Antminer S5 + 3D printed mods

[notlist3d]

Im using a prusa I3. I will for sure give a part of all sale to the OP.
Will let you know when I have a full kit. How many time it take and how much materiel I taked.

If everyone that sales gives a little bit to OP this could lead into a really awesome forum thread.  If there is lot's of custom 3d printed parts for mining that make improvements. 

I see this as a possible start of something really big.   Would be amazing to go here point to a file we want printed for out miners and pay someone with ability to do it.

[Mikestang]

If everyone that sales gives a little bit to OP this could lead into a really awesome forum thread.  If there is lot's of custom 3d printed parts for mining that make improvements. 

I see this as a possible start of something really big.   Would be amazing to go here point to a file we want printed for out miners and pay someone with ability to do it.

Totally, this could turn into something really cool, and bitcoin is the perfect way to pay for it all.

[Vorta]

Sorry for not doing the second part yesterday. Business dinner popped in and I couldn't do the post. But I plan to concentrate on fans, fan ducts and legs today after work. Before that, I'll try to answer the questions that came in after the first post.

what is the temperature difference between duct and no duct(using same fan)?

Since my fan is 130 CFM and original fan is 230CFM, the difference is about 15°C. I don't have written data of when I ran the first test. The problem are the temperature sensors. Thanks to having no duct, temperature sensors are highly affected by side airflow and read about ~5°C higher temperature, even though you can clearly feel the stronger heat radiating off the hash boards. Would not recommend using lower CFM fans without fan ducts. You will definitely shorten the lifetime of your miner and you won't be fully aware of how hot it is.

this is REALLY awesome work, and good for you for supplying these designs. I will print them out and test it on my S5 tomorrow. How did you attached the legs to the fan and fan to miner? I dont see a screw going in between them.

I'll get into this later today. The legs are designed so that they fit into Scythe's fan's gaps perfectly, you can also use screws that come with the fan to tighten them, but the miner is stable without that as well.

Out of curiosity what is time and cost to do this?  It is very very nice.

Also can I ask what 3d printer do you use for this?  And kudos for releasing designs that is above and beyond.

The cost is rather low if you're printing with ABS. Every part can be printed 80% hollow inside (20% infill). One duct is about 36g which equals about $1 of PLA in my case.
My 3D printer is Rostock Mini Pro. Just large enough to print a 120mm fan duct in it.

wow such a good looking miner..
how much temp difference using this mod?

Here are my stats after 7 hours of constant mining:

http://www.pohrani.com/f/1e/iW/1Jqovfj3/mine-black.jpg
http://www.pohrani.com/f/5/e5/4jz7lglA/miner-green.jpg

The new miner, which came with external heatsinks, has 5°C difference between hashboards. Still haven't found out why.

These are stats with only a push fan. Pull fan would lower temperatures by additional 5°C. I don't use it unless I'm looking at a really hot day or want to overclock miners to 425MHz.

Fantastic! Wondrous effort and very well done. I really I'm interested in the printed mods ..... and if possible an estimate as to cost.

With regard to the bottom cover, would a simple ridge at the intake end suffice rather than having the ridge all along the length of the cover? I assume that despite having underspecified the fans there is is still adequate air pressure to keep the stream flowing ....

I wonder what the result would be, hash wise & HW error-wise, to run the stock S5 fan, rather than your replacement ones (on original firmware) at an OC freq, say 412. I say this because I managed to drastically reduce the HW % out of an OC'ed S3 @freq 281, running a stock S5 fan in pull configuration at full pelt, aka no PWM. (PS. that also showed up my 550W PSU and had to switch to a 750W one!)

Thank you!
With my push-pull configuration I can run up to 450 MHz without overheating. But I noticed a drastic performance loss at 450 MHz. 425 seemed optimal. The reason why I wouldn't use stock fan is because it is louder than 4 Scythe Ultra Kaze fans. I think I have a way of demonstrating this.
When you see HW errors popping up in larger number, you can be certain you have some chips above 80°C.

Very nice work with the air ducting, I bet that's a huge difference.  I have an S1 and S3, plan on getting an S5, and I've noticed how much cool air comes out the bottom of the S1, vs the enclosed design of the S3.

Any plans to sell the top and bottom pieces?  I bet people would buy them.

Material-wise these are not a problem to make. But the print time of one on my printer is about 5 hours. Should there be enough interest nothing stops us from making bulk order from a company that does plastic products through molding. Both top and bottom ducts as well as top and bottom plates are easy to mold.

Excellent job, thanx! But it would be great if you design some stubs which will prevent air to blow away in all directions from the gaps between PCBs of the blades and plastic side panels. More directed airflow will make blades cooler. The bottom plate can be easily reworked for this purpose, it should be just about 20 mm wider from each long side, right?

Also for top and bottom plates - I believe you can put some voids in it to reduce the consumption of the material while the shape of the side faced to heatsinks will remain the same. Please sorry for my English. 

I need 3d printer now... For free... 

p.s. where is "Fan duct - Top.stl" on that photos? Is it for cooling S5 with 2 fans?

Yes, the bottom plate can be reworked to cover the gaps, top would be more difficult, but instead I made the bottom duct to direct the airflow and it is working good this way. Enough just to move the air around the external heatsinks, as internal heatsinks have much higher efficiency and air going out is just a waste.

The top fan duct is not there in the photos. I assembled the miners with one fan to make them less noisy. I'll put the second ducts up today and photograph it all. For now here are 2 photos of the prototype I dug out on my phone:

http://www.pohrani.com/f/34/qp/O5kodBV/prototype01.jpg

http://www.pohrani.com/f/2u/u2/3jIkPlVD/prototype02.jpg

This was the first prototype I ever printed. It didn't exactly match the fan and had rather thick walls, but it perfectly "clicked" into the miner.


Here's what the miners look now, powered up and mining:

http://www.pohrani.com/f/32/w3/imiAAaE/07.jpg

This is the power draw (350MHz):

http://www.pohrani.com/f/3r/xu/882xsqd/08.jpg

[Rabinovitch]

Don't you think that at the top you can (and maybe should) use the same duct as at the bottom? In that case the top fan will make additional vacuum for airflow through little radiators from back side of the blades to be more directed.

[Vorta]

Don't you think that at the top you can (and maybe should) use the same duct as at the bottom? In that case the top fan will make additional vacuum for airflow through little radiators from back side of the blades to be more directed.

That would reduce the vacuum done to the main heatsink, which is very important when miner is working in hot ambient, and would not affect the airflow on the side heatsinks. With the ducts as they currently are you can easily feel hot air moving on the sides. Which is enough. Giving it any more airflow by taking it from the main heatsink is just losing efficiency.

[Rabinovitch]

The more air blows strictly in direction from one fan to another - the more cooling efficiency, I guess. So if one fan pushes, and another pulls, then why should it lead to less efficient cooling?

[pekatete]

I'm thinking here that this is as good a precursory discussion to a complete redesign of the S5 housing as any, and laser cut perspex comes to mind to complement the clearly superior 3D printed fan ducts.
Just a question regarding creation of a vacuum, would that be with the pull fan running a tad faster than the push fan? If so, by how much more, say for a 130 CFM, 4200 RPM fan (max)?

[jstefanop]

If people in USA are interested I can probably start printing a few of these out that don't have access to a 3D printer.

[Rockett0]

Let us know cost for the kit with shipping. I'm interested.

80023 zip.

[valkir]

Should have the price for next wednesday 

[Vorta]

The more air blows strictly in direction from one fan to another - the more cooling efficiency, I guess. So if one fan pushes, and another pulls, then why should it lead to less efficient cooling?

The pull fan has effect in closed tubes where air meets resistance. You can, however, try using the same fan ducts on the top and bottom and measure differences. I'm expecting about 1°C higher temps with bottom fan duct also on the top.

I'm thinking here that this is as good a precursory discussion to a complete redesign of the S5 housing as any, and laser cut perspex comes to mind to complement the clearly superior 3D printed fan ducts.
Just a question regarding creation of a vacuum, would that be with the pull fan running a tad faster than the push fan? If so, by how much more, say for a 130 CFM, 4200 RPM fan (max)?

The pull fan would need to be much stronger than the push fan, to get the same airflow on the side where push fan is. I have no idea how much stronger.

I had a setup with a lower CFM pull fan (Corsair SP120L), this also worked better than a version without a pull fan. 2 ultra Kaze's seem best. Heatsink's thermal conductivity is too low to feel much difference with higher CFM. Heatsink's leaves tend to get cold further from the solid part and hot only up to 1cm from it. Because of the good airflow the rest doesn't even manage to get hot. If they were thicker, that would be a different story.

[pekatete]

..... Heatsink's leaves tend to get cold further from the solid part and hot only up to 1cm from it. Because of the good airflow the rest doesn't even manage to get hot. If they were thicker, that would be a different story.

So something like this may be a good one .... leaves are just 2cm tall and if all the airflow is focused on them, then ..... (not sure about the thickness of the base / solid part though on these).

[jstefanop]

..... Heatsink's leaves tend to get cold further from the solid part and hot only up to 1cm from it. Because of the good airflow the rest doesn't even manage to get hot. If they were thicker, that would be a different story.

So something like this may be a good one .... leaves are just 2cm tall and if all the airflow is focused on them, then ..... (not sure about the thickness of the base / solid part though on these).

Thickness does not matter for heat dissipation, just surface area(i.e. number of fins). Granted a thicker plate/fins will act like a bigger heat "buffer" so it could take more heat before it gets hot, but removing heat is only a function of airflow and the surface area of the metal (and a lot of other variables but we probably don't want to get into thermodynamics here).

You would most definitely fry out your S5 with the heat sink you linked, it probably has not even half he surface area of the stock heat sink.

What they should have done doubled the amount of fins with all that material they removed to make the heat sinks...since they could make the fins that thin they could probably have cut twice as many of them.

[aarons6]

if someone from usa makes these i would be interested in a top and bottom plate set.. i dont need the legs or the fan extension.

[Vorta]

..... Heatsink's leaves tend to get cold further from the solid part and hot only up to 1cm from it. Because of the good airflow the rest doesn't even manage to get hot. If they were thicker, that would be a different story.

So something like this may be a good one .... leaves are just 2cm tall and if all the airflow is focused on them, then ..... (not sure about the thickness of the base / solid part though on these).

http://i00.i.aliimg.com/wsphoto/v0/32330074820/2pcs-lot-300mm-140mm-20mm-High-Power-Silver-Aluminum-Heat-Sink-With-High-Quality-On-Sale.jpg

Thickness does not matter for heat dissipation, just surface area(i.e. number of fins). Granted a thicker plate/fins will act like a bigger heat "buffer" so it could take more heat before it gets hot, but removing heat is only a function of airflow and the surface area of the metal (and a lot of other variables but we probably don't want to get into thermodynamics here).

You would most definitely fry out your S5 with the heat sink you linked, it probably has not even half he surface area of the stock heat sink.

What they should have done doubled the amount of fins with all that material they removed to make the heat sinks...since they could make the fins that thin they could probably have cut twice as many of them.

It does matter if your fin is so thin that it dissipates more than it can conduct. If the heat is dissipated before it reaches the end of the fin - the fin is too thin. This, in result, also means the fin's total surface area isn't even fully utilized.

Btw, did anyone ever get the miner with heatsinks as pictured here?
https://www.bitmaintech.com/bm_tech_admin_at_123/userfiles/image/MAC_00504-s.jpg

[aarons6]

..... Heatsink's leaves tend to get cold further from the solid part and hot only up to 1cm from it. Because of the good airflow the rest doesn't even manage to get hot. If they were thicker, that would be a different story.

So something like this may be a good one .... leaves are just 2cm tall and if all the airflow is focused on them, then ..... (not sure about the thickness of the base / solid part though on these).

Thickness does not matter for heat dissipation, just surface area(i.e. number of fins). Granted a thicker plate/fins will act like a bigger heat "buffer" so it could take more heat before it gets hot, but removing heat is only a function of airflow and the surface area of the metal (and a lot of other variables but we probably don't want to get into thermodynamics here).

You would most definitely fry out your S5 with the heat sink you linked, it probably has not even half he surface area of the stock heat sink.

What they should have done doubled the amount of fins with all that material they removed to make the heat sinks...since they could make the fins that thin they could probably have cut twice as many of them.

It does matter if your fin is so thin that it dissipates more than it can conduct. If the heat is dissipated before it reaches the end of the fin - the fin is too thin. This, in result, also means the fin's total surface area isn't even fully utilized.

Btw, did anyone ever get the miner with heatsinks as pictured here?

yes the one i got looks like that.. the one OP has is the one with the s3 heatsinks.. i dunno if they are the first gen ones or if they ran out of s5 heatinks and took apart s3s to supply the newer ones.

[pekatete]

It does matter if your fin is so thin that it dissipates more than it can conduct. If the heat is dissipated before it reaches the end of the fin - the fin is too thin. This, in result, also means the fin's total surface area isn't even fully utilized.

Btw, did anyone ever get the miner with heatsinks as pictured here?
https://www.bitmaintech.com/bm_tech_admin_at_123/userfiles/image/MAC_00504-s.jpg

So really, my untrained conclusion from that is that the best fin would be an inverted V shaped type (thin at the top) ... and with the mods above, you'd not need the fins to be as long as they on a stock S5.

[Vorta]

It does matter if your fin is so thin that it dissipates more than it can conduct. If the heat is dissipated before it reaches the end of the fin - the fin is too thin. This, in result, also means the fin's total surface area isn't even fully utilized.

Btw, did anyone ever get the miner with heatsinks as pictured here?
https://www.bitmaintech.com/bm_tech_admin_at_123/userfiles/image/MAC_00504-s.jpg

So really, my untrained conclusion from that is that the best fin would be an inverted V shaped type (thin at the top) ... and with the mods above, you'd not need the fins to be as long as they on a stock S5.

That would be best for heat transfer, but V shape would give more space for the air to run through the part where fins are the thinnest.




P.S. I've updated the first post with missing details and photos.

[jstefanop]

if someone from usa makes these i would be interested in a top and bottom plate set.. i dont need the legs or the fan extension.

Im printing these out for few people in the US once I get my 3d printer up and running again. Ill charge cost in material plus a few bucks for my time and shipping cost. It will probably end up being around 5 bucks a piece.

A full set of the will take around 20 hours to print btw and I need to print 4 sets for me plus two people already wanting them so turn around might be a week or more. I have to buy myself one of those new laser resin printers 

(fyi "cheap" printing shops charge around $10 an hour to print, and were quoting me around 200 to print a set so i didn't have to wait for my printer to finish all of them, which is laughable)

[notlist3d]

if someone from usa makes these i would be interested in a top and bottom plate set.. i dont need the legs or the fan extension.

Im printing these out for few people in the US once I get my 3d printer up and running again. Ill charge cost in material plus a few bucks for my time and shipping cost. It will probably end up being around 5 bucks a piece.

A full set of the will take around 20 hours to print btw and I need to print 4 sets for me plus two people already wanting them so turn around might be a week or more. I have to buy myself one of those new laser resin printers 

(fyi "cheap" printing shops charge around $10 an hour to print, and were quoting me around 200 to print a set so i didn't have to wait for my printer to finish all of them, which is laughable)

If you do please share.  I would be blown away with one of the new resin printers.  Those are pretty amazing.

I think it might be better off with multiple PLA printer because of cost.  A tip for anyone looking for printers look into "make" community's in bigger cities normally.  There is a lot of printing gear with "makers" and they are normally very nice.   So that is one way possibly to get them printed.