First Real Attempt At Watercooling -- Advice Needed

[aigomorla]

And shame on Terry recommending Mayhem's. Ugh^22! That aurora junk will really mess up everything it touches in time. It's fine for "drag racing", i.e. use at a trade show and then when you come back home throw everything in the circuit away. That stuff is bad news. Most dyes are but that stuff...might as well put valve grinding compound in your crank case oil!

lol... i agree with you on the drag racing...

but the pearl effects are really really cool....

https://www.youtube.com/watch?v=o85DzpdoQjM


Also i used to test his dyes for him... i have a lot left over from the UV reactive, and basic composite.
His color dyes are not that bad, they are concentrated so u only need a couple of drops.

However dyes fade over time, and they eventually dye up your tubing and walls. Some people however still want dyes, and well, i cant say no i wont help you. I can only point people to the lesser of the evil.

 

[MiRai]

Also, the Swiftech block is a poor choice. Very high backpressure and subpar performance. EK (copper/acetel, forget nickel!), newest HeatKiller, etc. are far superior designs with lower backpressure and core temps.

So many people coming out from nowhere after I had already purchased the Swiftech XL. Where were you people when I needed you to begin with?!

After looking at several more reviews, I went ahead and grabbed the HeatKiller IV Pro, but it'll be here earlier than my back-ordered cable management stuff will be, so I'm still playing the waiting game.

And shame on Terry recommending Mayhem's. Ugh^22! That aurora junk will really mess up everything it touches in time. It's fine for "drag racing", i.e. use at a trade show and then when you come back home throw everything in the circuit away. That stuff is bad news. Most dyes are but that stuff...might as well put valve grinding compound in your crank case oil!

lol... i agree with you on the drag racing...

but the pearl effects are really really cool....

https://www.youtube.com/watch?v=o85DzpdoQjM

It's unfortunate that such dyes are harmful to a loop and are unable to be used for long periods, since they look really nice.

 

[MrTeal]

It's unfortunate that such dyes are harmful to a loop and are unable to be used for long periods, since they look really nice.

Yeah, that pearl looks crazy. It almost seems like it should be a viscous fluid and cool terribly, with how it flows through the loop.

 

[aigomorla]

Yeah, that pearl looks crazy. It almost seems like it should be a viscous fluid and cool terribly, with how it flows through the loop.

its actually pretty bad... it deposits on everything... you get a speckles on your blocks which u need to physically scrub.
And well, i can tell you first hand on the pains of re oringing a used gpu block...

its not something i like to do, nor would find an excuse to do when bored.

 

[Kenmitch]

No update?

 

[MiRai]

No update?

Unfortunately, no. It's been a slow moving process around here because of work (as well as the fact that I had to wait almost two weeks for cable combs), but I expect to get something done later this week, or this weekend.

 

[Aristotelian]

I was really hoping for a nice update too, MiRai. There are no red/black s-frames left that I can find and the other s-frame is almost EUR 900. I'm going to want to sand it down or find a way to strip the paint job and re-paint the case, which I'm sure will take a lot of time.

I want to colour coordinate a super build in an s-frame next year - after next gen video cards, and skylake-e. And this thread has already been super helpful so I know that I shouldn't expect any kind of colored liquid...

Come on, get that build up!

 

[MiRai]

Not to leave you guys hanging...

I did my cable management, but I think I want to redo it. So, in addition to some cable extensions (and SATA cables) that were already needed, I went ahead and got some more cable clamps that will be here later this week.

This last weekend I had to finish some last minute painting, as well as had to tackle what looked to be a game-breaking clearance issue with tube routing. It seemed as if the front glass of the chassis was going to interfere with routing, but then I got smart and figured out how to do it, so everything seems fine at the moment.

I will say that I wasn't prepared for the amount of force required to push the tubing onto the fittings, but again, I got smart (at least I hope so) and ran the ends of the tube under some warm water to allow it to flex a bit easier.

At this point I am hoping to have water in it by this weekend so that I can see if I've got any leaks, which I sure hope I do not. ::fingers crossed::

(I should also update the OP at some point because people stumbling upon this thread might have some incorrect information unless they read all the pages.)

 

[MiRai]

I've updated the original post with the following:

Alright, so the build has taken much longer than I had anticipated (and it's still not done). There were plenty of, we'll call them, mistakes that were made in the last two months, but in the end, I went with the following parts:

CPU Block - Watercool Heatkiller IV Pro
Radiator - Hardware Labs Black Ice SR2 Multiport 360mm
Pump/Reservoir - XSPC D5 Photon 170 Reservoir/Pump Combo
Misc - 1/2" OD Tubing / Bitspower Fittings (Barbs)

I had originally purchased 3/4" OD tubing with Bitspower compression fittings, but I just couldn't thread the compression "nut" onto the fitting itself after the tubing was in place without an unreal amount of force. I was very unhappy about that, so I switched everything up to barb fittings, and dropped the size of the tubing down a bit because I felt that the 3/4" OD was quite massive; the final shipment of tubing and fittings should be here on Wednesday. Based upon what has happened with everything so far, I can only assume that there will be additional problems once I finally have water in the system and am testing for leaks, like vibration or... something else. I will say that this was quite a learning experience and, if the water can actually keep the CPU at an acceptable temperature while it is overclocked, then I will likely do it again in the future, however, if/when I do it again, I'll be using a much more conventional chassis.

 

[MiRai]

Alright, so there is water in my loop. I had some issues with trying to figure out why the pump wasn't pumping, but I realized that I needed to have the ATX 8-pin connectors hooked up so that it could get a signal from the motherboard fan header. Hopefully, if there is a leak, that nothing gets fried.

One question at the moment... should I leave the top plug on, or off while I'm leak testing and trying to get the air out? I'm talking about the plug at the top of the reservoir where I filled it up. At the moment, I have it off so that any air can escape (at least that is what I believe is happening by leaving it off).

EDIT: I've also been tilting the chassis in order to help push any air bubbles around that might be trapped somewhere.

 

[aigomorla]

leaving the top plug off or on wont matter.

I would leave it on because ive had it where when i turned the system off, the water would overflow out the plug due to the disappearance of negative pressure the pump caused.

 

[guskline]

I agree with aigomoria. Let your system run and daily check the fluid level in the reservoir and make sure it's sufficiently above the return hose to eliminate cavitation in the reservoir.

In my 4790k rig, I use the Swiftech mcp35x pump with the small tube reservoir which attaches to the top return port. It took a number of frequent small fillings until all of the air bubbles had been eliminated from the system. The heighth between the top of the return port and the top of the tube is small so I had to keep an eye on the water level.

BTW, I always use just distilled water with a drop or two of Biocide.

 

[MiRai]

Alright, so I did some leak testing and then decided that I wanted to route one part of the loop slightly different, which set me back half a day but also gave me a reason to check the drain setup on the radiator (plus a flush of the system wasn't such a bad idea). All seems well for now and I'm busy dialing in my overclocking at the moment.

It looks like 4.6GHz @ ~1.26v is what I'm getting after a stress test that was done through most of the day, so that's exciting. The water seems to be keeping it nice and cool at ~70C after 8 hours.

I'll have proof of all of this and a long video that shows the build in the coming weeks. I have to sift through probably 40 hours of footage to see how I should piece everything together, but I'll bump this thread whenever it's ready.

Thanks, again, for the advice and the interest over the past few weeks/months.

 

[Kenmitch]

4.6GHz is a nice overclock for that chip.:thumbsup:

Looking forward to the update.

 

[guskline]

Alright, so I did some leak testing and then decided that I wanted to route one part of the loop slightly different, which set me back half a day but also gave me a reason to check the drain setup on the radiator (plus a flush of the system wasn't such a bad idea). All seems well for now and I'm busy dialing in my overclocking at the moment.

It looks like 4.6GHz @ ~1.26v is what I'm getting after a stress test that was done through most of the day, so that's exciting. The water seems to be keeping it nice and cool at ~70C after 8 hours.

I'll have proof of all of this and a long video that shows the build in the coming weeks. I have to sift through probably 40 hours of footage to see how I should piece everything together, but I'll bump this thread whenever it's ready.

Thanks, again, for the advice and the interest over the past few weeks/months.

WOW that's good. What mb and ram with that 5960X? Do you have it at 46x100 or did you raise your blck to 125?

 

[MiRai]

WOW that's good. What mb and ram with that 5960X? Do you have it at 46x100 or did you raise your blck to 125?

The motherboard I have is the MSI Godlike, and I'm using a straight multiplier here because BCLK is giving me less-than-stellar results. I don't want to derail the thread with non-watercooling stuff, but...

The RAM I purchased is advertised for 2800MHz. Unfortunately, I didn't realize that 2800MHz isn't a real "step" of memory speed, and the XMP profile sets the BCLK to 127MHz with the RAM at 2200MHz (or something similar) to reach the advertised 2800MHz. I say "unfortunately" because with such a high BCLK I can't seem to get the CPU to overclock reliably at all.

So, I dialed back the BCLK to 100MHz and the RAM to 2400MHz because I wanted to see what my chip could do by just playing the multiplier. After some quick testing I figured out that it likes 4.6, since 4.7 required another ~.6v in order to get it to run more than 30 minutes on a stress test. That extra .6v raised temps an additional 10C to almost 80C, which I wasn't happy with.

I then tried to mess with the BCLK a bit to see if I could increase the speed of the RAM while keeping the CPU around 4600MHz, but for some reason when using the BCLK to overclock the CPU it wants to use a lot more voltage than when the BCLK is set to 100MHz. This obviously raises temps, again.

It seems like 4.6GHz is where it's at, and anything beyond that is going to require more finagling than I'd really care to participate it, especially when stress testing takes forever. The only thing I'm trying to do at the moment is get the RAM to run stable at 2666MHz, but I'm continuously failing to do that and I'm not so sure I want to pump a bunch of voltage into my DIMMs for a measly 266MHz.

...maybe if they were water-cooled.

EDIT: Since guskline mentioned it, I decided to look a bit deeper into HW-E overclocking via BCLK, I decided to try it because I was reading that reaching RAM speeds over 2400MHz was very difficult w/o BCLK. So, I decided to set the BCLK to 125MHz (not the oddball 127.xx that the RAM's XMP profile used), and I was able to set the RAM to 2666MHz (using 2133MHz as a starting point), and the CPU to 4625MHz (37x125). So far, I'm 2.5 hours into stress testing this and so far, so good. Like I said above, using BCLK seems to suck more voltage and right now it's using between 1.304v and 1.312v, which pushes the temps up to 74C - 75C while under full, constant load. I don't know if I'm going to stick with this since I'm not entirely sure what the consequences are for the rest of the components when using such a high BCLK, but we'll see.