls as an essential part of the cooling process from the jet impingement flows, and is focussed on cooling < 15x15mm sized CPU dies (P4 is 11.3 x 11.3, Athlon XP is 11.6 x 11.1). |
|
11-03-2002, 08:22 PM | #158 | |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Quote:
There is a predicted difference of possible more than 0.06C/W between the "Side-to-side" and the "Central Inlet" with h=100kw/m*m*c in the JIA and a 6mm bp For a 70W Duron(50W Real) a 0.06C/W should be realised as ~ 3c which may or may not be detectable. Edit: Added curve for the perhaps more realistic h=50kw/m*m*c in the Jet Impingement Area. * Sums:- Total Cooling = (Cooling from Jet) + (Cooling from Wall Jet) = Qj + Qw = C(Wj/Cj) + C(Ww/Cw) Total C/W = 1/[(Wj/Cj) + (Ww/Cw)] For a 15x15mm Jet Impingement Area(JIA), 6mm bp,and 200LPH through 3mm nozzle(h=100kw/m*m*c) (Wj/Cj) is reciprical of "15x15mm Impinge Area" C/W value at 100kW/m*m**c on graph. (Ww/Cw) taken as the difference betwwen recipricals of (50x50mm) and (15x15mm) C/W s at 100LPH(split flow). (Wj/Cj) = 1/0.135 = 7.407 (Ww/Cw) = 1/0.199 - 1/1.06 = 5.025 - 0.943 = 4.082 Total C/W = 1/[(Wj/Cj) + (Ww/Cw) = 1/[7.407 + 4.082] = 1/11.489 = 0.087 For 12x12mm JIA,6mm bp,and 200LPH through 3mm nozzle(h=100kw/m*m*c):- (Wj/Cj) = 1/0.182 = 5.495 (Ww/Cw) = 1/0.199 - 1/1.62 = 5.025 - 0.617 = 4.408 Total C/W = 1/[(Wj/Cj) + (Ww/Cw)] = 1/[5.495 + 4.408] = 1/9.903 = 0.101 Repeat for various bp thicknesses with a JAI Heat Trans Coeff (h)of 100kw/m*m*c. Last edited by Les; 11-04-2002 at 09:18 AM. |
|
11-06-2002, 09:36 AM | #159 |
Cooling Neophyte
Join Date: Sep 2002
Location: Sweden
Posts: 66
|
If you for some reason had to do this block in aluminium, how much would that lower the performance? I mean the original design now, not nozzled.
Could you see it on one of those nice graphs, and would you make the base even thinner? Thanks. |
11-06-2002, 02:59 PM | #160 | |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Quote:
The production block that's currently selling differs quite substantially (relatively speaking) in dimensions from the prototype that's shown here. The base is even thinner and the channel heights are changed. Les's graphs are an excellent resource. I'm not quite sure what's going on though. I stepped through a range of base thicknesses with real blocks in 0.1mm steps, and while I don't have the most accurate test equipment in the world I was able to see significantly more optimistic results for thin bases than what Les is predicting. I don't know. Maybe `h' is higher in reality by around 20% or so. I imagine it gets pretty hard to model "furniture" in an impingement scenario. |
|
11-06-2002, 03:29 PM | #161 |
Cooling Neophyte
Join Date: Sep 2002
Location: Sweden
Posts: 66
|
Thanks, but would you mind describing the changes? The base is not 1,25mm thick, and the fins aren't 5 mm high?
|
11-06-2002, 03:36 PM | #162 | |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Quote:
The info will come out sooner or later. I'd prefer later. The final production block outperforms the prototype across the board. I spent quite a bit of money getting it that way so don't really want to give out final details before I get a chance to make my money back. |
|
11-24-2002, 02:18 AM | #163 |
Cooling Neophyte
Join Date: May 2002
Location: VA
Posts: 16
|
I've been reading this thread but not contributing cause alot of this talk is way outta my league. But at this point i do just want to say that this is definatly an awesome job done by you cathar (and everyone that has helped you) and that you should fully make your money back for all the time and effort that you have put into a project like this. Thank you for pushing the limits for us in the water cooling world.
slayher P.S. i would be interested in your blocks but right now i'm pelt cooling, which i dont think would work right with your block |
11-29-2002, 11:07 AM | #164 |
Cooling Savant
Join Date: Oct 2002
Location: MidWest USA
Posts: 176
|
Interesting thread and very nice block. One thing about the design has me curious. I'm sure the top o-ring prevents leaks, but does it prevent "cross-talk"? Have you considered isolating the 3 holes? I was wondering about o-rings like such:
I realize it would be a lot of work changing it, but I'm curious if the pressure added by the nozzle/slit is causing some of the coolant to channel between the 2 top plates and leaving the system without reaching the botom plate. I'm not criticizing your design, I just thought since you are obviously tweaking it for optimization, this might be worth looking at. peace. unloaded PS: have you ever setup one of these blocks with 3/8" inlet and 1/4" outlets? I'd be curious how it would do like that. Your results with low flow seem to indicate it would still do pretty good with this configuration and a lot of us are still using 3/8" setups. |
11-29-2002, 03:21 PM | #165 | |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Quote:
This question has been asked before and I'll answer it with an picture: As you can see, there is no gap between the middle and top plates, or if there is one, it is minute in comparison to the size of the micro-channels which can be seen in the lower part of the picture. If there is "cross-talk" it would be miniscule in comparison to the flow through the block proper. I do supply fittings for 3/8" systems and have already sold a few blocks in this configuration. I haven't specifically tested a 3/8" setup myself but have heard no complaints so I assume they're doing the job. I could do a 1/4" config too if someone desired that, it's no real issue since I fit the barbs on just before the block is packaged and sent. |
|
11-29-2002, 03:25 PM | #166 |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Oh, while I'm here, I thought I'd add some pictures of the blocks as they appear before shipping to people. The bases are all lapped to a 600-grit finish, and the rest of the block's externals are sanded back, polished with Brasso, and followed up by a coat of polish to make the blocks resistant to tarnishing (except the base of course - which is never polished)
|
11-29-2002, 08:00 PM | #167 |
Cooling Savant
Join Date: Oct 2002
Location: MidWest USA
Posts: 176
|
Ah, a pic is worth 1K words. I figured you had already adressed the question, but I've only read this thread and the one on [H] and saw no mention of it. I didn't realize the oring would mash flat enough to allow the surfaces to mate like they do. Another thing I didn't realize is the size of all three barbs are the same. For some reason I thought it was 1/2" in and the outs were 3/8", I guess I was just assuming. The more I see this block, the more impressed I am with it. With the holidays coming the price is a bit much for me atm. I'm not saying its overpriced, as I feel its the best value to be had if buying a block, its just over my budget. This will prolly change after the first of the year. How are sales going btw? First batch gone? Any signs of the price dropping soon for US buyers? I'm hoping for a US manufacturer or distributor so at least the shipping cost comes down a bit. I really wanna get one and pair it with this heater core:
I don't expect it would improve performance, but it would do away with the Y and look better and I'm a modder at heart, although you guys are swaying me to the dark side(functionality over form) 8P~ peace. unloaded |
11-29-2002, 08:29 PM | #168 |
Registered User
Join Date: Feb 2002
Location: Tucson, Arizona
Posts: 204
|
The only way you can run the three barb H/C is having supplying water from the H/C going to the pump and from pump to water block. This may or may not be the best setup. You always want the coldest water entering the block first and you can not do this with a three barb H/C. Maybe somebody has some insight as how much hotter would the water be after going through the pump first. I do agree it would look better than using a Y fitting.
Photo of the heater core already made. Last edited by DodgeViper; 11-29-2002 at 10:32 PM. |
11-29-2002, 09:01 PM | #169 |
Cooling Savant
Join Date: Oct 2002
Location: MidWest USA
Posts: 176
|
DV, I thought you'd want the coldest water from the rad going straight to the WB too. But it has been pointed out to me that this isn't the best. The argument is that the bit of performance hit by the heat from pump, is overshadowed by having the faster water directly from the pump to the cpu without the resistance of the rad in between. Not 100% this is true, but I took it as such. Maybe somebody will enlighten us further.
PS: If anybody is interested in that heatercore, its listed as one for a Ford Fairlane w A/C 1970-67 peace. unloaded |
11-29-2002, 09:24 PM | #170 | |
Registered User
Join Date: Feb 2002
Location: Tucson, Arizona
Posts: 204
|
Quote:
|
|
11-29-2002, 09:59 PM | #171 | |
Cooling Savant
Join Date: Jul 2002
Location: Toronto, Canada
Posts: 381
|
Quote:
__________________
Michael E. Robbins M.A.Sc. Candidate, University of Toronto 12.1 GHz of AMD's finest (17.7 GHz total) crunching proudly for the AMDMB.com Killer Frogs SETI BOINC: Dual Opteron 246s (Iwill DK8N) | XP2800+ (Shuttle SN41G2) | 3x XP2400+ (ASUS A7N266-vm) SETI BOINC: 2x P4 2.8E (ASUS P4R800-vm) | Crunching 24/7 |
|
11-29-2002, 10:05 PM | #172 |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
For the most part, the temperature of the water does not range over more than about 0.25C total difference between any two points in the system.
The pump before or after the waterblock only incurs the additional heat added by the pump itself, which for something like an Eheim 1250 means less than 0.02C added to the temperature of the water as it flows through it, and less for pumps like the Eheim 1048. If you are noticing any variations with either setup, then this will be more liekly due to mounting variations and therm probe inaccuracy. Following the physics through, it is extremely unlikely that you will ever be able to detect a 0.02C difference with a 0.5C pseudo-accurate uncalibrated CPU diode. |
11-29-2002, 10:11 PM | #173 |
Cooling Savant
Join Date: Jul 2002
Location: Toronto, Canada
Posts: 381
|
Actually, I was refering to the "faster water directly from the pump to the cpu without the resistance of the rad in between" part. I didn't notice any difference when I changed the order myself.
__________________
Michael E. Robbins M.A.Sc. Candidate, University of Toronto 12.1 GHz of AMD's finest (17.7 GHz total) crunching proudly for the AMDMB.com Killer Frogs SETI BOINC: Dual Opteron 246s (Iwill DK8N) | XP2800+ (Shuttle SN41G2) | 3x XP2400+ (ASUS A7N266-vm) SETI BOINC: 2x P4 2.8E (ASUS P4R800-vm) | Crunching 24/7 |
11-29-2002, 10:28 PM | #174 | |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Quote:
|
|
11-29-2002, 11:16 PM | #175 |
Been /.'d... have you?
Join Date: Jul 2002
Location: Moscow, ID
Posts: 1,986
|
Actually, this is going to sound totally f*cked, but it does make a slight difference in what order the parts are in. I've actually done tests on my system at home (before the latest project hit) in which I swapped positions of the rad and WB, and found that I was actually getting better temps and slightly better flow figures if I had the WB first. Further, since I never removed and reseated the WB, mounting errors cannot be added to the equation. When I tried putting the rad on first after the pump again, the temps normalized to their previous, higher temps.
Logically and theoretically, there is no reason to say that there should be any difference except a beneficial one from having the coldest water hitting the block (pump>rad>wb), but in real life situations it appears that the opposite is true. Feel free to tell me where I'm wrong, beccause I can't find a reason for this to be so, but it honestly does appear that under "normal" circumstances (not using a super-hot pump), that having the WB immediately after the pump is a good thing.
__________________
#!/bin/sh {who;} {last;} {pause;} {grep;} {touch;} {unzip;} mount /dev/girl -t {wet;} {fsck;} {fsck;} {fsck;} {fsck;} echo yes yes yes {yes;} umount {/dev/girl;zip;} rm -rf {wet.spot;} {sleep;} finger: permission denied |
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
Thread Tools | |
|
|