'Gainward Phantom, Phantom GS, Palit GameRock, GameRock OC RTX 3090, 3080 waterblocks'

Some sort of design flaw it is. How drastic, that I do not know. That huge air gap does not transfer heat at all. These are components that are designed to be cooled via thermal pads, but now there is no contact to pads. Those components do no have separate temp sensors, so there is no real knowledge of how warm they operating. Looks like not enough to crash the system, but enough to reduce card lifespan? I don't know the AC supply chain, but hiding problems can be very costly too. I do not care about looks.
I agree, it does look like a flaw not cooling those components...does other 3000 series blocks from AC have a similar cutout, or from other manufacturers for that sake?

If someone got some temp sensors laying around it would be interesting to see what the deviation in temps are and if the temp for the components in question is within specification(even though this would not be internal temps). I’m back from vacation in a week and will do some testing then.

Could it be that the cutout was from a commercial illustration image ment to show thermal pad placement and somehow got used in production?

Just to be clear this is purely speculation from my part on what looks odd, it would be good with a clarification from AC regarding the matter.
 
Last edited:
Hi @polym0rph, thanks for the pictures. The pads are correctly positioned according to the manual. I'm being honest here! Unfortunately, I can't tell you anything about the technical significance of the recesses. The creative mind that designed the cooler and the manual is currently out of the office. Unfortunately, I can't give you the technical reason for the time being, but I'll give it to you when I have it. I hope that's all right! How does the cooler run? Are the values OK?
Hi @Thomas_S

Any news on these cutouts/recesses from the coldplate?

The LR22 caps directly under these recesses have no direct cooling and Im a bit afraid it could lead to problems like cracking or early degeneration of the components.

I tried to find temperture operating range for the LR22 inductors and think it is -20 -> 120 degrees so it seems these can get quite hot.
 

Eddy

Iceman
Staff member
The exposed pads are coils not voltage transformers. Actually, the instructions are wrong, no pads belong there at all, because the coils do not need to be cooled.
 

M1xka

New member
The exposed pads are coils not voltage transformers. Actually, the instructions are wrong, no pads belong there at all, because the coils do not need to be cooled.
Thanks for reply. And it looks like you are right. I found Guru3D teardown and indeed, there are no contact:

3090.png
 
The exposed pads are coils not voltage transformers. Actually, the instructions are wrong, no pads belong there at all, because the coils do not need to be cooled.
Thank you for the clearification?

And a bit assuring to see that stock air cooler haven’t active cooling for these caps. However the airflow with stock cooler will be quite high as the caps are just below the heatspreader.
 
Last edited:

church

Active member
RosaPanteren: OK, even without heatsink obviously those elements will be somewhat cooled with stock cooler's air stream. But then again due rest of card, especially main heat generating components, being cooled much better and at lower temps with LC, those bits will dump/conduct way less heat onto PCB, reducing need to cool other bits without own heatsinks.
I'd still love to get active cooled backplate though for mem chips on back of card :). Yes, it's not that bad without it, yes, DIY ways like for pogostov also can do the job .. but still, nothing will beat looks wise propely designed one.
@Eddy by chance no news from that front? :) Also have you considered making dedicated thread to such (in case if it's being designed as universal and thus fit for many cards other then ours too)?
 

church

Active member
@Eddy feel free to dump here pics even with first concepts if it seems acceptable. While waiting for final product it's always interesting to find out about development process / what it might have been / what hurdles were met and so on :). Talking about it, suggesting something, rating possibilities, cons, pros, voicing wishes and preferences always more interesting then just plain waiting and speculating with guesses in absence of any info :)
 

Eddy

Iceman
Staff member
@church
I cant show concepts because we dont show ideas until we are sure we will make them and we have a very rough ETA for it. We also don't show discarded ideas because they might be taken up again later and the competition is also reading along. ;)
 

church

Active member
Pitty. I get business rationale behind, but it seems very interesting to see "how people do things" during developing phase of products i'm interested in. Probably some brainstorming, some 3d CAD modeling, 3d printing & cnc prototype machining, several testbench testrigs mix?
 

Eddy

Iceman
Staff member
The usual process in product development is.
Idea -> discussion about sense and nonsense development/CEO/marketing/design - features are defined -> design team makes first sketches - technical implementation is roughly checked - design team makes finished sketches and 3D renderings - these are discussed - development gets the design they have to implement - there is a debriefing on the technical details - the 3D drawings are virtually checked for compatibility and function. Finished development - 3D drawings are virtually checked for compatibility and function - Sample is finished - Sample is checked for function and performance - Refinements from development and design department follow if necessary - The whole thing goes into production - Instructions are created - Launch planning.
Participating departments are always Design and Marketing and CEO. Development then only in the technical area or when it comes to implementation. CAD models and prototypes are only a small part of the process.
 

Hakanese

New member
I am using Mp5works backplate my Junction/hotspot temps have not exceeded 60C even after 7hrs of torturing the card with 420W.
Any active cooling will result in big drops in memory temps, especially since the Alphacool block comes with pads for the backplate. I just wish they had done the active backplate sooner
 

Junglst

New member
Received my blocks at the start of August and thought I'd provide some feedback.

Have 2x Gainward 3080 Phantoms - ordered 2x blocks and the Acool Aurora GPX SLI connector.

Firstly some feedback regarding the SLI connector - the water jackets on the blocks and SLI connector are oval shaped - the O-rings are O shaped. I found the recess on the SLI connector ( my one at least ) was not deep enough so had an extremely difficult time having the O rings seat correctly on the connector when mounting onto the block.

Very frustrating and difficult to mount as the O-rings kept popping out when trying to connect the cards together.

After trying numerous times filling up my loop and getting leaks at the O-rings / connectors - I decided to scrape away some of the acrylic to make the recess deeper so the Orings would stay in place long enough so I could get a water tight seal.

I may have received a bad SLI connector but I think this may be a design fault as the recesses in the original blocks ( which are also oval shaped ) the O-Rings fit without popping out.

Eventually worked - but as a customer I shouldn't have to resort to doing this

@Thomas_S / Acool please review these blocks to ensure the O-rings seat properly and the recesses are within tolerance in SLI GPX connectors!

Either that or provide some oval shaped gaskets / O-Rings which will fit correctly and stay in place for installation

Now onto the blocks - used all the pads included and followed instructions. After I fixed the SLI connector issue booted it up and was getting roughly 76c whilst mining - a reduction from 95c with the air cooler.

After a couple of weeks I found one of my cards started to hit 90c while mining ( first card in the loop / the other card stayed around 76 )

From this forum - realised some pads may not be getting good contact with cold plate so decided to drain the rig and open the cards up again. Before doing that I ordered some Thermalright 12.8 w/mK pads - 1mm and 1.5mm.

When I opened the hotter card - found that the some pads covering LR22 chips was not making good contact.

Memory chips seemed to be ok but some could have been better - found an oily residue left on the mem chips from the Acool thermal pads provided

Replaced memory chips with Thermalright 1mm pads - replaced the other strips with 1.5mm as the forum suggested.

Did this with both cards.

With the Thermalright pads in - the hotter card has now reduced mem temps to 60c while mining - yes 30c less than stock pads.

The other card is roughly 4c higher - this can be expected as the the hot coolant from card 1 goes straight to card 2 before going into a radiator

Hope this helps someone!

Edit - Sorry guys didn't take photos as I really wasn't in the mood to take my time messing around with it. Wife was already upset that I had computer parts everywhere.

Edit 2 - uploaded mining mem temps. Ambient temps have picked up in Australia and no air conditioning in my room but at 5am before work temps were 60/64

 

Attachments

  • Screenshot 2021-09-26 200103.png
    Screenshot 2021-09-26 200103.png
    9.6 KB · Views: 18
Last edited:

Thomas_S

... the nice marketing guy next door
Staff member
I may have received a bad SLI connector but I think this may be a design fault as the recesses in the original blocks ( which are also oval shaped ) the O-Rings fit without popping out.

Eventually worked - but as a customer I shouldn't have to resort to doing this

@Thomas_S / Acool please review these blocks to ensure the O-rings seat properly and the recesses are within tolerance in SLI GPX connectors!
For this type of issue, please always contact our support. It is not intended that the customer "scratches around" on the acrylic to achieve functionality. Support would also have helped with the cards by replacing the pads, as I assume that one card is definitely also affected by Nvidia's package problems. Nevertheless, it is good to hear that it is now apparently working satisfactorily.
 

Junglst

New member
For this type of issue, please always contact our support. It is not intended that the customer "scratches around" on the acrylic to achieve functionality. Support would also have helped with the cards by replacing the pads, as I assume that one card is definitely also affected by Nvidia's package problems. Nevertheless, it is good to hear that it is now apparently working satisfactorily.
Thanks @Thomas_S - I know acool would have replaced the item but shipping to Australia would have been a minimum 2 week turn around perhaps longer.

Unfortunately I'm too impatient to wait for that.

@Thomas_S are you able to please have a look at a sli connector yourself and see if you have the same issue?
 
Top