X1900 replacement PWM heatsink - CNC

U

unixadm

Gawd
Joined
Jun 13, 2002
Messages
560
Hey guys,

Well I've been building my inventory of machinery so that I can try and start a side business down the road doing machining and fabrication. A week or so ago I picked up a small CNC Mill so that I can make some custom parts for vehicles, computers, etc. I'm working on some small engraving projects (like control panels with engraved letters above the switches) and recently I decided to look into machining a heatsink from T6061 aluminum. Thus far my mill is not up and running yet as I'm still waiting on the servos, but I hope to be machining parts via CNC by the middle of this month.

Anyhow I designed a replacement PWM heatsink for the X1900. It's my first part I've drawn to scale, modeled and created the CNC tool path. With that said it should fit like a glove. It's not super fancy, but then again it's functional. To cut this part I have to do it in two machining operations. One does the bottom of the heatsink and the other does the top. The stock heatsink is 3.67x.28x.5 or there abouts. My replacement is 3.67x.80x1.

Hopefully it will drop the temps on my X1900XTX a bit. What do you guys think?

http://www.dmwtech.com/gallery2/v/machining/ (X1900 album)

I also picked up a much larger mill that I'll be CNC enabling in the coming months:

http://www.dmwtech.com/gallery2/v/welding/machines/mill/

This machine will be able to do some pretty heavy cuts.
 
Looks nice. Wonder how much better the temps will be, as Iv seen my voltage regulators hit 99c when my card is overvolted. :eek: Or atleast ATITool says so. :rolleyes:

Id be VERY interested in buying one, if you make more than 1 that is. Should discuss that in pms tho.
 
Yes I can do copper. The issue with copper is that it's sticky. Machining it is not easy and I may have to use a slitting saw instead of an end mill. I'll make a couple out of aluminum and see how it goes. Copper is not cheap so I'd rather validate the design on cheaper material. Copper versus Aluminum on this probably would not big a big difference. The sheer size difference would most likely lower temps quite a bit no matter the material the heatsink is machined from.
 
I tell you one thing, if you setup a paypal account and make a few of them I am sure, people should buy them at $20 a pop. But you have to show that your design reduces temps better then what's already on their. You can use Systool from TechPowerup to reflect temps before and after.

If you use copper charge something like $25-$30 a pop
 
At $10 it would be a lost cause. I'd be losing money due to material and cutter costs. I would be highly upside down doing a copper part. That's just the reality of the world - metals are no longer cheap. Each heatsink will take close to two hours of machining time due to the fact that I have to work with 1/8" endmills due to the size of the slots. I also have to run more expensive 1/8" carbide endmills with a 1" cutting depth due to the depth of the slots. I can't make parts and sell them at a loss - common sense tells me that much. :)

I do a fair amount of welding as well and something as simple as a small gate frame could cost you just $50 in basic tubing material. One of my welding carts cost close to $200 just in materials alone.
 
Asian Dub Foundation said:
nice... if its any good... try make some for the nv 7 series and distributed around here!
Click to expand...
I don't have an NV 7 series card to reference. If you have a picture and some basic measurements I could look at what stock I could start with and come up with a rough design to look at. Things you have to watch for are component interference, hole patterns, width and the height. I'm waiting on the final parts so that I can wire up my CNC mill and start cutting some parts. I'm starting a side business doing fabrication and machining, so I wouldn't mind churning out some parts for people, but I can't do it at a loss. I also can't advertise anything here, but if you book mark my gallery page I'll post updates in the description as I make progress with this. You can always PM me and I'll send updates as I get them.
 
The source of the heat are the little voltage regulators.... those tiny little black squares under the stock heatsink.

The SMD air core inductors are not a heat source, and only get hot because they are next to the voltage regulators.

Concentrate on a contact surface that mates to the little black voltage regulators, and has bulk fins.

What the heatsink manufacturers do is start with a stock that has fins already, Then machine the mounting surface. Rather than using bar stock and cutting fins and mounting surface. Alot less work, just research up suppliers of bulk finned stock... like Alpha. Fins or Pins, I like Alpha's use of round fins. Great for results in an unpredictable (air flow direction) envoronment.

I'd also suggest altering the design of the mounting bung, rather than use the stock push pins, do something where a stock over the counter standoff, and two screws (one from under the card, one from above the sink) would make for a more secure and better aligned mount.

A dab of Artic Ceramique on each volt reg, then install the sink, and it'd be an excellent replacement.

I have no doubt you can show excellent results with your alternative. The stock sink with thermal pad is a really lousy solution.
 
Id be willing to spend 20-30 for one of those. Also Id be very interested in some memory coolers that aren't held on tho the video card with thermal glue- more along the lines of what viperjohn was doing for those x800's on his website. I believe the heatsinks were held in place by little brackets, I have wnated to get something like that for a long time.
 
xX_Jack_Carver_Xx said:
The source of the heat are the little voltage regulators.... those tiny little black squares under the stock heatsink.

The SMD air core inductors are not a heat source, and only get hot because they are next to the voltage regulators.

Concentrate on a contact surface that mates to the little black voltage regulators, and has bulk fins.

What the heatsink manufacturers do is start with a stock that has fins already, Then machine the mounting surface. Rather than using bar stock and cutting fins and mounting surface. Alot less work, just research up suppliers of bulk finned stock... like Alpha. Fins or Pins, I like Alpha's use of round fins. Great for results in an unpredictable (air flow direction) envoronment.

I'd also suggest altering the design of the mounting bung, rather than use the stock push pins, do something where a stock over the counter standoff, and two screws (one from under the card, one from above the sink) would make for a more secure and better aligned mount.

A dab of Artic Ceramique on each volt reg, then install the sink, and it'd be an excellent replacement.

I have no doubt you can show excellent results with your alternative. The stock sink with thermal pad is a really lousy solution.
Click to expand...
Thanks for the info. In the case of making one off or small runs of parts, it would be cost prohibitive to use a pre-finned stock. In a large production setting I agree that is the way to go. You limit your design to the stock in which you can purchase for the job. But in my case I'm doing custom parts, proto-types, etc. I don't mind spending the time machining a part to get the results or design I want. Heatsinks are a small portion of what I plan on doing. I'll test a few different options including using end mills to form the slots or a slitting saw. With a slitting saw I could cut much finer fins, but I'd have to see how much time that would add and if it would be worth the effort.

With this heatsink I plan to use nylon fasteners with a nylon locknut for securing the heatsink assembly to the card. I've used nylon fasteners in the past for this purpose with good results. Alignment isn't as critical as I've increased the width of the contact area to .030.

Time will tell. :)
 
I'm one step closer to making these. I got my coolant table going for my mini-mill. I just have to finish a few more screens to keep the water and chips from splashing out.

Pics: http://www.dmwtech.com/gallery2/v/machining/wt/

I should begin cutting this heatsink within two weeks.

It's water cooling for machinery, not computers for a change! :)
 
unixadm said:
At $10 it would be a lost cause. I'd be losing money due to material and cutter costs. I would be highly upside down doing a copper part. That's just the reality of the world - metals are no longer cheap. Each heatsink will take close to two hours of machining time due to the fact that I have to work with 1/8" endmills due to the size of the slots. I also have to run more expensive 1/8" carbide endmills with a 1" cutting depth due to the depth of the slots. I can't make parts and sell them at a loss - common sense tells me that much. :)

I do a fair amount of welding as well and something as simple as a small gate frame could cost you just $50 in basic tubing material. One of my welding carts cost close to $200 just in materials alone.
Click to expand...

Man your not thinking right. Use your little mill to make the parts . Test them when you get the results you want . Than take your part and blue prints to a job shop. thats set up to do volume parts . Than make money!!! Your little mill is not good for fast production work. But it will do fine for design and test and final . Than let the big boys produce them parts for you.
 
Involving a third party to make these would be cost prohibitive. These are $30 parts, not $400 parts. Due to the smaller working size, your cutter size is limited. You can only push an endmill so quickly through the material without breaking it. The advantage of this smaller machine is that I have a 10k rpm spindle. So I can actually machine parts more quickly at this scale than I could on say a Bridgeport. A nice machine center could get an even higher spindle speed, the cost per hour would be high. For doing engraving, aluminum and copper work for computer related equipment, this mill will work extremely well. Most people associate size with quality or capability and as long as the working envelope is there and you don't go crazy with your stepovers or your Z depth for each pass, you can achieve excellent results with a small machine. I certainly would not want to machine stainless or steel parts on a daily basis with this mill, but in a pinch it would do the trick.

I like to do as much as I can in house. That's just the type of person I am. I have a much larger mill than the one shown here. I just have to convert it to CNC. I'll use this smaller mill to machine my brackets for that conversion. This machine will take .250 off a block of T6061 using a 4 flute 1/2" endmill at about a 5000rpm spindle speed like it's nothing. As fast as I can spin the hand crank it will blow through the material with coolant. That's a full 1/2" width slot as well. For a machine of this size, that's fairly impressive.
 
Just a thought here - would horizontal orientated fins or a pin grid not be better?

I'm thinking this because of the original pwm coolers horizontal fins and the airflow of the cards stocker cooler. I'm thinking the volt regs will remain hotter because of less airflow getting in between the fins on your design. Only with cases with insufficient airflow of course.
 
I actually thought about that as well. The issue is machining time. But I'll play around with it tonight in the cad/cam program and run a simulation in Mach3 to estimate how long it will take to cut. Either way, size matters.
 
You must log in or register to reply here.
Back
Top