RTX A2000 Full Copper Low Profile Single Slot Edition - aka A2000 SpongeBob Copperpants Mod

BaK

King of Cable Management
May 17, 2016
831
850
I was just curious how a aluminum design would perform in comparison because it would be much cheaper to manufacture.
I was focused on efficiency only, but you're right, cost is another important factor!
I guess as a CNC owner the aluminium option will still be cheaper for you, but will the copper version without heatpipes still be that expensive?
I'm asking because from the tests I made with my CoolerCase (see sig), the vapour chambers were a bit disappointing in helping to spread the heat. This project is on hold but I will check that more deeply one day...
So maybe attaching the foam directly to the copper plate will work as good as with the pipes.


When the new heatsink is ready I'll try to summarize all experiments and make a "normalized" comparison.
Another idea I still want to try is soldering small copper fins (with or without heatpipes) to a solid copper plate
but first let's see how the new aluminum heatsink performs when it's ready.
Full of interesting things to come, sounds good! :thumb:
 

REVOCCASES

Shrink Ray Wielder
Original poster
REVOCCASES
Silver Supporter
Apr 2, 2020
1,749
2,436
www.revoccases.com
I was focused on efficiency only, but you're right, cost is another important factor!
I guess as a CNC owner the aluminium option will still be cheaper for you, but will the copper version without heatpipes still be that expensive?
I'm asking because from the tests I made with my CoolerCase (see sig), the vapour chambers were a bit disappointing in helping to spread the heat. This project is on hold but I will check that more deeply one day...
So maybe attaching the foam directly to the copper plate will work as good as with the pipes.





Full of interesting things to come, sounds good! :thumb:

A copper version is still an option if I can find a way to make it reasonably priced and with better thermal / noise performance compared to the aluminum cooler. Honestly I still haven't calculated the cost of my different ideas and I'm also not sure what the community would consider a good price / performance ratio. Maybe I'm just thinking too much about cost and people would gladly pay a premium for lets say - 5C better cooling?

Btw, I also try to keep the DIY community in mind and will share the designs here, e.g. if someone wants to get the CNCd aluminum cooler made locally.
 

scatterforce

Airflow Optimizer
May 21, 2018
345
224
A copper version is still an option if I can find a way to make it reasonably priced and with better thermal / noise performance compared to the aluminum cooler. Honestly I still haven't calculated the cost of my different ideas and I'm also not sure what the community would consider a good price / performance ratio. Maybe I'm just thinking too much about cost and people would gladly pay a premium for lets say - 5C better cooling?

Btw, I also try to keep the DIY community in mind and will share the designs here, e.g. if someone wants to get the CNCd aluminum cooler made locally.
The time with the CNC machine will be the same. Frankly, after spending 500-700 on a Low Profile GPU, I think copper makes the most sense. Keep prototyping with aluminum, as it's cheap.... then, try a solid copper block with fins routed out, fan on one side, just like your aluminum design. As cool as the copper sponge is, I think it adds complexity, cost, and a level of skill to apply it correctly each time.
 
  • Like
Reactions: REVOCCASES

Goatee

King of Cable Management
Jun 22, 2018
717
1,484
The time with the CNC machine will be the same. Frankly, after spending 500-700 on a Low Profile GPU, I think copper makes the most sense. Keep prototyping with aluminum, as it's cheap.... then, try a solid copper block with fins routed out, fan on one side, just like your aluminum design. As cool as the copper sponge is, I think it adds complexity, cost, and a level of skill to apply it correctly each time.
Additional option is to run run some heat pipes in channels on the aluminum, have them direct contact with the core to give you a third compromise option.
 

scatterforce

Airflow Optimizer
May 21, 2018
345
224
Additional option is to run run some heat pipes in channels on the aluminum, have them direct contact with the core to give you a third compromise option.
Yes, with large runs, that is by far the most cost effective solution. The issue is complexity.... you have to ensure the heatpipes are flat/smooth when in contact with the core and they need to be soldered to the aluminum fin stack. This adds time, and might not be worth the effort for such small batches/one off parts.

I realize copper is about x4 the cost of Aluminum, but is the extra time machining and soldering worth that price difference?

(the x4 was from google, I'm not in manufacturing so I don't really know what machine grade copper goes for)
 
  • Like
Reactions: REVOCCASES

Goatee

King of Cable Management
Jun 22, 2018
717
1,484
Yes, with large runs, that is by far the most cost effective solution. The issue is complexity.... you have to ensure the heatpipes are flat/smooth when in contact with the core and they need to be soldered to the aluminum fin stack. This adds time, and might not be worth the effort for such small batches/one off parts.

I realize copper is about x4 the cost of Aluminum, but is the extra time machining and soldering worth that price difference?

(the x4 was from google, I'm not in manufacturing so I don't really know what machine grade copper goes for)
Agreed, that's one way of doing it and @REVOCCASES already soldered on his original build so he could do that again, other options is to clamp a small piece of thin copper to the heat pipes like Streacom do. Pack it with thermal paste and it still works very well. Run the channels in the aluminium base and your good to go.



Other considerations are copper can be a bit challenging to CNC

Overall @REVOCCASES your doing a cracking job and love all the ingenuity whatever you come up with.
 

scatterforce

Airflow Optimizer
May 21, 2018
345
224

I came across these and thought of you @REVOCCASES

I'm not sure if tapping threads in one of these is an easy prospect, but it's kinda neat.
 
  • Like
Reactions: REVOCCASES

REVOCCASES

Shrink Ray Wielder
Original poster
REVOCCASES
Silver Supporter
Apr 2, 2020
1,749
2,436
www.revoccases.com
some minor updates on the project....

things have slowed down a bit due to COVID related lockdowns around my area but I was able to make two more aluminum prototypes:

1) I managed to slim down the CNCd aluminum design and am now very close to a "True 1-Slot" card (only ~ 0.6mm over specs). That should now allow me to cramp the A2000 in a 1L workstation (P350 Tiny) as planned.









2) I've also finished one sample based on @robbee 's idea - unfortunately this turned out more troublesome (expensive) to machine as I had hoped on my dated CNC XD



Also, I can't do any testing currently because I'm missing some screws etc due to the lockdown...

In the meantime I have contacted some suppliers that can do skived aluminum and copper heatsinks like @scatterforce 's idea. They confirmed that would work but unfortunately they are not interested in small quantities.



after the lockdown is lifted I still want to try one design with soldered-on copper fins and then summarize the results & settle for one solution...
 

Michael

Average Stuffer
Jan 3, 2016
74
34
Don't coolers normally do copper blocks with aluminum fins to save on cost without much performance loss?
 

SFFMunkee

Airflow Optimizer
Jul 7, 2021
284
248
Don't coolers normally do copper blocks with aluminum fins to save on cost without much performance loss?
It'd be completely dependent on use case.

My theory/2c ( disclaimer: not an engineer )

I assume cooler designs are heavily dependent on:
- the max design temperature for the die/junction (e.g. 95C)
- the max (and typical) heat generated per second (e.g. 65W = 65 joules/sec)
- the chip's surface area (e.g. 81sq. mm)

- how long the different heat loads would be sustained (e.g. 5 min? 90 min? 24/7?)
- how fast heat can be dissipated into another medium (air/liquid) - though obviously the design itself will determine this, and
- how much heat can be dissipated before reaching saturation (e.g. max operating temp of liquid systems or max ambient temperature)

The more heat being generated, or likewise the smaller the die surface area, the faster the heat will need to be moved away from the die surface, bearing in mind the (max) temperature at the die junction is what really matters. To this purpose, copper shims (AKA heatspreaders) or cold plates are often used to spread the heat away from the quickly, thereby effectively increasing the contact area between the die and the heatsink / dissipating element(s).

This is also probably why a lot of designs employ heatpipes/vapour chambers, as obviously the aim is to pull heat from the die into the heatsink and subsequently the air.

I imagine if the heat being generated is sufficiently low - or the surface area is already large enough - there may no be any benefit to adding any copper, if aluminium can already adequately spread/absorb the energy fast enough it won't blow past max operating temps at the die surface.
 

REVOCCASES

Shrink Ray Wielder
Original poster
REVOCCASES
Silver Supporter
Apr 2, 2020
1,749
2,436
www.revoccases.com
Don't coolers normally do copper blocks with aluminum fins to save on cost without much performance loss?

like @SFFMunkee explained it quite well it depends on the requirements. You don't always need copper and/or heatpipes.



The original A2000 heatsink is a extruded aluminum profile with skived fins & some CNC work. This is a rather simple and very cost effective solution but it get's the job done. Surely a more complex cooler with copper, heatpipes and fins would perform better but NVIDIA decided that it's not worth the cost increase.

For my single slot mod you'll also have to consider a few other points except sheer performance. I'm limited to 13~15mm cooler thickness only and we are speaking of very low quantities plus limited manufacturing capabilities. That being said I'm trying to find a solution that is "reasonably priced" while offering "good enough" cooling performance.
 

SFFMunkee

Airflow Optimizer
Jul 7, 2021
284
248
I assume skiving for a custom unit would be incredibly expensive for small production runs (lots of tool costs and labour in tooling setup etc) but would it be possible to CNC some sort of reasonably common heatsink to fit?

(Or if you really want costs low, some sort of extruded heatsink, but I assume that would be lower performance as you couldn't achieve the same extreme thinness of fins... plus you'd have limitations in the alloys that can be extruded, so probably needs to be maybe 6063 Aluminium..?)

Other thing to consider is - how consistent and directional is the airflow? If it varies a bit, would you perhaps get improved performance from something with round pins/elliptical fins instead of flat fins? Just throwing it in the mix for extra confusion :p
 
Last edited:

SFFMunkee

Airflow Optimizer
Jul 7, 2021
284
248
Oooooh, what about vapour chambers...?? (I assume v. expensive but have any of your suppliers quoted you on them yet?)

 

REVOCCASES

Shrink Ray Wielder
Original poster
REVOCCASES
Silver Supporter
Apr 2, 2020
1,749
2,436
www.revoccases.com
I assume skiving for a custom unit would be incredibly expensive for small production runs (lots of tool costs and labour in tooling setup etc) but would it be possible to CNC some sort of reasonably common heatsink to fit?

guess you missed one part from the last update... ;)

In the meantime I have contacted some suppliers that can do skived aluminum and copper heatsinks like @scatterforce 's idea. They confirmed that would work but unfortunately they are not interested in small quantities.


I've also asked them if they had something "off the shelf" which I could mod but nothing really suitable unfortunately.

Or if you really want costs low, some sort of extruded heatsink, but I assume that would be lower performance as you couldn't achieve the same extreme thinness of fins...

true, with extrusion you can't get the same fin density as with CNC. best fin density could be achieved with skiving but finding a supplier willing to set up their machines just for a few parts is near impossible.

Oooooh, what about vapour chambers...?? (I assume v. expensive but have any of your suppliers quoted you on them yet?)

a custom made vapor chamber cooler would be too expensive. only way would be finding a suitable cooler and mod it. I think heatpipes (as I used on the Spongebob Cooler) should be overall more suitable (and cheaper) for this project.

Speaking of Spongebob, I might just try one more design with copper sponge and then compare performance and cost vs the CNC aluminum version :p
 

SFFMunkee

Airflow Optimizer
Jul 7, 2021
284
248
Speaking of Spongebob, I might just try one more design with copper sponge and then compare performance and cost vs the CNC aluminum version :p
Yesssss, let's see what cooling performance can be achieved while still keeping as close to within-HHHL dimensions as possible!

I had seen the previous update but somehow my brain let it drop before processing :p
 

Schestex

Case Bender
New User
Mar 31, 2022
2
2
Am glad to have found this forum here, your 1Slot solution I find terrific.
Sorry that I have no suggestion for improvement to your idea, but I hope that you can help me.

I could not find anything where I could order an A2000 waterblock, could you make me a template for a heatsink?
Would be really strong, so I could then connect an adapter on it and connect them to my external radiator.
 
  • Like
Reactions: SFFMunkee