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

[GuilleAcoustic]

Searched on Digikey, but once you ticked "frameless" and 12V you're left with only 2 fans to chose from:

• Delta KFB04512HHAF0C - 49mm x 10.80mm
• Delta KFB0612HAFDB - 60mm x 12mm

Maybe it would be worth searching from Delta website for references.

EDIT: They do not have a tone of frameless fans .... https://www.delta-fan.com/technology/frameless-fans.html

 

[msystems]

Hmm 5v

Well. This is not going to help, but one interesting trick I discovered is that if Speedfan works..... then it can support a 5v fan (without step down) as long as it is not exceeding a certain PWM speed, equal to 5v. But it is a complicated solution since it requires extra motherboard header available, and speedfan compatible hardware controller, oh and also able to read gpu temp on smbus.

 

[REVOCCASES]

Searched on Digikey, but once you ticked "frameless" and 12V you're left with only 2 fans to chose from:

• Delta KFB04512HHAF0C - 49mm x 10.80mm
• Delta KFB0612HAFDB - 60mm x 12mm

Maybe it would be worth searching from Delta website for references.

EDIT: They do not have a tone of frameless fans .... https://www.delta-fan.com/technology/frameless-fans.html

Thanks for the links!

Looking at the requirements (12VDC, PWM, <=50mm diameter, <9mm thickness) the KFB04512HHAF0C seems to be the closest match. Unfortunately the airflow seems pretty low with 4.3CFM compared to 11.4CFM of the EFB0512HA I am using now. Not sure if the blower design would make up for it but could give it a try!

edit: it actually does not have to be frameless out of the box, that can be modded

 

[SFFMunkee]

I've also thought about this. Only issue would be that notebook fans require 5VDC instead of 12VDC and I don't think I can find enough space somewhere to integrate a small stepdown converter in the design

(Zener) Diodes? Alt wiring?

 

[robbee]

PS: would there be interest in a CNC machined heatsink for de-shroud / fan mods?

Looking at my test results I think something like this would work well enough with two 50/60mm fans...

I would love something like this. It's actually what I've been wanting to create myself but a CNC machined version would be a lot prettier.

What are the dimensions of this design?

 

[REVOCCASES]

I would love something like this. It's actually what I've been wanting to create myself but a CNC machined version would be a lot prettier.

What are the dimensions of this design?

Nice!

I haven't fixed the dimensions yet. If you tell me what you need (length, height, thickness) I could just make one for you accordingly.

 

[Michael]

I would love to purchase a single slot heatsink for my a2000. It would allow me to fit a meanwell psu in my case for a truly brickless build. I'd pay a pretty premium for one too!

 

[robbee]

I haven't fixed the dimensions yet. If you tell me what you need (length, height, thickness) I could just make one for you accordingly.

As wide and long as possible on the PCB. I guess that will be something like 55mm x 160mm. As for the thickness, if I'm not mistaking, the socket height is 3mm in your CAD file. If that is the case, a heatsink height of 20mm would be ideal in my case.

An alternative is to have the heatsink a little bit higher than the PCB (55mm) and have it at 60mm. The extra 5mm could protrude at the pci connector side if you leave a cutout for the locking tab. This way, you could even add mounting holes for the 60mm fans and squeeze some extra cooling performance out of them.

 

[rcodi]

Dropping a comment to say I'm interested in any improved cooler whether it's single or dual slot. Do you think there's any room to improve on the stock heatsink and make some sort of drop in replacement that fits under the shroud?

 

[SFFMunkee]

Dropping a comment to say I'm interested in any improved cooler whether it's single or dual slot. Do you think there's any room to improve on the stock heatsink and make some sort of drop in replacement that fits under the shroud?

Definitely room to improve, the original is a really basic aluminium extrusion from what I could see in @REVOCCASES's original post. Just about any heatpipe, material or airflow improvement would help thermals. For example the A4500 and above (maybe A4000?) use heatpipes as well, although they're obviously not LP.

TBH I'm keen to see custom-cooled A4000, stock is a single-slot full length cooler, but the PCB is tiny.

I'm estimating ~146mm, as it looks to be the same PCB length (obviously not height) as low power HHHL cards like the GT710 which also goes to the rear of the PCIe retention mechanism.

...Guess what else is about the same length, mini STX (147 x 140mm) !

 

[REVOCCASES]

As wide and long as possible on the PCB. I guess that will be something like 55mm x 160mm. As for the thickness, if I'm not mistaking, the socket height is 3mm in your CAD file. If that is the case, a heatsink height of 20mm would be ideal in my case.

An alternative is to have the heatsink a little bit higher than the PCB (55mm) and have it at 60mm. The extra 5mm could protrude at the pci connector side if you leave a cutout for the locking tab. This way, you could even add mounting holes for the 60mm fans and squeeze some extra cooling performance out of them.

good ideas, I've made a draft with two 60mm fans accordingly.... do you think that would work for you:

I would love to purchase a single slot heatsink for my a2000. It would allow me to fit a meanwell psu in my case for a truly brickless build. I'd pay a pretty premium for one too!

how tight is the "single slot" limit for you? my original plan was to make it a "true 1-slot" cooler but since I can not find a suitable fan with <9mm thickness that idea doesn't seem to work out. Now the question is if I shall continue with this project using a 10mm thick fan and just make it a 1.2 ~1.3 slot card (meaning 18 ~ 19mm thickness measured from the component side of the PCB).

Dropping a comment to say I'm interested in any improved cooler whether it's single or dual slot. Do you think there's any room to improve on the stock heatsink and make some sort of drop in replacement that fits under the shroud?

Definitely room to improve, the original is a really basic aluminium extrusion from what I could see in @REVOCCASES's original post. Just about any heatpipe, material or airflow improvement would help thermals. For example the A4500 and above (maybe A4000?) use heatpipes as well, although they're obviously not LP.

the stock heatsink is indeed quite simple but does the job just fine (70C under full load stress testing is acceptable IMHO). If you want to improve thermals I've found that by simply using better thermal paste (e.g. Thermalright TFX or TF8) and pads you can get GPU and memory temps down by about 5C. What I'm really not happy about is the stock fan curve. It doesn't go below 30% PWM and it ramps up veeeery slowly. Even at 86C GPU temp it only goes up to 60% PWM. So with better TIM&Pads and by setting a manual fan curve you should already see some good thermal improvements with the stock cooler.

 

[robbee]

Looking at the requirements (12VDC, PWM, <=50mm diameter, <9mm thickness)

I found another Delta fan that may be an option albeit also with low airflow. It has versions with 4 wires but I'm unsure if that means it has PWM support: https://nl.mouser.com/datasheet/2/632/BFB50x50x10mm-515795.pdf

It has a thickness of 10mm but perhaps if you cut away the frame, it becomes a bit lower?

EDIT: this is the version with PWM control. Mind you, the image on the website is wrong but the one on the spec sheet seems right: https://www.delta-fan.com/BFB0512HA-CX3D.html

There's also a similar fan from 'Gdstime' on aliexpress however that comes with only 2 wires so will be annoying without rpm readings nor pwm control: https://nl.aliexpress.com/item/1005001873960784.html

 

[Michael]

1.2 or 1.3 would work for me.

 

[rcodi]

the stock heatsink is indeed quite simple but does the job just fine (70C under full load stress testing is acceptable IMHO). If you want to improve thermals I've found that by simply using better thermal paste (e.g. Thermalright TFX or TF8) and pads you can get GPU and memory temps down by about 5C. What I'm really not happy about is the stock fan curve. It doesn't go below 30% PWM and it ramps up veeeery slowly. Even at 86C GPU temp it only goes up to 60% PWM. So with better TIM&Pads and by setting a manual fan curve you should already see some good thermal improvements with the stock cooler.

A noise reduction is my main hope for any sort of replacement, paste and 1.5mm pads did help but it's still the most audible part of my build currently. I love the 2x60mm design you mocked up!

 

[msystems]

Hmm, that reminds me, Noiseblocker makes a 50x10 12v fan which is dead silent. It doesn't move much air, but 2 or 3 of them in parallel could be effective on a low profile heatsink.

 

[REVOCCASES]

I found another Delta fan that may be an option albeit also with low airflow. It has versions with 4 wires but I'm unsure if that means it has PWM support: https://nl.mouser.com/datasheet/2/632/BFB50x50x10mm-515795.pdf

thanks robbee. I already tried this one but unfortunately it is still 10mm if you remove the housing and the airflow is really too low.

today I almost thought I found the perfect fan (from an old Quadro) - it fits like a glove for the design I made but then I ran a benchmark and found it performs much worse in terms of temps and noise compared to the normal DELTA 5010 fan I used before...

1.2 or 1.3 would work for me.

Awesome!

Actually I think a slightly oversized 1-slot card should work fine for most people - although I need to give up my personal plan to fit the A2000 in a 1L Lenovo P350 Tiny.

So here is the new plan for the next prototype: I'll slightly increase the thickness of the heatsink to fit a 5010 SUNON MagLev fan with a blade design for "mixed airflow" which (hopefully) should give a little bit more of both - good airflow and pressure

I love the 2x60mm design you mocked up!

If you're interested to give it a try, I am currently working on a quote for this design and should have a price by next week. Alternatively I have attached the drawings to this post so you could get a quote from a local CNC shop.

 

[REVOCCASES]

the new 5010 SUNON fan arrived today so I could do some further testing with the previous prototype heatsink I made...

I think the blade design looks very promising for the planned application...

...and with the frame removed, it should also push a good amount of air in radial instead of axial direction

as already mentioned, the previous prototype is too slim to fit a 10mm thick fan - so I had to improvise with some washers. It's not ideal to have a gap between heatsink and shroud but it should give a rough idea how well the fan performs.

before looking at the test results I'd like to note again that the stock fan curve is sub-optimal for my liking. Using the stock fan curve, the GPU will not thermal throttle with my heatsink design (e.g. go below 1200Mhz base clock) but it will get very toasty and therefore reduce boost/turbo clocks. I'm not quite sure why NVIDIA decided to have the fan running at only 60% RPM when the GPU hits 85C but that's how it is with my card. Long story short, for below tests I've set a custom fan curve in Afterburner which will increase fan speed linear from 30% to 100% between 50C to 80C GPU temp.

So here are the results which look quite promising IMHO:

Time Spy GPU score with the single slot cooler prototype is about the same as with the stock cooler:

With my custom fan curve, the A2000 settled around 77C in both Valley and TimeSpy which is about 7C more compared to the dual slot stock cooling solution. With the stock fan curve it would settle at 85C but I'm confident to get it down by a few degrees with a slightly optimized heatsink design.

Noise vs. Temps consideration:

• Idle: I feel the SUNON fan is more quiet in idle compared to the original blower, which is nice.
• Stress test: using the stock fan curve, noise is comparable with the stock blower - with my custom fan curve it is clearly louder but I'm confident that this could still be fine-tuned further down the road...

Summarizing I'm pretty happy with the results using the SUNON fan and have decided to make another prototype as shown below:

The new version will be about 20.30mm thick to fit the SUNON fan with a shroud and have three more "air channels" to increase the overall surface area. Hopefully this will get temps (and noise) down a bit more...

 

[BaK]

Dunno if the comparison is accurate with the fans variations, but what I see is:

Copper plate + foam and heatpipes:
GPU power: 68W
Fan speed: 3900RPM (48%)
GPU temp: 76°C

Aluminium heatsink fins:
GPU power: 67W
Fan speed: 5900RPM (88%)
GPU temp: 77°C

With noise linked to the speed of fans, the copper foam setup seems to be the winner.
Am I wrong?
If not, why ditching the foam solution?

 

[GuilleAcoustic]

GPU score is also higher with the copper foam

1. Stock: 6036
2. Copper foam: 5916
3. Aluminum Fins: 5790

@REVOCCASES, are you testing it in as open bench ? Is your ambient the same each time ? I see that your CPU temp is way hotter on the last version of the heatsink

1. Stock: 29°C (CPU temp)
2. Copper foam: 32.98°C (CPU temp)
3. Aluminum Fins: 48.95°C (CPU temp)

EDIT: It looks like the CPU frequency was way more stable (ie. always high, not droping) with the last test ... that could explain the CPU temperature difference.

 

[REVOCCASES]

Dunno if the comparison is accurate with the fans variations, but what I see is:

Copper plate + foam and heatpipes:
GPU power: 68W
Fan speed: 3900RPM (48%)
GPU temp: 76°C

Aluminium heatsink fins:
GPU power: 67W
Fan speed: 5900RPM (88%)
GPU temp: 77°C

With noise linked to the speed of fans, the copper foam setup seems to be the winner.
Am I wrong?
If not, why ditching the foam solution?

I was just curious how a aluminum design would perform in comparison because it would be much cheaper to manufacture. For now it seems the copper foam version performs better but also costs much more and has a few other disadvantages. Also, the conditions were not optimal for the test today. E.g. I had to offset the shroud to make the fan fit, therefore not all air was pushed trough the fins.

When the new heatsink is ready I'll try to summarize all experiments and make a "normalized" comparison.

GPU score is also higher with the copper foam

1. Stock: 6036
2. Copper foam: 5916
3. Aluminum Fins: 5790

@REVOCCASES, are you testing it in as open bench ? Is your ambient the same each time ? I see that your CPU temp is way hotter on the last version of the heatsink

1. Stock: 29°C (CPU temp)
2. Copper foam: 32.98°C (CPU temp)
3. Aluminum Fins: 48.95°C (CPU temp)

EDIT: It looks like the CPU frequency was way more stable (ie. always high, not droping) with the last test ... that could explain the CPU temperature difference.

the boost/turbo clock speeds are very sensitive to GPU thermals which explains the different GPU scores. I've seen user scores from 5500 to 6500 with the A2000 stock cooler so thermals are important to get the best possible performance. However IMO a small performance loss should be expected if going from a dual slot to a single slot design.

Yes, I'm testing open bench but ambient is not constant. Today it was about 5C hotter compared to my previous tests. CPU was probably hotter because my Antivirus decided to run a scan during benchmarks.

All things considered the conditions today are not suitable for a valid comparison but it gives me an rough idea that the new design with the SUNON fan should work and that it might be a less expensive alternative to the full copper version.

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.