It makes little logic or sense, but for some unknown reason I just want to fiddle with adding watercooling to a laptop. Lots of pics will be incoming at a later date. I have an old unused Dell Latitude D620 which has volunteered as the guinea pig. In usual Dell style the heatpipe/radiator/fan cooling is *just* adequate for the NB+SB+35W Core 2 Duo which heats up to about 80C at max, more like 90C when the hsf fins are gunked with fluff. I really like the setup that this guy did on a similar era laptop with low power dGPU: maybe 50W total heat production which was marginal on stock cooling but greatly improved with the added custom copper loop complete with screw disconnect to an external rad/reservoir/pump. The copper tube used looks like it's 3mm or 4mm (1/8 OD) Which is also the same basic setup used in the Asus GX700 which is factory hybrid watercooled. But spending a $5k PREMIUM over already premium priced similar spec hardware for a non-user-serviceable closed loop dock to gain a 30-40% overclock makes my idea look positively sane in comparison So yes, the aim of the project sounds like a huge catch-22, to improve (thermal) longevity by piping water through a laptop and to improve docked performance while not impacting portability which is the ONLY possible metric a laptop beats a desky in the first place. The purpose of the D620 is to trial it and see how it goes and get a gauge on the effect of flow, copper interface with heat pipes or fins, practice my soldering/brazing/bending/joining, planning purging/disconnecting, that kind of thing. The long term aim is to do it to a much larger, hotter, gaming laptop I got my hands on recently and am very much liking its 4K screen, and its BIOS-unlockable CPU and upgradeable GPU (hurry up mobile Polaris and Pascal I want to know whether to bother with either of you) So my todo/toresearch list at the moment: Get hands on a liquid cooling setup of any sort, to have a play with. get hands on some copper and start beating the shit out of it, cutting, bending, soldering, that kind of thing. Think about the balance between copper tube OD, thickness, malleability, interior space, fitting availability. Quick disconnect or screw fittings a must which 1/4 seem most available but 1/4 OD seems way too big for the laptop interior Dust off the D620, gut and clean, and plan a route Ingredients Copper tube, 6mm OD or smaller? 0.5mm wall thick enough? Soft or not? 1/8 OD? 1/4 OD? Will bendy copper be suitable for soldering/brazing will it harden, or crack? Connectors for side/base of laptop, and how to fix to copper tube Standard setup of rad/reservoir/pump/tubing for external portion and for testing, either AIO or bitzed from wherever Insulation for components where the loop will get too close
Tore the guinea pig laptop to bits. Impressed by how quick and easy it was to reduce it to pieces and a pile of screws. Click to view full size! Boy, it's cramped, there'll be no easy way to fit anything like 6mm tube in there anywhere, 2 or 3mm will be the maximum. Unfortunately it is a top-mount heatpipe arrangement which means any route following the heatpipe will be on top of the motherboard. Which is far from ideal. Click to view full size! People this is why you dust your fans - one big furball that originally covered the whole area between fan and radiator grill Click to view full size! The area next to the jack is an easy point for entry and perhaps exit as well, but right on top of the motherboard, will have to consider non-conductive coolant at least for testing Click to view full size! Click to view full size! A couple of potential routes. Depends on whether there is a route under the heatpipe or whether there could be space to sneak it in the space between fan and radiator in a way that minimises air flow loss since the radiator is only 10mm high (not including heatpipe on top) There is no need to have it route all the way around the end of the heatpipe, that's the integrated GMA950 graphics on the end of it which is only a few W worth, if I can route on top of the CPU that'll be enough I think The tradeoff will be: the smaller the diameter of the copper tube, the easier it will be to route, the less cutting and gouging to make space, but the lower the ID obviously flow is reduced, and the higher pressure means I can have no weak points after working with it I might order a few small bundles of copper pipe off aliexpress and mock them up. I think I can go as small as 3mm OD 0.5mm wall. The next issue is what can it connect to. I'm thinking 6mm air hose quick connects are about the smallest I'll find
So I've put together a bunch of bits, yes it's the cheapest of everything possible as this is the exploration phase. And to keep me crying about useless wrong parts to a minimum. Cheap weak 12V pump and small res $14 12V thermoprobe $1 Secondhand 360mm radiator (yes, waaaaay overkill) $45 10mm (~3/8) cheap vinyl tube I'll get from Bunnings, if I can't make the 8mm tube I already have fit. ($0, maybe $10) 3/8" barb to 1/8" BSP female. (<$5 for 5) 1/8" BSP quick disconnect with screw valve (a neat thing I found - used in tattoo/airbrushing size airtools). Male 1/8" with the slide mechanism and screw valve on the outside, female 1/8" will be mounted on the laptop case ($8 for 2) Male 1/8" BSP to 6mm compression - I also found elbow version as an alternative (<$5 for 5) 6mm OD soft copper tube ($12) Total so far under $100 Then the same again (6mm compression to 1/8", 1/8" screw/disconnect, to 3/8 barb. The screw valves should help draining/purging if required, as I want the laptop to remain portable not leaking all over itself as soon as it's put in a bag. Hopefully the quick disconnects are watertight. The only thing I'm not 100% on is the copper into compression fitting. May need oring or even soldering for a completely watertight join, which will be important since the inlet will be less than a cm from the DC jack. I think I've found where the 6mm tube will go. Depending on how much I can flatten it without kinking, I may need to grind a bit off the heatpipe housing (not the heatpipe itself), the fan housing and the radiator, and maybe the upper case (will ruin the space a BT module would have sat in but there isn't one there) it just depends on how much space the 1/8"-6mm fitting takes up (and whether I can shorten it). Waiting on china post to get it all to me now so I'll put the lappy back together and bench it in original form. Stage 1 will be putting the loop together. #2 will be plumbing it in, getting it functional. #3 will be improving bits as needed.
A few more bits procured during the agonising "20-45 day delivery" wait times, and straightened up a lot of bent fins on the very second hand rad. Some of which may have been my work. Flat (unribbed) square edged tweezers worked brilliantly $38 for a pair of 4 packs of what look like decent rad fans - Coolermaster SI2 - 1200rpm so fairly quiet at 19dba and okay static pressure - for $5 a pop fans - in a push/pull on a 120.3 there may be the option to dial down fan speed even less for near silence. They're not PWM but I wanted to keep it simple for now and I have the option of adding a voltage drop later on (a simple 2 speed $2 switch and resistor setup maybe) A few other bits from PCCG: thermal pads of varying thickness, more just in case but I needed a few anyway for another laptop. XSPC 4-way fan splitters, with outputs spaced at different lengths was a nice find Cheapest PSU I could find was $23. Shaw "EVO-685MAX"... their spec label BS has reached new levels. While the single 15A 12V rail was what I had running a GPU in 2003 - adding that 180W to the 12A 3.3V and 15A 5V it still only gets to a total of NOT EVEN 300 WATTS. Still I think it'll handle about 10W worth of fans... I may replace it with a 12v wall adapter off aliexpress down the track When I have it all I'll put up pics, plus a link to my aliexpress wish list which has most of the stuff on it.
Bloody hell. I have had everything else for 2 weeks now gathering dust except the cheap pump/res and the copper tubing. The pump/res was sent Economy so no tracking at all. The copper tube, Auspost tells me arrived in customs 2 weeks ago but nothing since. The 360mm rad with push/pull fans is set up and is complete overkill. Found a PSU switch in with some old Vidock stuff which was neat. Safer than paperclip. Thermoprobe hooked right in. Click to view full size! First problem: I figured out the quick disconnect is the wrong side for my plan. Click to view full size! The sealing part is under the spring loaded collar. The male part is just a tube. So the part of the loop inside the laptop would not be sealed. I can still use the screw valves for purging though so it should be OK. I saw the same quick disconnects without the screw valve which may still be useable inside the laptop but that'd be another 20-45 days away, and I'd need the opposite fittings as well since it'd be the other way around. I may also be able to save space by fitting the copper inside the end but it will take a lot of something (solder/JB weld?) to fill the ~1mm circular gap (6mm OD copper into 8.2mm ID fitting). But until I get stuff I'm just guessing! Overclocking, even overvolting the CPU in the 945GM chipset is a bust. The only possibility seems to be PLL pinmods but without ability to increase voltage it'd just be total experiment. I did however find GMABooster that increased the speed of the GMA950 iGPU from it's stock 200Mhz to its rated 400MHz of desktops it was in. The more heat it's producing, the better I can gauge its effectiveness. At least I have fine grained fan control, I can force the fan to fast (3700rpm), slow (2200rpm) and off through a util called i8kfangui. So other than being able to scale up the heat it's got what I need as a test bed.
So, now I FINALLY have a pump and res, things will actually get moving on this soon. I managed to destroy the original heatsink trying to fit the copper tube around it. Trying to do this in something so small and confined first up is clear to me I've bitten off more than I can chew, but I've got a good idea of what the tube can hack and how to bend/twist/flatten/shape it, how to dremel metal, and what not to do to heatpipes. I destroyed the heatsink by trying to blowtorch-solder the copper tube onto the channels I cut into the aluminium part of the heatsink, it melted the existing solder first - between the copper block and heatpipe, and the aluminium housing - and it all fell apart on me. I also punctured the heatpipe when the dremel slipped and made a weak spot which then led to a crack at that spot when it got hot with the blowtorch and I could hear all the gas inside fizz as it escaped. Oh well a new pair of heatsinks arrived - $6 (yep including post) from Lithuania so after I have done a couple other things (4K panel arrived today ) I'll have another go.
Finally got all the bits I need for a loop, tested ok. Have arctic silver epoxy adhesive too, heat transfer will be nowhere near solder but it'll be easier to start with! Have cut the path for the copper pipe, but it kind of looks scrappy so may redo it. I might try it simpler first: just do a short ~10cm pipe past the radiator and the end of the heatpipe, and then maybe try again at the expansive through-and-under setup I tried first. It will also be a good comparison of how much surface area is needed for effective heat transfer into the copper pipe.
Still haven't got onto this!!!! Have everything I need except time and space!!! I feel the bits staring at me every time I walk past! I lack a decent solid workbench in the garage - 1/4 of which is taken up with a stack of floating floorboards, which I planned to swap with a busted up old sturdy wooden desk or two from the MILs garage (hoarder) and put a solid vice on and reorganise the whole space, but I was only going to put those floorboards down in her back room once she got her butt to a lawyer and did a proper will... which hasn't happened even after a end of last year deadline expired TL;DR blameshifting...
I've been decluttering recently and got this back into gear. I have the loop components all set up and assembled and leak tested. Need to make like easier for myself by adding a drain port and also a way to purge the part of the loop sitting inside the laptop. But the 1/8" quick disconnects leak like anything. They must be real crap quality, no internal sealing I can see (just spring loaded metal on metal), if they can't hold water then no way they'd hold air (they are intended to be pneumatic, for small airbrush/tattoo equipment) Or maybe I was mistaken in assuming that because pressurised air finds leaks much easier that air couplers would be built to a better tolerance and spec than fluid couplers.
I've ordered some cheap 5mm o-rings from china and in the meantime have crudely cut my own and superglued them to the male end of the coupling. I will try them soon (not expecting anything magical) and will also remove the higher power MCP655 pump and try just the smaller alphacool DC-LT in case the water pressure from 10mm tube into 1mm opening internally in the small QDCs is exacerbating the sealing issue. Regardless the 10mm barbs attached to the cooper tube loop I made leaktested fine so I'll attach them with low temp solder (because I'm impatient) to a spare heatsink for a 7970M inside my spare P370EM. Trying to get the GPU to produce as much heat as possible led me to discover how to power play table edit the 7970M's stock 84W power limit - which has been done before but usually takes a vBIOS edit, and since that was from the legacy/CSM era years ago they don't work with UEFI (which this laptop is config'd for). Anything over 116W power limit leads to black screen crash though I think VRM cooling is insufficient - at 1050/1500/1.125V it maxes at 83C - ideally I'd like it up at 90C or so to be able to more accurately gauge the effect of the added heat removal capacity
Well the DIY o-rings were a bust but also validated the proof of concept as one of them was *almost* watertight... but I need 100% leakproof given the proximity of the fittings to important things like the power jack... for initial testing it'll all just be screwed in place. As for the loop I'm thinking a two-pump two-res setup with a drain line on the bottom res/pump and fill on the top res, since without QDCs I'll need to drain the loop before disconnecting. This will also allow fairly long tube lengths. The soldering will be interesting. I'm using a rosin flux paste, low-temp (~140c) solder paste and a heat gun which should stop me melting anything else (the problem I had with my first prototype, I desoldered everything and it all fell to bits), I managed to solder the fittings onto both ends of the copper tube pretty well with this setup despite approx a 0.5mm gap, and solder some 0.5mm copper shims onto the two pieces of radiator to create a one-piece heatsink this way. If I can bend and flatten the copper into having good contact it should be pretty solidly held together. I always have the thermal adhesive if things go pear shaped but would prefer solder for its better thermal conductivity
Teaser until I get back and do a full update 52C max on a 20 loop Firestrike stress test, loop max temp 35.5C... In 32C ambient (yep hot.) Without anything attached, this was pretty much max long term stress test stable clocks, reaching 84C (higher power draw would cause a black screen crash on stability testing). There is also a nearly 15W saving on wall draw just from cooler running.
This thread is overdue for an update but I have been busy modding heatsinks which is no fun at all. Since they all use low temp solder (heatpipes can damage at >180C) and since they are designed to spread heat around they have a nasty tendency to all fall apart all at once since that low temp solder is all that's holding them together! My master GPU soldering went well, the core heatplate shifted without me realising and was slightly misaligned (fixed with a 0.5mm shim), the slave works brilliantly for the GPU part of it, but I got too big for my boots and tried to solder the copper pipe across the woosy third thin CPU heatpipe and ended up causing myself all kinds of problems since I didn't have enough clips and clamps to hold all the bits together where they needed to be held together... Anyway when the 7970M was in the master slot I racked up world records for all the GPU benchmarks pushing it to 132W power limit, 1140mhz/1150mV core and 1500mhz memory. Temps were mid 50s maximum but the FETs would black screen crash at any more than 1.15V which was annoying. Still a fair bit over stock at 84W power limit, 850Mhz/1000mV core, 1200mhz memory. I'm still testing the 1070s but so far at 150-160W per GPU they hold at mid 60s in temps. I need to overclock them more; the curve editing capping them at 0.975V is constraining the cards. Unfortunately the CPU part of the heatsink has worse performance than before and struggles to maintain a 65W load without thermal throttle (whereas before it could handle 80W encoding long term in low 90s...) maybe it needs a shim but I think the heatpipes have become misaligned and are contacting the chassis or the solder's shifted where they contact the main radiator
And some benches: Firestrike 20537 with a GPU score of ~37500. Totally held back by CPU. These were on the same afterburner edited curve that maxes out at ~1911mhz, 0.975V, 2275MHz (9100qdr) memory. I had the power limits set at 180W per card; each was only using about 150W at those clocks and since the GPUs are mid 60s (compared with previously thermal throttling at 92C at stock 115W on the slave card) there is more headroom in it. These cards can go up to 1.013V; beyond that they get unstable (early MSI MXM 1070 bug)
End result is I can get both cards pulling ~170W sustained, before I start hitting the laptop's limits (if I grt too far over 500W wall draw it causes the CPU to throttle back to 1.2ghz) which is enough power to fully load the 1070s at 2000MHz @ 1.00V, with temps in the 70s. I've hurt the CPU cooling though (it used to be able to bench x45, now only x42) with all the desoldering dramas so will have a go at fixing that back up and maybe getting the copper tube closer to help cool the CPU since pulling heat off the heatpipe is good for only a few C.
