Overclock.net › Forums › Case Mods & Cases › Builds & Case Mods › Build Logs › [Build Log] Pandamonium (TH10 w/Ped, RIVE w/3930K, Tri-SLI EVGA 670 4GB SC, Water Cooled)
New Posts  All Forums:Forum Nav:

[Build Log] Pandamonium (TH10 w/Ped, RIVE w/3930K, Tri-SLI EVGA 670 4GB SC, Water Cooled) - Page 34

post #331 of 431
Thread Starter 
One other item, forgot to add that my fill line is a SHARED fill line between the 2 reservoirs and I do see liquid moving over the fill line between reservoirs; I assumed it was due to different levels of restrictions in the loop.

Quote:
Originally Posted by Electrocutor View Post

Actually, I figured it out, and that's not your loop. I'll draw it up in notepad so you can see what is happening. Your loop is never being pushed by more than two pumps. You have the potential to better control the flow if you split up the pump speeds into 2x2. One set powers the blocks, one set powers the radiators, but you'd have to make absolutely sure that the set power the rads always has a higher flow rate than the one powering the blocks, otherwise you will run the same water through the blocks without ever having gone through the radiators.

Thanks man, I REALLY appreciate the help.
post #332 of 431
Thread Starter 
Quote:
Originally Posted by Methos07 View Post

How on earth do you bleed a system like yours? I would be in tears.

I have a drain line in the ped. I actually roll it up to a shower and just let it drain. The only issue really is the 120.3, since the fittings are on the top.
post #333 of 431
Thread Starter 
Quote:
Originally Posted by cpachris View Post

Love the graphs. I hope to do something like that also when I'm done. Surprised at the large difference in temps for 30% vs 100% pump speed. How bad is the sound of the pumps at 100%. Are you going to settle somewhere in the middle?

Well I think Electrocutor is about to show me I've been a dumbass and I need to redo my loop. Previous to today's conversation in the thread I was planning to leave the pumps at 100% and fans at 40%.
post #334 of 431
Hmm, interesting. When I bled my 800d with a 360 / 140 and 4 blocks I was flipping it upside down and all sorts of crazy things to get the air out.

Can't imagine you doing that with this, lol. How does a drain line work?
post #335 of 431
Thread Starter 
Quote:
Originally Posted by Methos07 View Post

Hmm, interesting. When I bled my 800d with a 360 / 140 and 4 blocks I was flipping it upside down and all sorts of crazy things to get the air out.
Can't imagine you doing that with this, lol. How does a drain line work?

Basically I put a fitting with tubing and a valve on one of the bottom ped rads (they have 6 fitting openings, so I use 2 for in/out and 1 for drain). When I need to drain I open the fill port, open the drain port, and let gravity do it's things. When most of the water is out I tilt the case a bit to get "stuck" water, but there isn't much since the drain is at the bottom most part of the loop.
post #336 of 431
Quote:
Originally Posted by Reglar View Post

Well I think Electrocutor is about to show me I've been a dumbass and I need to redo my loop. Previous to today's conversation in the thread I was planning to leave the pumps at 100% and fans at 40%.

Lol. You have some pretty nice temps and delta if you have it setup inefficiently. Can it get even better?
post #337 of 431
Quote:
Originally Posted by Reglar View Post

Here's the final test results, I will put them in some graphs Friday. I can say that the results show that pump speed had a bigger impact on CPU temps, fan speed had bigger impact on Delta T. Since CPU temp is what I am after I plan to run at 40% fan speed and 100% pump speed, the noise is very manageable. Warning: Spoiler! (Click to show)
For some reason a bunch of my inline temp sensors stopped working today. I moved them to posts with working sensors, and no result, and when I moved working sensors to the posts they were using the posts reported just fine. Very weird I would lose 5 at once, I'll play with it later though.
I had enough left to do the testing so below are the results. Same test conditions as previous testing.
40% fans, 30% pumps, Delta T of 3.7

Same test with 100% pumps, no appreciable change on Delta T but CPU temps came down a few degrees.

50% fans, 30% pumps, Delta T of roughly the same, the screenshot just was at a moment when it blipped up but you can see the average in the box.

Same test with 100% pumps, again we see the CPU temps come down a bit, so did Delta T.

60% fans, 30% pumps, Delta T similar to 50% fans, 100% pumps, but the CPU temps were higher with the higher fan speed and lower pump speed.

60% fans, 100% pumps, overall similar performance as low pump levels.

70% fans, 30% pumps, Delta T is lower but CPU temp is pretty much unchanged.

The trend continues, pump speed is definitely lowering CPU temps more than the fan speed.

The rest of the results follow a similar trend, included below for completion.






Just saw these. I hadn't clicked through on the email notification initially. They are awesome. Aquacomputer should use something like this in their advertising. Amazing how you can monitor, record and save all of this data with a screenshot. I definitely plan on doing something similar. I usually think in terms of gallons per minute, instead of per hour. Just noticing that the flow rate is actually pretty low at 30% pump speed. Even at 100% pump speed, I would have thought it would get more than 1 gpm. Maybe Electrocutor is on to something.
post #338 of 431
Code:
(Res)+-(Rad)<-(Rad)<-(Rad)<-(Rad)<-(Rad)<-(Flo)<-(Pmp)
  |  |                                             ^
  v  '--.                                          |
(Pmp)   '------------------------------------.   (Pmp)
  |                                          '--.  ^
  v                  .->(GPU)-.                 |  |
(Pmp)->(Flo)->(CPU)->+->(GPU)-+-->(Chp)->(VRM)->+(Res)
                     '->(GPU)-'

I have a number of ideas for changing your loop, but at the very least you should add a valve to your fill line. When the CPU and GPUs are at load, you want to have as much water going through them as you can, so close the valve most of the way to add some of the pressure from the other pump to the blocks.

The one thing I haven't quite figured out is why at 100% the radiator pump is only pushing right at 1gpm. I wouldn't have figured that just some elbows and the Aquaero block would be that restrictive. It might have to do with your choice of using all low-profile elbows. The curved (longer) elbows have a lot less restriction than the stark right-angle ones. It might also have to do with having two right-angle elbows between the res and the rad pump inlet: they may be starving the pump and that's why you see double the flow rate in the rads than blocks at lower speeds and almost no difference at all at higher speeds.

You haven't posted GPU temps yet, but I bet they won't be especially great. With max block flow at 0.9gpm, each GPU will only have 0.3gpm flow rate. With the pumps you have here, I see no reason why you wouldn't be able to get 1.5gpm or higher for all blocks and radiators if you worked on the loop efficiency a bit. If you're interested and would like me to help, just ask; otherwise, your CPU temps aren't so bad if you'd rather stop messing with it and wasting coolant, but I would at least add a valve to your fill line.
post #339 of 431
Thread Starter 
Quote:
Originally Posted by Electrocutor View Post

Code:
(Res)+-(Rad)<-(Rad)<-(Rad)<-(Rad)<-(Rad)<-(Flo)<-(Pmp)
  |  |                                             ^
  v  '--.                                          |
(Pmp)   '------------------------------------.   (Pmp)
  |                                          '--.  ^
  v                  .->(GPU)-.                 |  |
(Pmp)->(Flo)->(CPU)->+->(GPU)-+-->(Chp)->(VRM)->+(Res)
                     '->(GPU)-'
I have a number of ideas for changing your loop, but at the very least you should add a valve to your fill line. When the CPU and GPUs are at load, you want to have as much water going through them as you can, so close the valve most of the way to add some of the pressure from the other pump to the blocks.
The one thing I haven't quite figured out is why at 100% the radiator pump is only pushing right at 1gpm. I wouldn't have figured that just some elbows and the Aquaero block would be that restrictive. It might have to do with your choice of using all low-profile elbows. The curved (longer) elbows have a lot less restriction than the stark right-angle ones. It might also have to do with having two right-angle elbows between the res and the rad pump inlet: they may be starving the pump and that's why you see double the flow rate in the rads than blocks at lower speeds and almost no difference at all at higher speeds.
You haven't posted GPU temps yet, but I bet they won't be especially great. With max block flow at 0.9gpm, each GPU will only have 0.3gpm flow rate. With the pumps you have here, I see no reason why you wouldn't be able to get 1.5gpm or higher for all blocks and radiators if you worked on the loop efficiency a bit. If you're interested and would like me to help, just ask; otherwise, your CPU temps aren't so bad if you'd rather stop messing with it and wasting coolant, but I would at least add a valve to your fill line.

Thanks for the help, I am not quite following the fill line valve comment. Since the valve lines are with the reservoirs, I don't understand how a valve will help here. You think the fill line is causing a pressure drop on the water going into the pump blocks?

One concern I have is that perhaps my PWM cable is causing issues, it was a custom cable I made. It seems to be working but since I am just getting a sensor from 1 pump maybe it's misleading. I will try tonight removing the PWM so that there is no possibility of bad PWM signaling.



GPU temps were ~37C.
post #340 of 431
GPUs were 37C at full load? That's actually really impressive.


With pumps at the same speed, valve open will allow block and rad loops to act independently. Which, at least at lower pump speeds means the rads will have twice the flow as the blocks. This should bring down DeltaT, but increase block temps due to lower flow rate. With the valve closed, both flow rates are in-line and any excess pressure the rad pump has is added to the block pump, thus increasing block flow. Since your DeltaT is so low anyway (from tons of rads), I would bet you'll get the best results just leaving it all the way closed.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Build Logs
Overclock.net › Forums › Case Mods & Cases › Builds & Case Mods › Build Logs › [Build Log] Pandamonium (TH10 w/Ped, RIVE w/3930K, Tri-SLI EVGA 670 4GB SC, Water Cooled)