Pro/Forums - View Single Post

Long Haired Git · 07-20-2005, 09:42 PM

1. Traditional PSU design has the hot component (Mosfets? Transistors?) mounted on a heatsink or pair of heatsinks. The heatsinks are typically, however, live. As in 120/240v AC power. Thus, the way it was done by BladeRunner and an Aussie I can't recall (but Google might) is to take out the heatsink, and replace it with a custom waterblock (bent copper pipe for the Aussie, flashy shiny thing for BladeRunner) and then electrically but not thermally isolate the component from the waterblock using mica shims. So far, haven't mustered the balls to do this myself as I've been zapped by 240v and I didn't like it.

2. Recent innovation has so called "silent" and "fanless" PSUs. Not sure how these work. Perhaps its just got better live heatsinks and lots of holes in the PSU casing, and components that can handle hotter temps?
Perhaps they thermally connect the live heatsink to the casing? In which case, you could then add a block to the casing, or replace that thermal interface with one to a WB.
Remember, the worse the interface between the heater (Mosfet) and the coolant, the hotter that component has to get to get its energy into the coolant. For a fanless PSU, that temp could probably get pretty hot before you care.

3. 100 WBs ain't going to matter much if 99 of them are basic straight-through-flow types. Eg: 10mm ID copper pipe soldered to the side of a copper HDD cage, or to the outside of a fanless PSU. Remember also that flow rates don't fall proportionatly as the resistance is so impacted by flow rates and vis-versa.
Eg: Laing D4, 2m x 1/2" tubing, 1 x 120.3, 1 x 120.2, 1 x LRR Whitewater block = approx 7.7 LPM
Add additional 20cm of tubing and a Swiftech 5002 (to represent a medium impendence GPU block) = 6.8 LPM
Add additional 20cm of tubing and a Swiftech 5002 (to represent a medium impendence NB block) = 6.1 LPM
Add additional 20cm of tubing and an Atlantis block (to represent a low impendence HDD block) = 5.9 LPM
Add additional 20cm of tubing and an Atlantis block (to represent a low impendence PSU block) = 5.7 LPM
Add additional 20cm of tubing and an Atlantis block (to represent a low impendence Mosfet block) = 5.5 LPM

Now the above are all approximated but you get the idea.

BTW, Eheim 1250 with the above gets 4.5LPM, MCP350 gets 3.5 LPM which is all "enough".

4. WRT the "hot" water at the end of the loop, I give you this:
http://www.geocities.com/lh_git/wckb...p_changes.html
So, 400W @ 3.5 LPM (MCP350) gives water 1.6 degrees hotter at the hottest point than the coldest.
400W @ 5.5 LPM (Laing D4) gives water 1.0 degrees hotter at the hottest point than the coldest.

To me, I'd have the CPU right after the big radiator, and I'd put the GPU next, the NB next, the HDDs next, the mosfets and then the PSU. Who cares if the water is 1 or 2 degrees hotter by the time it leaves the PSU?

5. Design it thinking of hot air rising, and getting the coldest air, and venting hot air away from the PC and you'll be fine. Remember also that once you remove the fans and silence the drives, you'll hear things like power mosfets "whistling". True silence is ambitious.

6. WRT coolers, I can't help you with those rads (being an Aussie), but I can point you to http://thermal-management-testing.com/ThermoChill.htm and predict that **typically** a radiator of the same size will perform approximately the same. You'll definately need more than 1 x 120.2 as shown in that site for 450W because you've not got good air flow, but then you can probably live with a higher difference in temp between coolant and air. You never get in trouble in buying the biggest radiator that will fit.

Whilst you are there, check out Graph 15. See how steep the graph is getting? Well, passive radiators will struggle to get 0.5 CFM even if horizontal, so you'll be in the LHS area of the graph.

Best to go large then!
Graph 13 is also scary, but remember that BillA was keeping the delta-T to 5 degrees. As per previous, my 260mm x 150mm of passive heater cores was able to keep approx 60W cool using no near zero air flow but with a much larger delta-T (probaly 15 to 20 degrees).

7. Define serious? Cathar designed the best blocks so far twice in a row (LRR, Cascade) and Silverprop make one of the best GPU blocks, and are Aussie like me. What are you buying from Germany? WC PSUs are rare and are typically DIY. Eheim pumps are also popular. However, you could buy all your components from say DangerDen or Swiftech, except you'd have to DIY the PSU and HDD blocks.

07-20-2005, 09:42 PM	#27
Long Haired Git Cooling Savant Join Date: Jan 2003 Location: Sydney, Oz Posts: 336	1. Traditional PSU design has the hot component (Mosfets? Transistors?) mounted on a heatsink or pair of heatsinks. The heatsinks are typically, however, live. As in 120/240v AC power. Thus, the way it was done by BladeRunner and an Aussie I can't recall (but Google might) is to take out the heatsink, and replace it with a custom waterblock (bent copper pipe for the Aussie, flashy shiny thing for BladeRunner) and then electrically but not thermally isolate the component from the waterblock using mica shims. So far, haven't mustered the balls to do this myself as I've been zapped by 240v and I didn't like it. 2. Recent innovation has so called "silent" and "fanless" PSUs. Not sure how these work. Perhaps its just got better live heatsinks and lots of holes in the PSU casing, and components that can handle hotter temps? Perhaps they thermally connect the live heatsink to the casing? In which case, you could then add a block to the casing, or replace that thermal interface with one to a WB. Remember, the worse the interface between the heater (Mosfet) and the coolant, the hotter that component has to get to get its energy into the coolant. For a fanless PSU, that temp could probably get pretty hot before you care. 3. 100 WBs ain't going to matter much if 99 of them are basic straight-through-flow types. Eg: 10mm ID copper pipe soldered to the side of a copper HDD cage, or to the outside of a fanless PSU. Remember also that flow rates don't fall proportionatly as the resistance is so impacted by flow rates and vis-versa. Eg: Laing D4, 2m x 1/2" tubing, 1 x 120.3, 1 x 120.2, 1 x LRR Whitewater block = approx 7.7 LPM Add additional 20cm of tubing and a Swiftech 5002 (to represent a medium impendence GPU block) = 6.8 LPM Add additional 20cm of tubing and a Swiftech 5002 (to represent a medium impendence NB block) = 6.1 LPM Add additional 20cm of tubing and an Atlantis block (to represent a low impendence HDD block) = 5.9 LPM Add additional 20cm of tubing and an Atlantis block (to represent a low impendence PSU block) = 5.7 LPM Add additional 20cm of tubing and an Atlantis block (to represent a low impendence Mosfet block) = 5.5 LPM Now the above are all approximated but you get the idea. BTW, Eheim 1250 with the above gets 4.5LPM, MCP350 gets 3.5 LPM which is all "enough". 4. WRT the "hot" water at the end of the loop, I give you this: http://www.geocities.com/lh_git/wckb...p_changes.html So, 400W @ 3.5 LPM (MCP350) gives water 1.6 degrees hotter at the hottest point than the coldest. 400W @ 5.5 LPM (Laing D4) gives water 1.0 degrees hotter at the hottest point than the coldest. To me, I'd have the CPU right after the big radiator, and I'd put the GPU next, the NB next, the HDDs next, the mosfets and then the PSU. Who cares if the water is 1 or 2 degrees hotter by the time it leaves the PSU? 5. Design it thinking of hot air rising, and getting the coldest air, and venting hot air away from the PC and you'll be fine. Remember also that once you remove the fans and silence the drives, you'll hear things like power mosfets "whistling". True silence is ambitious. 6. WRT coolers, I can't help you with those rads (being an Aussie), but I can point you to http://thermal-management-testing.com/ThermoChill.htm and predict that typically a radiator of the same size will perform approximately the same. You'll definately need more than 1 x 120.2 as shown in that site for 450W because you've not got good air flow, but then you can probably live with a higher difference in temp between coolant and air. You never get in trouble in buying the biggest radiator that will fit. Whilst you are there, check out Graph 15. See how steep the graph is getting? Well, passive radiators will struggle to get 0.5 CFM even if horizontal, so you'll be in the LHS area of the graph. Best to go large then! Graph 13 is also scary, but remember that BillA was keeping the delta-T to 5 degrees. As per previous, my 260mm x 150mm of passive heater cores was able to keep approx 60W cool using no near zero air flow but with a much larger delta-T (probaly 15 to 20 degrees). 7. Define serious? Cathar designed the best blocks so far twice in a row (LRR, Cascade) and Silverprop make one of the best GPU blocks, and are Aussie like me. What are you buying from Germany? WC PSUs are rare and are typically DIY. Eheim pumps are also popular. However, you could buy all your components from say DangerDen or Swiftech, except you'd have to DIY the PSU and HDD blocks. __________________ Long Haired Git "Securing an environment of Windows platforms from abuse - external or internal - is akin to trying to install sprinklers in a fireworks factory where smoking on the job is permitted." (Prof. Gene Spafford) My Rig, in all its glory, can be seen best here AMD XP1600 @ 1530 Mhz \| Soyo Dragon + \| 256 Mb PC2700 DDRAM \| 2 x 40 Gb 7200rpm in Raid-0 \| Maze 2, eheim 1250, dual heater cores! \| Full specifications (PCDB)