PARVUM FURIA

[JR23]

I think there is time for a small update before sleepy time, not everything just yet but a good chunk of progress to keep you all entertained, hopefully!

Straight in with the cutting at Parvum HQ, unusually we didn't actually use our swarf sucker for any of FURIA's parts. The reason for that was the tiny 1.5mm single flute tool used for the o-rings could barely reach out of the foot which surrounds the action. But on the positive side we got to see far more action.

So this is the first of 5 main layers which make up FURIA's monoblock, cut from 10mm frosted perspex this has all of the features from Bitspowers original CPU and motherboard block top.

And this is layer 2! Joining the coldplates together and sending the coolant off to where it needs to be for the next layer. It's sealed with just two 1.6mm o-rings and fills out all of the free space on the motherboard.

Screws, lots of screws, actually not 6-32 UNC this time but the rather more conventional M3 and M4 variety as used by Bitspower. Just nice simple oxide black steel screws with plain socket countersunk heads.

I actually changed the CPU block top design somewhat to allow for better packaging in the block. Coolant comes in through the centre, passes through the original jetplate and fins then leaves through both sides simultaneously not dissimilar to the original however rather than having the two sides then join asymmetrically they both flow out into the block with identical bends.

Time for layer 1 and layer 2 to be mated together. Actually I missed out a boring half day of work where I made all of the o-rings and tapped all the holes right here, but you've seen that before, right MATE?

At this point lots of the screws are missing because they thread into a coldplate, simultaneously applying pressure to multiple seals.

These are the original Bitspower parts going into the reworked block, the board block is M3 and the CPU M4.

First the jet plate presses into it's locating features, then the coldplate is placed over the top with original o-ring in place.

Much the same story for the motherboard block only without a jet plate.

Then flipped and lots more screws added, some of which were cut to length so they could utilize the most thread possible in the copper.

So this is the side which the motherboard sees.

And these are all of the components which went into the first major stage of the monoblock.

Once that was complete a 3rd layer was added, this is actually the only 15mm thick part in the build and is used to mate the two main parts together and leave a big enough gap for the GPU to fit in among it all.

5 M4x30mm screws attach layer 3 securely to layers 1 and 2. 4 further M4 threaded holes then attach the entire assembly to layers 4 and 5.

That pretty much completes this one tiny sub-assembly!

Just for some scale you can see it here mounted to the motherboard. The 4 pronounced screws hold it directly to the ILM socket, applying pressure to the CPU while smaller screws are inserted from the back of the board to mate it securely with the VRM's and chipset.

So there we go, phase 1, or 2 or whichever complete and presented!

JR

 

[Hukkel]

Art...it is art.

 

[OJ46]

Daaaaaamn that's some fine work, looks absolutely stunning! One quick question though: When you make custom blocks like this, what are the temps like compared to the standard components? Part of my head says better, part says worse and part is saying identical, so guess at least part of me is right

Awesome work though, can't wait to see the finished article in all its perfectly machined beauty

 

[Cenedd]

Stunning. Also, I think we now know why Parvum have been so busy!

 

[davido_labido]

I had a sneak preview of this and I can honestly say the whole thing is pretty damn awesome <3. Great work as per usual Joe.

 

[mods1965]

Amazing. You have now started providing specialist cooling as well as Reservoirs.
You are really taking PC design in an interesting direction.
Is this all going towards your final year project for your engineering degree?

 

[JR23]

Thanks guys, this is just one tiny piece of the monoblock, more coming soon.

Daaaaaamn that's some fine work, looks absolutely stunning! One quick question though: When you make custom blocks like this, what are the temps like compared to the standard components? Part of my head says better, part says worse and part is saying identical, so guess at least part of me is right

Awesome work though, can't wait to see the finished article in all its perfectly machined beauty

It's quite hard to say exactly how it compares as we don't build a loop just to test things beforehand, in theory the performance will be identical as for the majority of parts the features remain the same, they are just all combined into one part.

One huge difference though is the flow rates and restriction of the loop as a whole, because there are no unnecessary bends and any changes in cross section are gradual. Plus the actual loop is really really short with barely any difference in height so a normal DDC gets a huge amount of flow. That actually can make it very difficult to bleed because the air loves to just fly around with the coolant so we really haven't given much attention to how well optimized the pump volute is.

In order to keep the complexity and thickness down in this build the two sides of the pump aren't actually sealed from each other in this, still the pump is way too fast! I may try to get hold of a low RPM DDC from a Predator or something similar to calm it down, but actually it's running great! After a full full day of benchmarking the GPU temps peaked at 32°C with the two fans at 900RPM.

JR

 

[Cenedd]

You could always 7V it. I know that DDCs are supposed to have a starting voltage of 9V but I ran an 18W DDC off a 7V molex mod for quite some time and it never failed to start or run. Also quieter as a bonus. Currently run the same pump 5+ years later off an Aquaero at 8V and it's also fine.

 

[TechMinerUK]

This looks amazing, I just can't believe the sheer amount of design/engineering that goes into this with the custom waterblocks as it looks fantastic

It certainly makes the motherboard beefier when fitted though

 

[OJ46]

That makes sense, cheers JR! So you need to run the pump fast to push air out of the radiators, but slow to stop it flying straight through the reservoir... Can see why that would cause issues Still, when it looks that awesome an extra hour or 2 getting the air out is more than worth it

That's seriously cool, I'm running a full fat 360 + X-Flow 240 with fans (water temp controlled) peaking at 1200RPM and the GPU is usually around the 40°C mark :/

 

[JR23]

You could always 7V it. I know that DDCs are supposed to have a starting voltage of 9V but I ran an 18W DDC off a 7V molex mod for quite some time and it never failed to start or run. Also quieter as a bonus. Currently run the same pump 5+ years later off an Aquaero at 8V and it's also fine.

I have tried a 10W DDC on 7v before but it just didn't seem interested in starting at all, that was when we were trying to find a quieter solution for MATE. In the end a PWM variant resolved all issues for that and with that ASRock board it ran great off the CPU header slowing right down to 2000RPM with 40% PWM. Also I had the same issue with Chocolate Box which again is now running an EK 3.25 PWM. So I thought I would repeat that however the PWM on this motherboard is relatively poor and just doesn't get on with the DDC at all, anything below 77% and it just starts turning off and on repetitively.

I want to try and find the open vein design low wattage DDC from certain EK products, my friend had one with an MX-res that was astonishingly quiet. When I have the desire to rebuild this a little then i'll hunt for one of those and wire it to the PSU at 12v.

That makes sense, cheers JR! So you need to run the pump fast to push air out of the radiators, but slow to stop it flying straight through the reservoir... Can see why that would cause issues Still, when it looks that awesome an extra hour or 2 getting the air out is more than worth it

That's seriously cool, I'm running a full fat 360 + X-Flow 240 with fans (water temp controlled) peaking at 1200RPM and the GPU is usually around the 40°C mark :/

I've been working with 8 Pack doing a Fury X build in a VEER1.0 over the last couple of days and that's actually running in the 50's under sustained load, but with 500RPM fans. We've come to the conclusion that actually because the Fury X in FURIA is surrounded by waterblocks on all sides it really creates a good thermal environment. And actually it just seems like a good sample of the card that overclocks well.

JR

 

[8 Pack]

This build is truly immense. Crazy temps for the size.

Had fun the last few days testing and building with JR this rig and like JR says a Fury X build for AMD VR world conference.

 

[JR23]

I'm back with part 2 of the super update! This time I won't be showing absolutely everything as it would really detract from the final post, essentially the entire build is one big waterblock hence this will be largely details.

In the last update we looked through layers 1, 2 and 3 hence we are picking it up on the much much bigger layer 4! Here you can see what is essentially the GPU block top, power delivery portion on the left and core to the right.

Just like MATE in FURIA a DDC is mounted directly to the waterblock only this time rather than a 15mm block cut from both sides with a 5mm lid there is a stack of two 10mm blocks each machined from only one side. This is the lower portion of the pump top containing just the volute and holes through to the final layer.

The memory waterblock also mounts directly to the lower face of layer 4, exactly the same height as the GPU which gives some idea of just how small this build is and why the motherboard section had to be so thin.

Layer 5 which forms the 'front' face of the build joins everything else together distributing and linking each little channel, the reservoir and fill port are also in this level.

Because the GPU top originally had a copper plate covering the bridge between GPU core and power delivery elements of the block the two identical features original in the top of the top could easily be moved up to this level.

As mentioned earlier the pump top was split over two 10mm layers, this is the upper part which feeds and exhausts from the pump volute. The two sides of the pump are not positively sealed from each, not only due to size but also because the part over the pump impeller the o-ring would mate against is so thin I was concerned that it would leak. So instead of making an entire plate thicker and increasing the size of the build by 5mm mass was left behind to distribute the flow.

A small finger of layer 4 extends over the radiator ports however to keep the build as small as possible the in and out channels to the radiator had flow directly over each other.

Both layer 4 and 5 have a channel cut along to one of the ports and an intermediate layer seals the exposed channel forming two passages to the radiator ports within 25mm thickness.

Layer 3.5, 4 and 5 were all screwed together and the GPU coldplates added.

That just leaves the memory block which actually presented some substantial complications. In order to assemble the two main pieces the cold-plate had to be attached the RAM even though it seals against layer 4. Not to complex in itself however it meant the screws had to pass all the way through layer 5, 4 and into the coldplate, 25mm. Finding M2.5x25mm countersunk socket head black screws is not an easy job. Hence the coldplate had to be tapped out to M3.

A few days before I planned to assemble the build I decided tapping to M3 was the best way forwards, drilled the old threads out with a 2.5mm (the tapping diameter for M3) and then immediately sheared a tap in the first hole. Getting that out was impossible hence I had another block sent from a niche european retailer who can be bothered to stock awkward parts like memory water blocks from Bitspower (props for being rapid). This time things couldn't go wrong so I found an extremely gradual taper tapped and employed lots of cutting compound.

Thankfully it went awesome and the build could continue! The tap survives for another day!

Time for some coolant, I had no idea how it would appear in the block so just to get it started I added some Mayhems blue dye to X1 clear for the leak testing.

After spotting a small leak early in the filling process I managed to pinpoint a small bur one one of the tapped holes between layer 4 and 5 holding them apart. Quickly drained and re-assembled things went very smoothly considering the immense complexity and shear quantity of seals.

Yeah, i'm not about to make it obvious how it works or appears as a whole

Right then, with the one watercooling part and leak testing out of the way it's time for some sleeving. No less than MDPC-X b-magic, grey mkII, shade 19 and black!

Because things were so tight surrounding the GPU there was no room to have the cables sleeved where they leave the dual 8-pin PCIe's. So I choose to use all black insulated 16AWG wire and join them as tight to the GPU as possible where they are hidden deep inside some other stuff.

Now with everything complete there really isn't much to show without showing the whole build. So here is the big piece. Most of the loop, graphics card, reservoir, pump, fans, rad mounts, PSU mount and to the right it's clear where the smaller piece mates via the memory block and layer 3.

And this is the baby piece, the motherboard hangs down directly over the GPU and each of the tiny cables flow round into the PSU.

I suppose next time we will be seeing the full build! There are some little details to complete, figuring out how it stand and a neater riser cable solution but the build is running and performing magnificently, decimating builds several times it's size!

Firestrike - i7 5820K @4.4 Fury X @1200/540

JR

 

[Cenedd]

Absolutely incredible. Would be interested in clarification of the DDC cut outs across layers as I'm planning something similar (much less grand and impressive but containing two DDCs) for Parvum to cut. I can see the classic whirl where the impeller sits, just not seeing the part on the next layer where you say it deliberately doesn't fully seal.
My plan is essentially to integrate something similar to XSPC's dual DDC top into the end of my block but have the input and output where I want it. My wife should probably lay the blame at your feet for inspiring this course of action...but I won't tell her if you won't!

 

[Hukkel]

It is just too much. Just too much...

 

[OJ46]

Well that's certainly made me look like an incompetent ape, cheers for that JR

Haha fair play though, that is some absolutely top level craftsmanship on the blocks / reservoir / beastly lump of nerd porn, always astounds me how you can cram a full loop into such a minuscule space. Just the planning stage and figuring out how to run the loop through that many separate layers would have caused me a major brain implosion, without the ball ache of actually trying to put it together with that little room for manoeuvre! Thought I'd done well fitting a loop into an ATX case

 

[TechMinerUK]

I keep saying I can't believe how much effort goes into these build but it is clearly worth it as teh waterblocks look smashing with the blue coolant running through them, also you have done a great job at sleeving the cables (I'm surpised at the tight bend radius to get them to the GPU being effected by the sleeving).

Not that I don't like watching the build log but I'm glad the wait is almost over to see the full system as I can't wait in anticipation much longer

 

[Diagro]

wow...

 

[8 Pack]

Great Firestrike mate for water cooled Fury X. Very nice.

 

[JR23]

Absolutely incredible. Would be interested in clarification of the DDC cut outs across layers as I'm planning something similar (much less grand and impressive but containing two DDCs) for Parvum to cut. I can see the classic whirl where the impeller sits, just not seeing the part on the next layer where you say it deliberately doesn't fully seal.
My plan is essentially to integrate something similar to XSPC's dual DDC top into the end of my block but have the input and output where I want it. My wife should probably lay the blame at your feet for inspiring this course of action...but I won't tell her if you won't!

Sure it's quite simple once you see it and figure how they sit together. So the semi-spiral volute is cut into the bottom side of layer 4, the inlet and outlet are cut all the way through...

...and they go straight through into the bottom of layer 5. The inlet is coming from the res here on the right and straight into the centre of the pump, flowing through the impeller/volute and out through the thin slither which aligns with the channel on the left.

The inlet and outlet sides of the pump aren't fully sealed from each other coolant could sneak through the tiny gap between layers 4 and 5. There will definitely be some efficiency/head pressure loss but that wasn't a huge concern. The reason why they couldn't be sealed from each other is the piece left in layer 4 is really thin and would bend under the pressure of the o-ring, so the o-rings go all the way around the whole thing.

Here you can see how the layers fit over each other a little more clearly. This view is looking down from above level 5 so the front is solid with channels in the back, then the next layer has the pump volute open on the back, the pump then screws on behind that with original o-ring.

Well that's certainly made me look like an incompetent ape, cheers for that JR Haha fair play though, that is some absolutely top level craftsmanship on the blocks / reservoir / beastly lump of nerd porn, always astounds me how you can cram a full loop into such a minuscule space. Just the planning stage and figuring out how to run the loop through that many separate layers would have caused me a major brain implosion, without the ball ache of actually trying to put it together with that little room for manoeuvre! Thought I'd done well fitting a loop into an ATX case

Yeah figuring out where the coolant will go is one weird headache, then how the build can actually be assembled is another. Placing each screw so everything seals, everything can be done in order and one doesn't pop out in the middle of some coolant on the layer underneath is something else. But after doing all that, the build just had to work out. There was only one mistake made and that was just leaving enough clearance for the GPU cables to come out, luckily I realized before final assembly, actually as soon as I saw the parts cut and managed a decent work around.

I keep saying I can't believe how much effort goes into these build but it is clearly worth it as teh waterblocks look smashing with the blue coolant running through them, also you have done a great job at sleeving the cables (I'm surpised at the tight bend radius to get them to the GPU being effected by the sleeving).

Not that I don't like watching the build log but I'm glad the wait is almost over to see the full system as I can't wait in anticipation much longer

It wasn't entirely down to how tight they had to bend but just the shear added cross section of sleeving for the 16 wires on the end. When you look where they come out they have to immediately bend straight around avoid the memory block which mounts to the open hole above the GPU. Then they have to fit inbetween the pump and the GPU, quite a tight gap. I made them join together right at that point so everything you can see without really trying is sleeved. The splits are hidden under more cables too so it was quite a neat way out for the PCI power.

Great Firestrike mate for water cooled Fury X. Very nice.

I managed to find another 500 points since then, the CPU had some more to give and it really helped with the combined result. Just could use a mental 10 core now to smash away the physics result! Actually perhaps the HBM could find a more comfortable clock, it would go higher but it didn't seem to be beneficial but maybe it hadn't made it to the next bin with 560.

Firestrike - i7 5820k @4.6 Fury X @1200/540

JR