(Anti) SFF fun house

[tabascosauz]

Hmm. So i got my new 5900x and the problems persist. My write speeds are always messed up. Across all caches the read/ write/ copy ALWAYS vary across l2 and l3. For the l1 cache it very often. From what i understand a variation in over 20 gb/s shows some sort of instability. This problem is even on stock. I cannot figure it out i have replaced everything. Last thing i have is i replaced my PSU with the same PSU model. Maybe 5900x doesn't play nicely with corsair rm850x? But then again my laptop is starting to come out of sync too. (4800h). Could i have like bad electricity that is frying my hardware?

Did you already RMA your 5900X and this is a new CPU? Got any screenshots of this?

I have to admit I don't care about cache results much. There's too much that can impact that - AIDA being AIDA, unstable CPU OC, unstable RAM OC, unstable IF, insufficient VSOC or minor voltages, bad windows config (too corrupted/damaged, or too many background programs), bad CPU, bad board...... As long as it's roughly in the right ballpark I haven't paid too much attention.

After my experience, I would probably look towards the board and the RAM. Did you say you were on DJR? If you're able to I'd just pick up a B-die kit from Amazon (30 day returns) and see if anything changes.

20GB/s seems wayyyyy too stringent a requirement for any L1/L2/L3. If you're running at stock then even the boost algorithm alone will naturally account for run-to-run variances bigger than that.....100GB/s is where I'd call it instability, and 200GB/s a bug.

 

[Mussels]

Hmm. So i got my new 5900x and the problems persist. My write speeds are always messed up. Across all caches the read/ write/ copy ALWAYS vary across l2 and l3. For the l1 cache it very often. From what i understand a variation in over 20 gb/s shows some sort of instability. This problem is even on stock. I cannot figure it out i have replaced everything. Last thing i have is i replaced my PSU with the same PSU model. Maybe 5900x doesn't play nicely with corsair rm850x? But then again my laptop is starting to come out of sync too. (4800h). Could i have like bad electricity that is frying my hardware?

No, cant be bad electricity.
You really do need to fill out your system specs, people can't help you with part of the information.

The top link in my sig is a work in progress guide for people like you stuck with mystery problems, go read through it and try the troubleshooting steps - remove un-needed hardware and software, disconnect from the internet, etc.

 

[tabascosauz]

Polling the water crowd:

Anyone knowledgeable about building loops can check over my parts list, and help with fittings? I figure it'll be a while before I can even think about Caselabs in realistic terms, so considering that I probably won the QC lottery on my old white Cerberus it seems like a good opportunity to dip into water with a CPU-only loop (Barrow/Corsair blocks not available for 2060S/2070 FE anymore).

I was originally thinking of a simple 280mm AIO, but the Cerberus makes it hard without having the PSU covering the board.. Pure Loop (lack of quality) and Coreliquid R (no 280mm) not really viable.

This is my list so far, minus fittings and other miscellaneous tools:

So while I'm pretty sure the other choices won't change too much (Iceman res and Optimus block bought already), but fittings........I was originally was just going to put 12 x regular fittings due to the EK configurator saying I only needed 6.................until I realized that I'm probably going to need at least 2 x 45 degree fittings, probably 2-4 extenders and a few 90 degree and 90 degree swivel fittings. Not to mention any drain valve/T-splitters etc.

Got a couple of days to work out the fittings - have taken a bit of a liking to the EK fittings, and the rest of the order needs to wait a bit since DDC isn't in stock until 19th.

 

[Mussels]

I dont use aqua, but i got some tips and pics
I am *new* to the watercooling world, but i got a good 6 months of trial and error going on here - my setup is ghetto, over complicated and i love it cause all the dumb ideas were my own.

(Hardest anti-lazy lesson: do not twist soft tubing to attach or remove fittings. Hose on/off first)

Use rotaries a lot, they're LIFE savers.
The build will be 3 dimensional, so sometimes you'll need more bits than a 2D diagram can show you - or simply need to have a hose move off to the left 45 degrees, before it lines up with the next component

QDC's basically dont exist small. I own some, and gave up on ever making them fit.

You need two drain/fill valves, EK has some super compact ones, or you can use ball valves (bigger, cheaper) - do NOT skip drain and fill ports! You can use a Y fitting (I use the cube style) to help fit these in odd places.
You definitely want one at either end of the loop... you have no idea how much easier it is to drain when you can open both sides up.

Mine needed tipping and tilting and took a dozen tries to drain, but then i added one cube style joiner at the top rad and bam, cracking it just drained the left side of the loop instantly.
The right side (with my res) was still full and not pouring out - closing the left side one and cracking open the top of the res on the right, bam that side drained.

Here is my two drain/fill locations, since the pump is located below the res it stops the pump from needing to spin, to drain. My res has 4 plugs on it which is overkill, but simply twisting those to open lets air in, and WHOOSH coolant goes out the lower drain tap. I can also use the left one to fill up the GPU from above, seal that, then fill the res (instead of half filling, start pump for 10 seconds, turn off, fill again, etc)

Remember that to fill from those, i can screw in a G1/4 fitting and some spare hose and safely fill away from the PC (same with draining from a distance) - and blow air into them for draining

I also used a flow meter to help extend and stabilise that odd angled res to GPU intake position (face hugger optional, requires hoarding 80's action figures)

Top res has one rotary extender, then the Y splitting cube thing - back fitting is just a stationary extender, then a rotary 90 degree (keeps them level, and out of fans)

Then due to an annoying height difference, i used two 90 degree fittings and an extender for in and out of the front rad, and the worlds shortest soft tubing.
It's still flexible for small shifting and adjustments of pump/res/front rad movement, but using the extra fittings made it stay where the heck i wanted it to be.

I also use a SATA to USB adaptor (the kind made for 2.5" laptop drives) with an optional 12V power brick, so i can power NOTHING but the pump - the PSU isn't connected so while filling and testing
This is the newest variety of them: add a 12V brick, and bam - sata with 5V and 12V (Mine was keyed so i couldnt directly attach normal SATA devices, so i snapped a small piece off a sata double adaptor to live with it)

 

[Valantar]

I dont use aqua, but i got some tips and pics
I am *new* to the watercooling world, but i got a good 6 months of trial and error going on here - my setup is ghetto, over complicated and i love it cause all the dumb ideas were my own.

(Hardest anti-lazy lesson: do not twist soft tubing to attach or remove fittings. Hose on/off first)

Use rotaries a lot, they're LIFE savers.
The build will be 3 dimensional, so sometimes you'll need more bits than a 2D diagram can show you - or simply need to have a hose move off to the left 45 degrees, before it lines up with the next component

QDC's basically dont exist small. I own some, and gave up on ever making them fit.

You need two drain/fill valves, EK has some super compact ones, or you can use ball valves (bigger, cheaper) - do NOT skip drain and fill ports! You can use a Y fitting (I use the cube style) to help fit these in odd places.
You definitely want one at either end of the loop... you have no idea how much easier it is to drain when you can open both sides up.

Mine needed tipping and tilting and took a dozen tries to drain, but then i added one cube style joiner at the top rad and bam, cracking it just drained the left side of the loop instantly.
The right side (with my res) was still full and not pouring out - closing the left side one and cracking open the top of the res on the right, bam that side drained.

Here is my two drain/fill locations, since the pump is located below the res it stops the pump from needing to spin, to drain. My res has 4 plugs on it which is overkill, but simply twisting those to open lets air in, and WHOOSH coolant goes out the lower drain tap. I can also use the left one to fill up the GPU from above, seal that, then fill the res (instead of half filling, start pump for 10 seconds, turn off, fill again, etc)

Remember that to fill from those, i can screw in a G1/4 fitting and some spare hose and safely fill away from the PC (same with draining from a distance) - and blow air into them for draining
View attachment 225493

I also used a flow meter to help extend and stabilise that odd angled res to GPU intake position (face hugger optional, requires hoarding 80's action figures)

Top res has one rotary extender, then the Y splitting cube thing - back fitting is just a stationary extender, then a rotary 90 degree (keeps them level, and out of fans)
View attachment 225495

Then due to an annoying height difference, i used two 90 degree fittings and an extender for in and out of the front rad, and the worlds shortest soft tubing.
It's still flexible for small shifting and adjustments of pump/res/front rad movement, but using the extra fittings made it stay where the heck i wanted it to be.

View attachment 225502

I also use a SATA to USB adaptor (the kind made for 2.5" laptop drives) with an optional 12V power brick, so i can power NOTHING but the pump - the PSU isn't connected so while filling and testing
This is the newest variety of them: add a 12V brick, and bam - sata with 5V and 12V (Mine was keyed so i couldnt directly attach normal SATA devices, so i snapped a small piece off a sata double adaptor to live with it)
View attachment 225503

- Having two fill/drain ports is pretty much a must if you go that route, as water doesn't come out unless air can come in, and doing both through a single small hole at the end of a tube is a major hassle. One + QDCs (with a spare QDC set and some tubing) gives you three openings, which is even better than two, so that will work well.

- Rotary fittings are an absolute necessity, yes. I wouldn't even try a build without them, unless I had tons of space.

- Alphacool's industrial nylon QDCs are reasonably sized. Worth considering. They still aren't small, and the need for enough flexibility to be able to connect and disconnect them is a consideration that is easy to miss, but I still fit three in my Meshlicious.

- If you have a DDC that can run off of a fan header, powering that externally is pretty easy.

@tabascosauz I'll try to give some more detailed feedback when I'm back from work

 

[tabascosauz]

Lots of good info to take in, thanks guys.

@Mussels for the time being I'm keeping my SGX650 around as my bench/tester PSU, and I've got a billion of those self tesr 24-pin plugs so I can run the pump off that when filling while the PC and SF750 remain off. AFAIK EK's DDC 4.2 has a PWM header and a SATA power lead? Not sure if that means I have to use both, or the PWM would be fine if I hooked up to the beefy PUMP header on my board or an Aquaero wait no I'm stupid lol, it's okay I'll have to get a sata power extension sooner or later

@Valantar I'm not sure I have anywhere in this loop that's high enough for a second fill port. The Iceman res has a fill port on the top that's supposed to be where the coolant goes on, as well as the drain port on the bottom next to the DDC, although I've never seen that one in action yet which is why I was entertaining the idea of another drain on the rad. Seems like using the drain on the res would require a lot of tilting of the PC........

If I was doing an internal EK 80FLT or 120TBE certainly would solve that problem, but fitment would be such a bitch (I'd maybe mount it horizontally hanging off the top fan bracket, and that's if I'm lucky), would prefer to keep it as last resort if I can.

 

[Chrispy_]

Is there really no waterblock for it, and if no - is it not worth just flipping it for a GPU that you can find a block for?

 

[tabascosauz]

View attachment 225509

Is there really no waterblock for it, and if no - is it not worth just flipping it for a GPU that you can find a block for?

Maybe 6 months ago I could have gotten a 3070 or 6700XT. Right now the GPU drought is real bad where I am, in-store stock reduced to a single lonely GTX 1050 Ti. At least there used to be near-MSRP 3070 Ti and 3080, and inflated 6700XT/6800XT.

There is no waterblock for it because it looks like this in case you forgot, if you manage to get past all the glue without breaking something in disassembly:



There used to be a short Corsair block that kinda works if you fold the pigtail over on itself to come out where it should have. There's a grand total of one dude on the internet that's tried this. Also a Barrow block that could potentially work.

Both blocks are discontinued. Plan is to run the 2060S as is, and upgrade to something more deserving of being in the loop maybe next year. It's just past 2 years old now.

 

[Chrispy_]

Maybe 6 months ago I could have gotten a 3070 or 6700XT. Right now the GPU drought is real bad where I am, in-store stock reduced to a single lonely GTX 1050 Ti. At least there used to be near-MSRP 3070 Ti and 3080, and inflated 6700XT/6800XT.

There is no waterblock for it because it looks like this in case you forgot, if you manage to get past all the glue without breaking something in disassembly:

View attachment 225510

There used to be a short Corsair block that kinda works if you fold the pigtail over on itself to come out where it should have. There's a grand total of one dude on the internet that's tried this. Also a Barrow block that could potentially work.

Both blocks are discontinued. Plan is to run the 2060S as is, and upgrade to something more deserving of being in the loop maybe next year. It's just past 2 years old now.

There are official RTX 2060FE/2070FE/2060SFE blocks from both Corsair and alphacool that simply leave a cutout for the power cable.
I'm well aware of this as I'm using a 2060FE myself in the HTPC until I can find a 3060Ti that's small enough to replace it.

Buying a new GPU is mad right now but If you're prepared to spend the money on a waterblock it's probably easier to sell the 2060SFE and just slap a vanilla 3060 in there for which there are ample waterblocks and plenty of very short PCB designs out on the market already. 3060 LHR cards aren't even commanding that much of a premium in the messed-up GPU market right now.

 

[Valantar]

Polling the water crowd:

Anyone knowledgeable about building loops can check over my parts list, and help with fittings? I figure it'll be a while before I can even think about Caselabs in realistic terms, so considering that I probably won the QC lottery on my old white Cerberus it seems like a good opportunity to dip into water with a CPU-only loop (Barrow/Corsair blocks not available for 2060S/2070 FE anymore).

I was originally thinking of a simple 280mm AIO, but the Cerberus makes it hard without having the PSU covering the board.. Pure Loop (lack of quality) and Coreliquid R (no 280mm) not really viable.

This is my list so far, minus fittings and other miscellaneous tools:

View attachment 225483

So while I'm pretty sure the other choices won't change too much (Iceman res and Optimus block bought already), but fittings........I was originally was just going to put 12 x regular fittings due to the EK configurator saying I only needed 6.................until I realized that I'm probably going to need at least 2 x 45 degree fittings, probably 2-4 extenders and a few 90 degree and 90 degree swivel fittings. Not to mention any drain valve/T-splitters etc.

Got a couple of days to work out the fittings - have taken a bit of a liking to the EK fittings, and the rest of the order needs to wait a bit since DDC isn't in stock until 19th.

View attachment 225491

First off, a question: what is the flow direction here? Beyond that, here's some more thorough feedback:

• I don't really see the point of the extender stack in the front. Is there a specific idea behind it? That's a good place for the temp sensor though. Still, I'd just run the two tubes from the rad in similar ways - that's likely to look the cleanest anyhow. No reason to overspend on extenders that don't do anything much.
• That pump-res-to-CPU connection looks tricky. If that is the inner connection on the res, that run is going to be very short. Given how stiff ZMT 10/16 tubing is, you might have trouble getting it in place unless your fittings happen to line up perfectly - even a slight bend in a run that short will be difficult to do. You might want to get a double-45 fitting (or just stack two 45-degrees) to get sufficient flexibility in the angles of your tubing.
• One useful trick for working with ZMT is to soak it in warm water if you're doing anything small that needs bending, as that makes it significantly softer and easier to work with. Similarly, it can be hard to get onto thick barbs or compression fittings, but wetting the inside of the tube slightly helps massively with that.
• The drain valve on the bottom of the res is useful, but will require quite some tilting of the case to drain the loop, simply because it's above most of the loop.
• Having another plug at the top of the res will alow water to run out the bottom of it easily, but you'll still struggle with the tilting. And tiliting a case around with open plugs in the loop can get pretty messy
• If you don't want the expense of QDCs all around, where you've added them is a good place for one - allows for relatively easy filling/draining of both the reservoir and the radiator. Remember to buy at least one more than what's in the loop if you plan to use it for filling and draining. If you want a second QDC in the loop, I would put it on the CPU-radiator line. And, as I said, Alphacool's industrial nylon QDCs are great as long as you can live with them not being entirely drip-free. Nothing a bit of paper towel can't handle, and you easily get two of those for the price of a single Koolance or other brass one.
• If the CPU-res connection is too tricky, you can likely use the outer connection on the res for that and use two 45-degree fittings on res-rad tube to offset that tube slightly and clear the SO-DIMM.2 (i.e. one 45-degree angling the tube up towards the top of the case slightly, and the second then pointing it straight ahead again, but slightly higher up).

I'll find some pics of my various setups so that you can se what bending ZMT in tight quarters is like - I've struggled with it before, but have gotten more comfortable with it over time.

Edit 1: Here's my OG water setup in the H200i. (I had a loop in a define R4 before that, but it's not worth mentioning in this context.) That CPU-rear rad run you can see is quite tricky. It has a rotary 45° on the CPU block and a barb on the rad, and it was tight. Definitely put some uncomfortable stress on that CPU fitting to get the tubing in, but it worked. Plz ignore the anti-kink springs - I added them, but they did essentially nothing - ZMT is plenty anti-kink on its own. You can also see how that rad-GPU run bends outwards and back in to reach the GPU - that should give you some indication of the bend radii possible with this tubing.

Edit2: Here's the CPU side of my Meshy. That micro res-front rad tube bend is about as tight a bend as you'll get with ZMT without it pushing back with quite some force.

Edit3: And here's the GPU side, demonstrating what happens when you dont' have a lot of space, have QDCs, a limited amount of fittings, and a sub-optimal flow path. There isn't enough clearance between the GPU and side panel for the tubing to cross there, hence the overlong runs and general mess. I've since gotten confirmation from EK that the GPU block is bidirectional, so I'll be swapping around the inlet and outlet ports and significantly shortening the runs at some point - probably the holidays.

Edit4: Here's the H200 with the Aquanaut. As you can see, I could no longer fit the rear rad with that installed - it conflicted with the ports, and flipping it around messed with the runs too much. That CPU-GPU run is another demonstration of the bend radius of 10/16mm ZMT. It's routed that way for cleanliness + enough slack to easily connnect and disconnect. You can se I use double-stacked 45s to angle the tubing past the fans on one port.
Edit5: And apparently I forgot to paste the last image link, lol. Here goes:

 

[tabascosauz]

Holy crap, ordered the Foundation block last night and UPS delivery is already scheduled for tomorrow evening. Meanwhile the Iceman will take two weeks...... really starting to think about EK 80FLT mounting to the top bracket.

@Valantar no real idea behind the extenders, just saw a bunch of cerberus bottom rad loops do it haha. Figured it might help with the bend if I just stack extenders. Not pretty, but the way it is right now I figure if I only use two regular fittings (+water temp) on the rad, the bend will push down on the GPU (esp. the long res-rad run).

As for flow direction, on the iceman the inlet is closest to the CPU/board and outlet is away from the board. Thinking I might have to take a chance on the SO-DIMM.2 and make the long res-rad run connect to the inlet. Should give me more room to work on the res-block bend.

You reckon it might be easier to run the tube further out a bit from the res outlet, then do a big bend back towards the block and end in a 45" fitting (do I need a swivel there too)? Not sure which is in/out on the block yet, but the acrylic top is reversible which ever way.

Not sure how I feel about the QDC if it just naturally leaks/is a significant weak point. I'm not sure why I put it in the res/rad run. I don't swap hardware at all.

Here's the main inspiration. This build was also with iceman, I'm not sure how they managed the res/CPU bend with just a 45 degree and a rotary. Also stacking extenders here, I thought double stacking might alleviate the res/rad and rad/CPU bends.

 

[tabascosauz]

There are official RTX 2060FE/2070FE/2060SFE blocks from both Corsair and alphacool that simply leave a cutout for the power cable.
I'm well aware of this as I'm using a 2060FE myself in the HTPC until I can find a 3060Ti that's small enough to replace it.

Buying a new GPU is mad right now but If you're prepared to spend the money on a waterblock it's probably easier to sell the 2060SFE and just slap a vanilla 3060 in there for which there are ample waterblocks and plenty of very short PCB designs out on the market already. 3060 LHR cards aren't even commanding that much of a premium in the messed-up GPU market right now.

Can't find the alphacool block anywhere, the Corsair one is discontinued like I said. On the face of it, there's not much difference between FE and reference PCB aside from the pigtail, but if you have to hang that out the side, then what's the point of spending money to integrate it into the loop if it's just gonna look like puke?

The only other dGPU I have lying around is my GT610 for benching.

 

[Valantar]

Holy crap, ordered the Foundation block last night and UPS delivery is already scheduled for tomorrow evening. Meanwhile the Iceman will take two weeks...... really starting to think about EK 80FLT mounting to the top bracket.

@Valantar no real idea behind the extenders, just saw a bunch of cerberus bottom rad loops do it haha. Figured it might help with the bend if I just stack extenders. Not pretty, but the way it is right now I figure if I only use two regular fittings (+water temp) on the rad, the bend will push down on the GPU (esp. the long res-rad run).

As for flow direction, on the iceman the inlet is closest to the CPU/board and outlet is away from the board. Thinking I might have to take a chance on the SO-DIMM.2 and make the long res-rad run connect to the inlet. Should give me more room to work on the res-block bend.

You reckon it might be easier to run the tube further out a bit from the res outlet, then do a big bend back towards the block and end in a 45" fitting (do I need a swivel there too)? Not sure which is in/out on the block yet, but the acrylic top is reversible which ever way.

Not sure how I feel about the QDC if it just naturally leaks/is a significant weak point. I'm not sure why I put it in the res/rad run. I don't swap hardware at all.

Here's the main inspiration. This build was also with iceman, I'm not sure how they managed the res/CPU bend with just a 45 degree and a rotary. Also stacking extenders here, I thought double stacking might alleviate the res/rad and rad/CPU bends.

View attachment 225568

Ah, I see. I think the point of the extenders is to make removing and inserting the side rad easier, as not having the tubes bend down makes them less prone to snagging and gives them a more easily accessible point of rotation. It also allows the use of rotary fittings there instead of just having the tubes clamped down to the radiator, again making it easier to shift the side rad out of the way. I'm a bit stumped by that single QDC though, as I would have assumed they would want the ability to remove the rad entirely? Though I guess there could be a second one all the way towards the rad that isn't visible in that shot. Still, unless you plan on adding a side radiator later, I would skip the extender stack.

I doubt the tubing bends will push down meaningfully on the GPU, as the stiff tubing will naturally want to bend around. It's of course possible, but I woudlnt' worry about it.

If I understand your description correctly, that does sound like an easier way of making the res-CPU run work. I'll throw together a sketch of what I think would be a sensible combination of fittings etc. in a little while.

To be clear: the Alphacool QDCs don't leak (at least in my experience), they just let out a few drops of liquid every time you disconnect them. I just wrap a sheet of paper towel around the tube while disconnecting, then give them a slight shake, as that catches everything. Drip-free ones are supposed to not even do that - AFAIK Koolance QDCs are drip free. It definitely takes this from "needs a bit of preparation and presence of mind" to "connect and disconnect at will, no worries", so that's potentially an advantage for people who often make changes to their loops, but if not (like in my case), I don't see 1-2ml of liquid escaping as a problem (and getting Koolance QDCs where I am is not easy). It's still miles better than needing to drain the loop for every little thing, and I've had a lot of luck using my spare res and pump for filling. Your Iceman res does give you some advantages I lack, but I would still recommend a secondary fill/drain point, and QDCs are far better for that than any splitter+ball valve setup (as those take a lot of space).

To be clear, all angled fittings should be rotary fittings unless you have a very specific use case - non-rotary angled fittings essentially end up pointing in random directions depending on the threading, which is hardly ideal.

Do you have a link to any more pics of that build? I'm curious about the CPU-res run as well (and I don't quite understand where the res is, given that rear fan? Is it just blowing straight into the acrylic part of the res?). Looks tight enough that compression fittings are essentially a requirement, but it looks like things lined up nicely for them.

 

[tabascosauz]

Ah, I see. I think the point of the extenders is to make removing and inserting the side rad easier, as not having the tubes bend down makes them less prone to snagging and gives them a more easily accessible point of rotation.

I doubt the tubing bends will push down meaningfully on the GPU, as the stiff tubing will naturally want to bend around. It's of course possible, but I woudlnt' worry about it.

If I understand your description correctly, that does sound like an easier way of making the res-CPU run work. I'll throw together a sketch of what I think would be a sensible combination of fittings etc. in a little while.

Realized I will need an offset fitting for res/rad run because the SO-DIMM.2 does get in the way. Fortunately it's not that big and the inlet/outlet already sit 20-30mm above the GPU. But that creates an issue in that wouldn't the entire tube just want to fall back down onto the SO-DIMM.2 instead of staying offset (to the north) as it should be?

I thought about it a bit more and I don't think I have the room to make a huge about-face on the res/CPU run - ie. to have it go past the block then come back diagonally into the 45° on the block. But seeing as the res/CPU tube will be on the outside, there might be a chance to run it through a 90° straight up to the 45° on the CPU. Eyeballing it, the angles look close, and the swivel on the 90° might do the trick.

Still working on the layout of the rad side, but how's this:

To be clear: the Alphacool QDCs don't leak (at least in my experience), they just let out a few drops of liquid every time you disconnect them. I just wrap a sheet of paper towel around the tube while disconnecting, then give them a slight shake, as that catches everything. Drip-free ones are supposed to not even do that - AFAIK Koolance QDCs are drip free. It definitely takes this from "needs a bit of preparation and presence of mind" to "connect and disconnect at will, no worries", so that's potentially an advantage for people who often make changes to their loops, but if not (like in my case), I don't see 1-2ml of liquid escaping as a problem (and getting Koolance QDCs where I am is not easy). It's still miles better than needing to drain the loop for every little thing, and I've had a lot of luck using my spare res and pump for filling. Your Iceman res does give you some advantages I lack, but I would still recommend a secondary fill/drain point, and QDCs are far better for that than any splitter+ball valve setup (as those take a lot of space).

Okay that's fine then. I'm starting to think I could benefit from just having 2 QDCs on the two longest runs across the GPU, so that if I ever need to do anything in the GPU/bottom of case area, or need to do anything with the fans/rad I have the space to do it.

To be clear, all angled fittings should be rotary fittings unless you have a very specific use case - non-rotary angled fittings essentially end up pointing in random directions depending on the threading, which is hardly ideal.

Yeah, non-rotating angled fittings sounded about as mindbogglingly stupid in my mind, but sleep-deprived me wanted to be sure

Do you have a link to any more pics of that build? I'm curious about the CPU-res run as well (and I don't quite understand where the res is, given that rear fan? Is it just blowing straight into the acrylic part of the res?). Looks tight enough that compression fittings are essentially a requirement, but it looks like things lined up nicely for them.

The res is supposed to be directly mounted to the case, but he's done something custom and installed both it and the fans on some really long screws or something. So the res is like an inch off the back of the case, so the 92mm can exhaust out. So far I don't see a need to have a fan there, the 92mm there in any M1 layout case doesn't do much except maybe help out an air cooler a little.

 

[tabascosauz]

There are official RTX 2060FE/2070FE/2060SFE blocks from both Corsair and alphacool that simply leave a cutout for the power cable.
I'm well aware of this as I'm using a 2060FE myself in the HTPC until I can find a 3060Ti that's small enough to replace it.

Buying a new GPU is mad right now but If you're prepared to spend the money on a waterblock it's probably easier to sell the 2060SFE and just slap a vanilla 3060 in there for which there are ample waterblocks and plenty of very short PCB designs out on the market already. 3060 LHR cards aren't even commanding that much of a premium in the messed-up GPU market right now.

Turns out Google isn't omnipotent yet - found both the Alphacool block and Corsair block on PPCS. I have to have a good think about this though, this late into its life. Trying to keep it simple for a first effort. Not sure I stand to gain too much from putting it under water, probably won't move it up a segment in performance.

The Bitspower block is VERY tempting though

 

[Mussels]

God, those alphacool disconnects are $15 USD but $45Au.
No wonder i never looked at them.


Oh and the reason behind using the USB-Sata adaptor is safety: with a 12V 2A brick, i'm at 24W of danger. Not 650W or whatever the spare PSU is.


I'm now even looking at Bykski quick disconnects... they may be a cheap option for you, too.

 

[Chrispy_]

Not sure I stand to gain too much from putting it under water, probably won't move it up a segment in performance.

The Bitspower block is VERY tempting though

This.

Watercooling the 2060S is academic at this point, the only advantage is that it'll let you build the loop for GPU watercooling now, which makes life less of a headache when you do eventually replace the 2060S.

 

[Valantar]

Realized I will need an offset fitting for res/rad run because the SO-DIMM.2 does get in the way. Fortunately it's not that big and the inlet/outlet already sit 20-30mm above the GPU. But that creates an issue in that wouldn't the entire tube just want to fall back down onto the SO-DIMM.2 instead of staying offset (to the north) as it should be?

I thought about it a bit more and I don't think I have the room to make a huge about-face on the res/CPU run - ie. to have it go past the block then come back diagonally into the 45° on the block. But seeing as the res/CPU tube will be on the outside, there might be a chance to run it through a 90° straight up to the 45° on the CPU. Eyeballing it, the angles look close, and the swivel on the 90° might do the trick.

Still working on the layout of the rad side, but how's this:

View attachment 225609



Okay that's fine then. I'm starting to think I could benefit from just having 2 QDCs on the two longest runs across the GPU, so that if I ever need to do anything in the GPU/bottom of case area, or need to do anything with the fans/rad I have the space to do it.



Yeah, non-rotating angled fittings sounded about as mindbogglingly stupid in my mind, but sleep-deprived me wanted to be sure



The res is supposed to be directly mounted to the case, but he's done something custom and installed both it and the fans on some really long screws or something. So the res is like an inch off the back of the case, so the 92mm can exhaust out. So far I don't see a need to have a fan there, the 92mm there in any M1 layout case doesn't do much except maybe help out an air cooler a little.

Is the outside fitting on the Iceman res above the top of the SO-DIMM.2? If not, then your proposed layout will still conflict with it.

Anyhow, I made a quick sketch of how I would attempt this. Please note: I have so far never successfully planned a loop; my tubing runs and routing always end up changing. But this is my current best guess at something that might work:

Some of my reasoning:
CPU-res connection: A single rotary fitting gives you some flexibility in routing (can point down towars the socket or be angled more upwards for a larger curvature), while not having two saves you from needing to match up their angles. This run will bend quite sharply, and the perspective of my sketch makes it look much longer than it will be. But it should be doable. Not having a rotary fitting on the CPU also makes it easier to install, as you'll have a more secure base to push the tubing onto. Install sequence: tube on the rotary fitting first, bend as needed, then onto the CPU. Be mindful that adjusting the length of tubing will help you adjust the bend radius as well. Cut a bit longer than you think you need, then adjust as necessary. I would use compression fittings here for additional security.
Res-rad connection: Using two 45-degrees creates less of a flow restriction than using an offset fitting, but either will work, and this isn't a very restrictive loop anyway. The offset fitting will likely hold the tubing more in place (i.e. prevent it from sagging down onto the GPU/SO-DIMM.2), but they should be functionally the same. Either should let you get past the SO.DIMM.2 easily.
The CPU-rad connection is identical to your sketch.

Using two QDCs like that sounds like a good idea. That way you can easily remove your radiator for access to parts, or just for draining/cleaning, as rads are by far the most difficult part to drain. You also get a straight shot out of both the reservoir and CPU block for draining those. Protip: fill and bleed your radiator outside of the case. Those are always where the most air gets stuck, and with QDCs you can pull it out, fill it, shake it around to get the air out without having to shake the whole PC around, then put it back in once done. Really convenient!

Also, QDCs obviously let you add a GPU to the loop relatively easily later on.

Oh, and one additional advantage of the Alphacool nylon QDCs: they're obviously non-conductive, so there's no risk of them shorting out anything if they touch, nor is there any chance of them scratching other parts.

The ones I have have G1/4 female threads on both ends, but there is also a variant with male threads if you want to screw them directly into something (like the radiator). Could save you a couple of fittings that way, and you'd kind of get your extender stack back, just made of QDCs If you go that route, one male + two female should get you what you need: one male threaded for each radiator port, and matching female threaded parts for adding fittings and connecting to tubing.

God, those alphacool disconnects are $15 USD but $45Au.
No wonder i never looked at them.

Whoa, that's crazy. Regional price differences can be all over the place, sadly. Koolance isn't particularly expensive in the US, while here in Europe they're seemingly made from unobtainium, and a single part (i.e. half of an actual QDC) costs more than those Alphacool ones.

Turns out Google isn't omnipotent yet - found both the Alphacool block and Corsair block on PPCS. I have to have a good think about this though, this late into its life. Trying to keep it simple for a first effort. Not sure I stand to gain too much from putting it under water, probably won't move it up a segment in performance.

The Bitspower block is VERY tempting though

Hey, if you really want a water cooled GPU you can have my old Fury X for cheap

 

[tabascosauz]

Res-rad connection: Using two 45-degrees creates less of a flow restriction than using an offset fitting, but either will work, and this isn't a very restrictive loop anyway. The offset fitting will likely hold the tubing more in place (i.e. prevent it from sagging down onto the GPU/SO-DIMM.2), but they should be functionally the same. Either should let you get past the SO.DIMM.2 easily.
The CPU-rad connection is identical to your sketch.

After thinking more about it, I probably only need the medium or even small offset EK fitting, because the inlet tube already *almost* clears the SO-DIMM.2. Don't think the massive offset of 2 x 45° will be necessary.

The outlet tube is well clear of the SO-DIMM.2 and will not be an issue regardless of how I route the res-CPU run.

Or, alternatively, if I opt for the horseshoe bend below per Optimumtech, I could run the CPU block to the inlet instead so it's closer to the board and basically is what OT does. Then, the res-rad run would require no offset fitting at all because it doesn't have an issue with the SO-DIMM.2.

Is the outside fitting on the Iceman res above the top of the SO-DIMM.2? If not, then your proposed layout will still conflict with it.

Anyhow, I made a quick sketch of how I would attempt this. Please note: I have so far never successfully planned a loop; my tubing runs and routing always end up changing. But this is my current best guess at something that might work:

Man, that's a sweet diagram. Thanks for making one

I think Optimumtech makes a horseshoe shaped loop up and back down into a 90° on the CPU block actually. Not sure how I didn't notice this. He uses the X570 Impact as well so the socket placement is the same. Dare I say, the 180° bend has a bit more character than the other options. Cerberus is similar but a little bit bigger than M1 in every dimension that matters, so I think it might be doable.

Hey, if you really want a water cooled GPU you can have my old Fury X for cheap

But in all seriousness, I would probably get stuck at the glued-in fan connector during disassembly, and it also has some interesting coil whine that's usually covered up by the always-on fans (it's already my second FE, the first I sent back for nauseating coil whine). Should be enough reason to prevent me from impulse buying the Bitspower block.

 

[tabascosauz]

Welp, decided to just go for it and order most of my parts. Switched to PPCS as they seem much more reputable. A few remaining things (pump, Cryofuel and some misc and extra fittings) gonna order from EK, gotta wait for the DDC 4.2 to come back in stock first. Pray to god my Iceman res isn't getting stuck somewhere.

 

[tabascosauz]

Parts are all arriving now in droves, just waiting on the last EK order. I think this is the layout I'll go with.

No QDCs for the time being as I just don't see myself adding a GPU in this case. If I were to do it, I think I would just put it after the CPU, replacing the current curvy stretch going into the GPU and then the rad, without changing the rest. But having seen the few Iceman Cerberus builds out there I'm definitely not willing to make my life impossible with a side rad, so GPU is probably not happening.

Questions @Mussels @Valantar @HammerON :

• With new parts, do you guys usually individually rinse the block/tubing/reservoir with distilled water before you assemble?
• Alternatively, would you rather fill the loop first with distilled water, leak test it, then drain and fill with Cryofuel?
• Recommended procedure for pre-cleaning rads? I watched a couple of videos and tentatively decided on 4 x flush with warm diluted vinegar water, then 4 x flush with distilled water

In the future, that res will definitely be an internal res. Having to route it all the way to the back yet having to put the rad on the floor is not looking very optimized.

 

[Mussels]

built it outside the PC first, so you can learn how the fittings work, how the pump behaves and what sorta shit goes on as you cause a dozen leaks


I had lots of gunk fluff and oddities show up in my hardware, some of it because the coolant and/or dyes i'd used were the problem

So yeah... measure twice, cut once, test your coolants and shiz cause cleanups a PITA

 

[Valantar]

Parts are all arriving now in droves, just waiting on the last EK order. I think this is the layout I'll go with.

No QDCs for the time being as I just don't see myself adding a GPU in this case. If I were to do it, I think I would just put it after the CPU, replacing the current curvy stretch going into the GPU and then the rad, without changing the rest. But having seen the few Iceman Cerberus builds out there I'm definitely not willing to make my life impossible with a side rad, so GPU is probably not happening.

Questions @Mussels @Valantar @HammerON :

• With new parts, do you guys usually individually rinse the block/tubing/reservoir with distilled water before you assemble?
• Alternatively, would you rather fill the loop first with distilled water, leak test it, then drain and fill with Cryofuel?
• Recommended procedure for pre-cleaning rads? I watched a couple of videos and tentatively decided on 4 x flush with warm diluted vinegar water, then 4 x flush with distilled water

In the future, that res will definitely be an internal res. Having to route it all the way to the back yet having to put the rad on the floor is not looking very optimized.
View attachment 226644

The first rad I bought I never flushed - but that was in ... 2008? and I knew exactly nothing about anything. Can't remember having problems with it, so I guess I was lucky? My second rad, in 2017-ish, an EK PE, I did a simple flush with warm dilute vinegar, think I ran through it a couple of times. Did the same for the Corsair XR5 I got earlier this year. Neither had any notable amount of gunk in them, so either I got lucky or they were well flushed at the factory. It's probably a good idea to be a tad more careful than I am, as I've seen some really bad examples. Due to the relative difficulty of getting ahold of large quantities of distilled water where I live I have never, ever used that for flushing. I know the tap water where I used to live in Norway was great, but here in Skåne, Sweden it's supposed to have a ton of lime - but I've had zero problems. I refuse to believe that whatever minerals are left over in the few ml of liquid I can't get out of the loop matter whatsoever, and any algae, bacteria, or other problematic stuff are easily neutralized by the coolant. I wouldn't run a loop with tap water (perhaps for testing, not for long though), but I have zero issues using it for flushing and the like. Then again, there are no doubt places in the world with far worse tap water than where I live. Still, considering how little water will be left after a flush, I can't really imagine it actually being a problem. Those repeated flushes with distilled water sound to me like either outdated wisdom from the times when radiators were assembled in a shed and barely held together, or just plain superstition. We're at least two decades into PC water cooling, and these components are made by large companies. They should arrive reasonably clean. I would do a couple of vinegar flushes, and then run water through until you can't smell vinegar any more.

A sealed, factory-new block from any decent brand I would just install straight off the bat. If EK or someone similar sent me a dirty block, I would be pretty pissed, as that IMO is not an acceptable level of QC. The Aquanaut needed to be built, so I knew that was clean, but I never flushed my 6900 XT Liquid Devil. Radiators with their complicated solder joints and the like are ... acceptable if there's some flux and minor gunk left over. They might also have some sort of anti-corrosive added for transport and storage. Milled copper, brass, acrylic, POM, etc? Should be entirely squeaky clean, period. There is zero excuse for the factory to leave those dirty.

For the record, I've never flushed anything new in an assembled loop, as that just seems wildly unnecessary to me. Flushing individual components makes things drastically easier, and saves you from filling and draining the loop multiple times.


Not very applicable to you right now, but I've also used warm dilute vinegar to remove corrosion from that 2008 rad, which was out of commission for something like 8 years (I had a leak and got scared) and was stored in some random box with ill-fitting stop plugs, but was used in my loop with the 1600X and Fury X from 2017 till I got the Aquanaut last year. Vinegar truly does wonders in removing corrosion. It won't make things look good, as it doesn't remove the tarnish itself (you likely need abrasion for that), but it removes the insulating layers of corrosion and anything likely to flake off.

That loop layout looks good to me. Simple, clean, and reasonably easy to plumb. Compression fittings are a good choice considering that short (and potentially sharply bent) res-CPU run. I would probably think a 45 is good on the CPU-rad run, as that needs to bend outwards anyway to cross the other tube and reach the outer radiator port.

I think Optimumtech makes a horseshoe shaped loop up and back down into a 90° on the CPU block actually. Not sure how I didn't notice this. He uses the X570 Impact as well so the socket placement is the same. Dare I say, the 180° bend has a bit more character than the other options. Cerberus is similar but a little bit bigger than M1 in every dimension that matters, so I think it might be doable.

This is an example of something that I would only ever attempt with compression fittings. At least with my 10/16 EK ZMT, that tube would be exerting some significant force on the CPU-side fitting, and would likely wrench itself off any barb fitting in existence. I guess a barb with screw-tightened tube clamps would work as well, but those are hardly SFF-friendly. Seems you're going for compression fittings though, so that saves you that hassle. Just remember to put the outer ring onto the tubing before pushing it onto the fittings


Oh, did you order that leak tester on your parts list btw?


What I would do, step by step, if I was building this:
- Flush the radiator. Get it as dry as is reasonable (and likely plug the ports with paper towels to avoid drips when moving stuff around).
- Flush the tubing if it seems dirty. Likely not.
- Install parts, cut tubing to length, assemble the loop, being reasonably careful to ensure that all fittings are screwed down properly and all tubing is secure.
- Leak test with air.
- Fill and bleed the loop, using an external power source for the pump.

If you don't have a leak tester, I would be a lot more careful about installing everything, and then fill the loop slowly while looking out for leaks, then being even more on the lookout for this as the pump is powered on and the loops is pressurized. Leak testers make this part of the process so much simpler.

 

[tabascosauz]

SUPER pissed off at EK - they pulled the FEK5B 5% off code the day before I ordered (after it being valid for like 2 years, for a week prior while I was putting together my order, it had been working every day).

By then the "Black Friday" sale was already going (promising BIG DISCOUNTS), but was really underwhelming because discounts were only on some crappy bundles. So I reached out to them via email and asked; they said it isn't coming back, and I need >$350USD to get a discount. The rep didn't say ANYTHING about the ongoing sale in our correspondence earlier this week - all of a sudden, without warning today the "Black Friday" sale evolved, and now everything on the site is on sale. IT'S STILL THE SAME SALE!!!!! THE BANNER AND TIMER HAVE NOT CHANGED!!!! NO WARNING WHATSOEVER!!!!

Not to mention when it came time to order, the shipping costs suddenly went up and have stayed up since, like BRUH

Not to mention that they don't have stock counters unlike any of the other watercooling stores, so trying to find the DDC 4.2 in stock is like playing a slot machine. When I ordered it said it was backordered until the 26th.....................then subsequently shipped out immediately despite being "backordered" LOL

Yeah, I get it, you want to make a quick buck from me, but are you kidding me? The only consolation is that EK is still being stingy af and reducing their prices by like $1 on most of the items I ordered (Cryofuel down to $16 from $17.99 LOL). In the end it was only about $15 difference.

PPCS might have sent me the wrong rad, but they sent me the right one at no cost arriving within 3 days, and let me just keep the incorrect rad. And I hold them in high esteem, because of that. Maybe I should have just ponied up for some Bitspower fittings from PPCS instead, and stayed away from EKrap store entirely.

For the record, I've never flushed anything new in an assembled loop, as that just seems wildly unnecessary to me. Flushing individual components makes things drastically easier, and saves you from filling and draining the loop multiple times.

It's just the block really, I had a 90 degree on once and off once before I decided I wanted regular fittings only on the block, and there's one or two tiny bits of I'm guessing acrylic shavings in there. I'm probably just OCD, I mean they are tiny, size of a speck of dust. Figured maybe I could run a bit of distilled water through there once I had the two tubes on.

Water here is pretty soft (~10mg/L), but I can get 4L distilled water for $2 so it won't be a problem to pick up a few jugs of it.

I will be taking my time flushing the rad obviously, but not sure how early I want to start. The last few parts are on their way over the Atlantic now but will still be a couple of days. Is it okay to just leave the radiator out to dry after vinegar- and water-flushing it? Or just do it day of, and go right into building?

This is an example of something that I would only ever attempt with compression fittings. At least with my 10/16 EK ZMT, that tube would be exerting some significant force on the CPU-side fitting, and would likely wrench itself off any barb fitting in existence. I guess a barb with screw-tightened tube clamps would work as well, but those are hardly SFF-friendly. Seems you're going for compression fittings though, so that saves you that hassle. Just remember to put the outer ring onto the tubing before pushing it onto the fittings

Yeah, thought the U-bend was a good idea, until I actually got my hands on the block and tubing and realized it's way tighter on space than I thought it would be. Probably extra flow restriction too from 2 x 90 degrees and the fact that the tubing bends all the way back. Still, even doing the 45 degree out and regular in routing may be a bit challenging

Literally all I needed was a way to mount a 80FLT/DDC combo onto the back of the SF750, but as hard as I tried I just couldn't find anything of the sort, short of resorting to velcro (!!!!). Would have made for an easier time, undoubtedly. At least the Iceman has good build quality, got it today.

The ball valve is a REALLY tight fit on the Iceman, because unfortunately the threads end with the valve handle pointing towards the case. Just my luck, but it still fits. If I have any 45s left over after building the loop, i might put it on to angle the ball valve outwards.

What I would do, step by step, if I was building this:
- Flush the radiator. Get it as dry as is reasonable (and likely plug the ports with paper towels to avoid drips when moving stuff around).
- Flush the tubing if it seems dirty. Likely not.
- Install parts, cut tubing to length, assemble the loop, being reasonably careful to ensure that all fittings are screwed down properly and all tubing is secure.
- Leak test with air.
- Fill and bleed the loop, using an external power source for the pump.

Yep I did buy the EK leak tester. So I will be a connecting it to the res fill port and pumping it up to 0.25-0.3bar for 15 min or so, before I consider filling the loop.

 

[Mussels]

Wow, that's a rough experience


I mean sure, sales people likely wont wont know about upcoming deals - that's higher ups.
Stock counters, yes, they're poopy. I had weeks of waiting, and added delays.

Buuuuuuuut I've had three bottles of coolant and a new piece for my 3090 block sent out completely at their cost, when i've had issues with their products.