Hi guys.
First a short summary of how I got myself into the mess I find myself in
Feel free to skip down if all you want is to read about my dilemma. I thought this information might be useful for others who can get their hands on a pulse easier than a nano.
I'm doing a DIY case build of a dual-boot hackintosh somewhat to S4 size specs. One of the greatest hurdles I've met is that the Vega 56 Powercolor Nano – which would fit the size requirement of my case – costs about 800$ (EDIT: I live in Europe and has to import the Nano from Newegg). The Vega 56 Sapphire Pulse costs 500$, and has (almost) the same PCB (it takes 2x8 PCI-e for power rather than 1x8 + 1x6).
I thought the 300$ I saved on the card could get me a good ways to better cooling than stock Nano has anyways, so I ordered the Pulse. Then, I went ahead and looked at cooling solutions. My power situation is perhaps somewhat precarious (HDPlex 400 DC-ATX, waiting for the 400 AC-DC and currently on a 220W brick with a 65W CPU). With undervolting and overclocking some people have said that the Vega 56 cards can perform like a GTX 1080 and I also love me a silent computer, so the cooling headroom I have been looking into is 200W-250W, although the card is rated at 180W.
First I looked into water cooling, but it's very difficult to get a pump, reservoir and radiator+fan combo under the S4 size specs. There are two water cooling blocks to choose from specific to the Vega Nano PCB, one from Bykski and one from Alphacool. Since AIO systems puts the pump in the CPU block, these are not useful for a build like this, and the two options I have seriously considered are the Airplex Modularity System 140 mm integrated pump from Aquacomputer and the Ion from XSPC. Both of these solutions would necessitate a size increase of the case which I'm reluctant to accept.
In aftermarket air cooling, the most useful GPU coolers seem to be ARCTIC aftermarket air coolers. From among the ARCTIC coolers the Accelero Twin Turbo II seem best because there would be no space for the backplate of later iterations anyway, and very importantly, because the heat pipes traverses the breadth of the card, which would allow me to cut it to length by chipping off the heat fins at the ends without touching the heat pipes. The ARCTIC coolers runs into the same business as necessitates a specific GPU block for water cooling, namely that it does not cool the whole card. This seems important on the tiny Nano PCB, where VRM and load balancers heats up quite a bit. Also, they haven't been updated in years and there is no official support for AMD GPU's. Then again, with small radiator parts fixed to important parts of the card, the ARCTIC cooling solution may work. Does anyone have experience cooling hot cards using a GPU aftermarket cooler and small radiators?
MY CURRENT DILEMMA
So all this research comes full circle. There are no great solutions. The Pulse stock air cooler won't work because it is too long. 27-28 cm, and it can probably be 18-19 cm, which means almost a third of the radiator length would stick out. I can still return the card and get the Nano. Otherwise, there's the option of modding the stock cooler.
I see two options. Either I remove the fins over the left third of the card and bend the heatpipes over a 40mm diameter. I am afraid this might damage the wick inside if a powder-type wick is used, and capillary action over the bend might be compromised. Furthermore, as the bend radius is 20mm, I need to remove at least 20mm of useful fins this far into the part of the radiator (or do some very difficult counterbending of the fins to retain airflow. In this scenario, I would then likely bend the heatpipes 'under' the radiator (to the top side on the above images).
THE BONKERS IDEA
The other option is to cut the heat pipes when I cut the radiator.
I would then 'simply' reseal them using compression fittings.
Doing so would be a somewhat involved process – ensuring semi-perfect water content in the heat pipes, bringing the water to a boil, then resealing when the water vapour has displaced all air in the heatpipe. The ambitious plan is to calculate ideal water content, add more, seal it up, test it, remove a bit, test again, remove a bit – and so on, until performance begins to deteriorate, then add water back. This might require more compression fittings/olives than I care to use.
I would use some non-hardening grease on the olives (middle part on the picture above) of the compression fittings to prevent gas transport. When the heat pipe cools, the water inside condenses, creating a near-vacuum. This lowers the boiling temperature of the water to be in the working temperature range of the GPU.
Initially this didn't seem like a good idea, but these fittings are used for natural gas, water, oil, etc., and if they can work for industry, they should be useful for a small home DIY project. Even if the seal fails and needs a few retries, the layout of my case is such that the leak would fall directly out of the case and onto the table, making diagnosing pretty easy.
I am currently pretty confident this would work the best and be the easiest solution, but I haven't been able to find any examples of others cutting open their heat pipes, which makes me reluctant. The only concern I have been able to pinpoint is if the working fluid of the heat pipes isn't water and dries out on the wick when I open it, but I have a hard time imagining a working fluid not wanting to boil off with a little help. The only serious contender for working fluid seems to be acetone, which I could probably also use if I wanted to.
If it works it is a pretty easy DIY fix to make any radiator conform to any smaller size. I think I'm 20$ out for the fittings and will probably be about 20$ out for the grease, making this a 260$ savings with improved cooling, IF IT WORKS.
What do you think? Is this completely bonkers?
(I'm new here, I hope I'm doing things as I'm supposed to. If not, feel free to point out errors etc.)
First a short summary of how I got myself into the mess I find myself in
Feel free to skip down if all you want is to read about my dilemma. I thought this information might be useful for others who can get their hands on a pulse easier than a nano.I'm doing a DIY case build of a dual-boot hackintosh somewhat to S4 size specs. One of the greatest hurdles I've met is that the Vega 56 Powercolor Nano – which would fit the size requirement of my case – costs about 800$ (EDIT: I live in Europe and has to import the Nano from Newegg). The Vega 56 Sapphire Pulse costs 500$, and has (almost) the same PCB (it takes 2x8 PCI-e for power rather than 1x8 + 1x6).
I thought the 300$ I saved on the card could get me a good ways to better cooling than stock Nano has anyways, so I ordered the Pulse. Then, I went ahead and looked at cooling solutions. My power situation is perhaps somewhat precarious (HDPlex 400 DC-ATX, waiting for the 400 AC-DC and currently on a 220W brick with a 65W CPU). With undervolting and overclocking some people have said that the Vega 56 cards can perform like a GTX 1080 and I also love me a silent computer, so the cooling headroom I have been looking into is 200W-250W, although the card is rated at 180W.
First I looked into water cooling, but it's very difficult to get a pump, reservoir and radiator+fan combo under the S4 size specs. There are two water cooling blocks to choose from specific to the Vega Nano PCB, one from Bykski and one from Alphacool. Since AIO systems puts the pump in the CPU block, these are not useful for a build like this, and the two options I have seriously considered are the Airplex Modularity System 140 mm integrated pump from Aquacomputer and the Ion from XSPC. Both of these solutions would necessitate a size increase of the case which I'm reluctant to accept.
In aftermarket air cooling, the most useful GPU coolers seem to be ARCTIC aftermarket air coolers. From among the ARCTIC coolers the Accelero Twin Turbo II seem best because there would be no space for the backplate of later iterations anyway, and very importantly, because the heat pipes traverses the breadth of the card, which would allow me to cut it to length by chipping off the heat fins at the ends without touching the heat pipes. The ARCTIC coolers runs into the same business as necessitates a specific GPU block for water cooling, namely that it does not cool the whole card. This seems important on the tiny Nano PCB, where VRM and load balancers heats up quite a bit. Also, they haven't been updated in years and there is no official support for AMD GPU's. Then again, with small radiator parts fixed to important parts of the card, the ARCTIC cooling solution may work. Does anyone have experience cooling hot cards using a GPU aftermarket cooler and small radiators?
MY CURRENT DILEMMA
So all this research comes full circle. There are no great solutions. The Pulse stock air cooler won't work because it is too long. 27-28 cm, and it can probably be 18-19 cm, which means almost a third of the radiator length would stick out. I can still return the card and get the Nano. Otherwise, there's the option of modding the stock cooler.
I see two options. Either I remove the fins over the left third of the card and bend the heatpipes over a 40mm diameter. I am afraid this might damage the wick inside if a powder-type wick is used, and capillary action over the bend might be compromised. Furthermore, as the bend radius is 20mm, I need to remove at least 20mm of useful fins this far into the part of the radiator (or do some very difficult counterbending of the fins to retain airflow. In this scenario, I would then likely bend the heatpipes 'under' the radiator (to the top side on the above images).
THE BONKERS IDEA
The other option is to cut the heat pipes when I cut the radiator.
I would then 'simply' reseal them using compression fittings.
Doing so would be a somewhat involved process – ensuring semi-perfect water content in the heat pipes, bringing the water to a boil, then resealing when the water vapour has displaced all air in the heatpipe. The ambitious plan is to calculate ideal water content, add more, seal it up, test it, remove a bit, test again, remove a bit – and so on, until performance begins to deteriorate, then add water back. This might require more compression fittings/olives than I care to use.
I would use some non-hardening grease on the olives (middle part on the picture above) of the compression fittings to prevent gas transport. When the heat pipe cools, the water inside condenses, creating a near-vacuum. This lowers the boiling temperature of the water to be in the working temperature range of the GPU.
Initially this didn't seem like a good idea, but these fittings are used for natural gas, water, oil, etc., and if they can work for industry, they should be useful for a small home DIY project. Even if the seal fails and needs a few retries, the layout of my case is such that the leak would fall directly out of the case and onto the table, making diagnosing pretty easy.
I am currently pretty confident this would work the best and be the easiest solution, but I haven't been able to find any examples of others cutting open their heat pipes, which makes me reluctant. The only concern I have been able to pinpoint is if the working fluid of the heat pipes isn't water and dries out on the wick when I open it, but I have a hard time imagining a working fluid not wanting to boil off with a little help. The only serious contender for working fluid seems to be acetone, which I could probably also use if I wanted to.
If it works it is a pretty easy DIY fix to make any radiator conform to any smaller size. I think I'm 20$ out for the fittings and will probably be about 20$ out for the grease, making this a 260$ savings with improved cooling, IF IT WORKS.
What do you think? Is this completely bonkers?
(I'm new here, I hope I'm doing things as I'm supposed to. If not, feel free to point out errors etc.)
Last edited:
