[Build Log] From Chiller to MO-RA

[Arcmanov]

Ah boy...the good ol HAF-X gettin it done.

That said though...not me and that GPU liquid cooling nah.  I don't overclock my GPUs except for the odd benchmark or two.
Still, the thermal headroom is nice to have.

...but then I used to say that about CPU liquid cooling too. 





Samsung Galaxy Note 2 | Omega v16 | SwiftKey | Tapatalk 4 Beta

[captainchris]

Ah boy...the good ol HAF-X gettin it done.

That said though...not me and that GPU liquid cooling nah.  I don't overclock my GPUs except for the odd benchmark or two.
Still, the thermal headroom is nice to have.

...but then I used to say that about CPU liquid cooling too. 





Samsung Galaxy Note 2 | Omega v16 | SwiftKey | Tapatalk 4 Beta

when you see yours card maxing at 35c during hrs of sweaty caspian border, or a wicked running of gulf of oman...

ni**uhs be like

[MessiaaH]

Yup, and them kepler gpus like to get up when they ready, when u can load at 40 degrees overcooked, its nice. And oc on these cards actually make a substantial difference n performance. It meant the different between 50-55 fps fluctuations, and a solid 60 in TR with everything on.

D'oh worry arc, soon enough u going to join d gpu lcs party ....

[MessiaaH]

Parts Reach!

[Netizen1]

18 fans!!! WDA?!?!?!?

[Spazosaurus]

The stuff gamers' wet dreams are made of. I think this thing is better cooled than my van!!

Sas controllers why though. You using enterprise SSD's? Why not a PCI-E SSD?

[MessiaaH]

18 fans!!! WDA?!?!?!?

lol, I bought 2 external power bricks to power them, but I did some math and I think my PSU can comfortably power all 18. So I going to rig it up to my front bay switches. So I can turn on/off a side of fans at a time. And then too they will come on/off with the machine.

The stuff gamers' wet dreams are made of. I think this thing is better cooled than my van!!

Sas controllers why though. You using enterprise SSD's? Why not a PCI-E SSD?

It's a SAS/SATA controller. The 2 cables in the pic right below the controller is min-sas to sata extender cables.
So 1 mini-sas port gives 4 sata ports. And I didn't really buy the controller for my SSDs. I had a raid5 on my motherboard, and I was tweaking my overclocking, and 2 drives out of my 4 drive array decide to say they aren't members anymore. It took a few hours and extreme patience to rebuild that raid and retain my data. Not me and intel shit raid again! So I immediately put in an order for an areca, my data too important. So that card going to really be used for 6 * 3TB drives. I'm toying with the idea of putting my 2 SSD's on it as well. But I not sure how UEFI going to play with it. I'll test and see.

[captainchris]

Mother of ..

Sent from my Nexus S using Tapatalk 2

[Spazosaurus]

Very nice @ that raid card. I assume that's the ARC-1223-8i model? That thing expensive boy. Does it have a bios that is accessible outside of the OS? Something like this is really chalk and cheese compared to a raid controller built into a board. There's just no comparison.

[MessiaaH]

Yes, that's the model self. I used to have a 1220 in meh beast ah while back (MH-LCSx11 ah think). But I end up sending the card trini to use in we storage server home. But this onboard raid quality just suck boy, I have to go back to an areca. Onboard RAID 1 sucks hard, anytime power go, or machine unexpectedly restart, stupid raid start checking for parity errors, which can take anywhere from 1-2days, during which time performance degraded. I switch to Windows 8 Storage Spaces, but performance on that was crap for VMs. Then i switch to onboard RAID5 and that mess happen. That was the last straw. Back to areca.

Unfortunately i loose a PCI-E slot, so will have to stick to Triple SLI or Dual Slot Quad. But given the power of these new GPUs, that should be just fine.

And yes, there is a bios accessible from outside the OS. It's just like a server raid card, ctrl+i or whatever to enter raid bios on boot. Areca actually make a lot of the raid controllers in servers, HP/Dell/IBM does just rebrand they tech. If u search amazon u go see areca controllers going upward of $1000 deneros.

Check this bad boy:
http://www.amazon.com/Dual-Height-Support-Internal-External/dp/B006HUNAV8/ref=sr_1_8?ie=UTF8&qid=1374159888&sr=8-8&keywords=areca+raid

Something like this is really chalk and cheese compared to a raid controller built into a board. There's just no comparison.

Yup!

[MessiaaH]

I found that my top gpu temps was a bit high, so whiles machine is down decided to investigate to see why
and also install a pair of quick disconnects for the gpus to make thing easier in the future.


Machine Tear-Down:


















Ah Ha! look the bastard there!





After waterblock clean:




Quick Disconnects:

[MessiaaH]

Day 5

Pump Heatsink and fan install:

Also had to build a custom switch bay to control the MORA Fans and ligthing, with externally accessible connectors for the rad fans
.
Rigged the switchbay to be powered by the psu.

Had to also do a little mod to the front bay beneat the resevoir, as it no longer fits properly with the heatsink/fan installed.

A couple custom Cables as well.

[Philosophical45]

Wayz, serious like kamla and a bottle ah puncheon! Look at d size of that flipping thing!

[TriniXaeno]

hahhaha

it just ridiculous

Man building skynet

[firstchoicett]

things looking real nice.

[MessiaaH]

Day 5

GPU Overheating Woes   :

I’ve mentioned before that one of my GPUs was running hotter than normal.
I did find some build-up of gunk in the block and cleaned it out so I thought that was the issue.
So I proceeded to bleed test the GPUs and add it back into my machine:







To my surprise the primary gpu was still running hotter than the rest. 20 degrees difference to be exact.
On idle, the primary gpu would be like 55, and the other 2 @ 25degrees each. On load the Primary gpu would quickly rise to 90~100 degrees!

So obviously this was unacceptable, so I proceeded to do more testing. I used the features of the motherboard to disabled each PCI-E Lane 1 by 1, and individually tested the cards, same issues. The gpu on the top just runs hot! I thought it may be the motherboard, so I dropped the entire triple setup down 1 notch. So GPU 1 was now in slot 2, did not make a difference.
When I finally exhausted all other options I figured it out.

When using the chiller, the FlowRate and Pressure in the loop was very high, because the chiller offers no resistance at all, it’s basically a giant reservoir. The addition of the MORA on the other hand, offered significant flow/pressure reduction, which I saw initially. But the significant drop affected the way I have my liquid flowing in the loop.

My current setup is:

RAD -> CPU -> Parallel split -> (GPU / NorthBridge / RAM) -> Reservoir

GPUs -> Parallel Split:

                 


   
I suspected that the flowrate/pressure drop was affecting the GPU parallel setup. Most of the flow/pressure was going to the 2nd and 3rd GPUs, and the 1st hardly got any water at all. So I removed the EK LCS SLI Bridge, and rigged up a Series loop for testing:




Turns out that’s exactly what was going on, idle temps on all 3 gpus are now 25 a piece, and load 40 degrees max. CPU idles around 27, load’s at 50. Which in of itself is pretty impressive for the MORA.
It’s actually competiting on an equal footing with the Chiller. I know the chiller is capable of extremely low temps, but running your machine at that constantly would eat a lot of current and wont be practical.

Thus far I’m extremely impressed with the performance of this radiator. 5.0GHz OC with 3 OCed 680s, NB, RAM and those temps? Kinda epic.

I’ve ordered a new EK LCS SLI Bridge in Series configuration, so more waiting!



When it arrivs hopefully I’ll finally complete the build!

O BTW, just when I thought this would be my first build without a cut, this happened:



So it's official.

[Philosophical45]

great analysis there.. you know men serious when they bring out all the applied sciences lol

[captainchris]

i have been saying this a while now, the idea for parallel is not efficient for temps, but for flow. all my builds and ones i help others with i so serial..that way each both/gpu gets the same flow rate and volume of water...

its not really applied since boy shoto, just common sense!

[MessiaaH]

parallel has it's advantages as well.
Reason I went it was because I using a chiller, so I wanted that chilled water to hit all 3 GPUs simultaneously.
And it's been working at optimal temperatures for the better part of a year.

But that same setup cant work with the dramatic drop in flowrate offered by this MORA.

I usually avoid parallel loops, and run independent loops if anything. Also, trying to cool cpu, 3 gpus, 2 banks of memory, nb, and mosfet in a serial config, the flowrate in that loop going to get bone hard. And using a chiller, yUh want optimal flowrate, cause then you have things like condensation to consider.

So using a chiller, that was the best configuration for that particular loop.
But I basically just plug the chiller out and replace it with a heavily restrictive rad, so yea, things had to change up.

So don't cuss the parallel setup eh, it have it's uses, and I think d 5GHz OC and Temps I had before can attest to that

Multi loops are also great, but 1 advantage to be had from a unified loop is, when combining yUh pumps yUh get increased head pressure and fail-over if 1 pump fails. Individual loops, if 1 pump fail, whatever on that loop boned. Unless yUh running a 4 pump loop lol.

[captainchris]

the 5ghz oc temps dont really have much to say with respect to the cards running parallel.. that getting water flowing through it like normal.

but i will hit the parallel hard because they dont return the best results..

if there is an option to choose..which there is..give me the one with the lowest temps!