EK Active Backplate Waterblock install for 4090 FE

[Jay-G25]

I posted a few pics of this in the main 4000 series thread but decided to do a separate thread in this section showing the process i took of installing the EK Active back plate on the 4090 FE , will include pics of the temps i achieved after doing do


Setup to accommodate air cooled 4090 after upgrading from a 3090 Water cooled Strix

Time to watercool that 4090 FE

Front , back and side of the EK Active back plate

Disassembly time.
The back plate is very easy to remove and just has a small clip to release next to the io shield , no screws as its held in magnetically . Very little cooling going on hence the lack of thermal pads to get the pcb out needed to release 2 fan ribbons and an rgb cable then just a case of unscrewing the clamp for the die and removing 6 torx screws from the io shield, a couple of screws around the pcb and the pcb should then lift out .

PCB removed , the stock thermal paste was like cement ... rock solid !

PCB cleaned ready for thermal pads , i went for better thermal pads than EK supply ... not cheap but keep the memory and vrm as cool as possible

Many thermal pads

s i am using Liquid Metal as the TIM it is imperative to protect the smd surrounding the GPU , for that used Thermal Grizzly varnish which i got from OCUK .. nice shade of red

Then on to the boring task of cutting all the thermal pads to shape and size .. i swear this stage took me a good 2 hours to do both sides

Conductonaut Liquid Metal applied to die , i also gave the mounting face of the block a light coating.

Backplate fitted and pressure testing for leaks

Time to drain the loop and get the pipework sorted . I actually got a bit lucky and was able to reuse the top complex hose for the gpu all it needed was a little cutting down but had to make a new lower pipe as previous was a bit short .

Filled loop and all back together

RGB Yo!

And on to some Temp data , At 450w the delta will be 20deg over water temp but in normal gaming loads which are around 350w this drops to a 13-14 deg c delta over water temp . Hotspot is 6-7 degc over core temp . The result of this is the card will now boost a few bins higher than previously and will hold its boost stable . Previously i could get a max of 3075Mhz running Time spy extreme where as now that is 3120Mhz due to the core being cooler. Overall very happy and it has done exactly what i wanted , bit more performance , runs much cooler , looks nice and doesn't make any noise

 

[Jay-G25]

Yes mate they have redesigned the port termainl for these blocks and can use any configuration you want

From EK site :

Direct Link Terminal​

These new Vector² GPU water block and active backplate sets feature a Direct Link flow solution - the first multiple block solution to implement a conventional 4-port terminal layout. That allows serial and parallel connections from adjacent or opposing sides of the block. It was made possible by the addition of a machined brass insert in the terminal itself that diagonally links the water block directly to the active backplate inside the acetal terminal. Direct Link proves invaluable when flexibility is required for the right setup, like in a case of a distribution plate or vertical GPU mount.

 

[Jay-G25]

Thanks mate
Much better solution
Being tied to certain ports was a bad design really

Yeah i was a bit worried about it myself at first thinking have i got it the right way round then when i found out what they actual designed was like oh that is actually a very good idea

 

[Jay-G25]

Can you overclock your memory as high as you could with air cooling. I've read on overclock.net people say it need to be warm and water-cooling is actually reducing its overclocking potential. Im thinking of using EK's stock thermal pads now because of this. Hoping its an isolated case though as like you ive just spent loads on after market pads.

I'm glad you asked this as that is exactly my findings, I'm no longer able to clock the memory as high due to it being cool . I thought I was imagining it at first but no I am no longer able to hold 1500mhz stable on the memory and had to reduce it to 1450mhz so although it's only a very minor decrease it is a decrease none the less . From all the videos I have watched on ln2 overclocking while not to the same degree of cold have heard many times that the memory clocks had to be reduce to get it to pass due to it not liking being so cold and imagine with this pulling so much heat out the pcb and memory is having the same effect

 

[Jay-G25]

Interesting to hear. Ill use the EK stock pads then. On the other forum they were saying memory likes to be around 45C to perform normally. It seems its all changed now which is strange considering its still GDDR6X all be it 2GB instead of 1GB.

Yeah for now best to use the ek pads and see how you get on , tbh will still be more than cool enough with ek pads just hopefully not to cool to effect memory overclocking, would be interesting to see your results when done as I could always swap them out but would really only matter in extreme benchmarks as 50mhz won't lose hardly anything as I am getting better scores due to the core boosting higher now . Temp wise the memory struggling to get out mid to low 30 currently so sounds like it's definitely needs to be a touch warmer in games I have it set to+1200mhz which is what I had on air and it remains rock solid stable at that clock and low temps on memory, its just when trying to get that very last bit in benchmarks

 

[Jay-G25]

Good pictures and information.

Hopefully non active backplate and standard EK thermal pads will keep the memory in its optimal window.

Got my block sitting here now too.

You should be fine I imagine but would be good if you can share your results aswell I've never had to take a card apart to try make it warmer and have worse cooling but will see how you guys get on

 

[Jay-G25]

It’s interesting that the stock FE cooler has minimal thermal pads and then EK go ham with them covering all the coils and everything.

Yep, I’ll share before and after, will be a little while still as going to build new rig on air for testing first while I clean all my WC gear.

Thanks both , will be good to see both your results when done

Yep EK went crazy with the amount of extra thermal pads they used considering when you look at the back of the stock back plate it basically does sweet fa for cooling ... EK with all the extra thermal pads pulling out the last bit of heat that makes the memory stable , Its no bad thing having a cool card just maybe they went a bit overkill in certain places .

 

[Jay-G25]

Excellent, post here to keep this thread a good resource for the EK 4090 FE block and then not lost in other threads.

Is it possible to try the ABP block without sending coolant around the backplate? Or would that require a different terminal connection like the stock blocks have?

I am not sure if you could use the active back plate without coolant running through it , the back plate is thick enough ( 11mm copper afaik ) so would be up to the job of cooling dry but would have to somehow swap the distribution block on the front to a stock one which should allow the coolant to flow through just the bottom section but not sure if it would physically fit this setup with the abp being in the way

 

[Jay-G25]

Ah ok OC2000, still good to see all the different configurations you can test.

Was just a thought to test if the ABP actually does anything in exactly the same conditions.

Jay, Have you noticed any coil noise since padding them to the block?

Yeah i am happy to see all sorts of configurations so share whatever you got

I think coil whine has actually reduced now , i had small audible bit on air but would say its definitely quieter ... i expected to hear it more as now no fan noise but no its very very quite

 

[Jay-G25]

I'm looking at the 9th, 11th and 12th pictures in your first post showing the die after cleaning. Is there a protective film on there to protect whilst you were varnishing, or is that really how clean you could get it, you say it was cement like? It somehow looks like TIM has "leached" into the die or at least tarnished it, never seen that before.

Only the smd have varnish on them which is red and protects them. The die is perfect, the patterns you see on the die is left over residue from the Noctua alcohol wipes nothing else or anything to be worried about

 

[Jay-G25]

Ah, good, I thought it looked odd for a clean

They mention about the residue in this video , they say it is only visual and can be polished if needed which i didn't do as was getting covered with liquid metal anyway

 

[Jay-G25]

Another temp test , this time running furmark with core clock and memory clock stock but with power and voltage maxed . Result is delta of water to core of 25 deg at 570w give or take so 22deg water , GPU core 47 deg and hotspot at 55 deg and pump speed is set at 2.5l per min / 150l per hour according to the byski flow sensor This is somewhat of a useless test as the only game i know that will have power usage this high is Metros Exodus Enhanced edition when running 4k with RT will happily use 540w+ but most games are around the 250w to 380w range and delta is a lot lower in this case . either way this is what it is at 540w+ .

 

[Jay-G25]

Great results, I especially like the delta between core and hotspot. I have the EK 4090 block on the way for the TUF. I had been debating getting different thermal pads but don't want to run into any issues. I presume the Gelid pads you used would be similar to the ones I would need if I happened to change them up. Did you get exactly the same size ones as EK or was it a bit trickier adding them together or so forth? Thanks.

I used Gelid Ultimate pads which are 15w/mk BUT i would strongly advice against this and would suggest you use the supplied EK thermal pads at 3w/mk and the reason is due to the Gelid pads overcooling the vram and not allowing it to get to a stable temp for max oc ability , the vram on these 4090 like to be above 43 deg from the info i can find and my own testing ... anything under this will reduce your memory oc by 50Mhz or so to get it stable. Im not to bothered about it on mine as in games i have the memory set to +1200 in AB and it is rock solid stable at that clock and has no issue with the temp being to cold just if trying to go for +1500 is not possible and has to be +1450 under 43 deg memory temp. Pad size i used exactly as EK stated so 1mm, 1.5mm and 2.0 mm on my card

Which pads did you use @Jay-G25?

I cancelled the active backplate order and ended up with the regular passive backplate, but would still like to use better pads if possible.

Presumably you also used the different thicknesses?

See above , stick to stock EK pads imo on the memory at least ... VRM etc can upgrade no issue. Used exactly the same size as EK state to use

 

[Jay-G25]

Finally put my EK water block on. I used EK pads for ram, but unfortunately didn't work. Ive gone from a stable 1850 to 1680 on the ram. What's worse is I don't think ive even gained any core bins! Seems water-cooling has proven nothing but negative results so far. Still early though.

Yeah the ram is strange on these , have seen people even report that on air it will black screen if they run it cold at max oc and have to reduce it to get it stable. Mine dropped just 50mhz on the memory to get stable although as I mentioned earlier in games I have it set to +1200 and is rock solid stable no matter what temp . Not ideal for extreme benchmarks but perfectly fine with a good stable gaming oc on the memory.

Strange you haven't gained any boost bins though. What does it do now and what would it do previously? For reference using heaven as an example previously I was able to get the core to 3135mhz at 55deg but under water could now reach 3180mhz at 40 deg ( artifacts at 3180 ) so around 3 boost bins due to temp . It will drop its 1st bin at about 43 deg . 3165mhz is about the max stable I can get my core without it producing artifacts.

I will agree from a performance point of view watercooling hardly offers anything over air cooling this time . Only plus I can see is a lot less noise , potentially better looks although the FE is stunning design and the card being a lot smaller in a waterblock .

 

[Jay-G25]

Just can't get memory over 38C really annoying. Port Royal used to run at 1850, now only 1500. Madness. May remove a few pads on the back. Coil whine is horrendous so I need to remove the VRM ones anyway.
Increased my speedway score to 11150 with loads of things in the back ground running on stock CPU and ram on XMP so plenty of headroom. Same with port royal got to 28600 so head room there too.

If they are genuine score with no artifacting then that is good imo . the best on my FE is 10940 for speedway and 28299 for Port royal . Your current score would put you no1 on the speedway thread on here so something is clearly working What were your best scores on air to compare ?

 

[Jay-G25]

Removing VRM pads eliminated all coil whine. Put EK pads all over back except over the CPU area. Re-ran Port Royal at 3075 mhz and 1730 mhz memory and score 28970
cpu is still stock no optimisations or anything. I still have loads of headroom for more. before removing VRM pads on the back I couldn't go higher than 1500 on mem.

Sounds like its certainly working now then , so you removed the pads from the rear of the pcb and it has allowed you to clock memory higher? I've not bothered stripping mine down as am happy with where its at temp wise . Next year when I switch cpu will look at redoing the thermal pads but currently don't see the need to chase that last 50mhz when my biggest issue holding me back in most benchmarks is the cpu and ram setup being so old now you have some really solid scores now though, very impressive

Also just to add can't say the coil whine has got any worse on mine , if anything was the opposite for me and became even quieter

 

[Jay-G25]

Obviously nothing is for certain, but removing the VRMs pads on the back has helped so far. Maybe I just didn't it enough time to bake in. A lot of variables.

What is for certain though is ASUS card with coil whine amplify it with a water-cooled back plate on the back. EK 2 years ago told me to remove them and it eliminates it. It has this time too. VRMs on the back never get hot so its totally fine.

I still need to try colder water to see if I can get a higher bin, but need a nice day to put the MO-RA outside.

My Strix squealed like a pig in its EK water block so I certainly know what you mean there interesting that removing the thermal pads shuts it up, never knew that! My FE is silent so will leave the pads on but good to know for future reference

As for a cold day I'm not sure where you are but here in London it's absolutely freezing at the minute.. like 3 deg outside . Perfect for some pc / rad outside benchmarks if mine didn't weigh 60kg I'd have it out in the garden getting chilled

 

[Jay-G25]

Didn't see this post.
I was getting the black screens too on air with max oc applied on my old system. I thought it was the DDR4 3800 CL14 Ram degrading so down clocked it to XMP.
I've only tried Port Royal and Speedway so far. On Air I think Port Royal maxed out at 3060mhz and Speedway 3105mhz. Under water Its 3075 Mhz on Port Royal and 3105mhz on Speedway. I haven't tried a curve yet. Heaven will be a good one to try too.
Im also using this new 17 mk/w Ocool Apex thermal paste which may need some time to cure. Its at least a month of testing which will be fun for me, but horrible for the family as it can get quite addictive

Looks like you got a top grade card if you can achieve over 3.1Ghz in port royal , my max is 3075/3090 on port royal... your memory goes pretty high as well so will all help towards the benchmarks . I can get my core to 3180Mhz in heaven but will start to artifact so assume this is the limit of the core at my current temp which ranges from 30-45deg in heaven. My FE would blackscreen on air when memory was set to +1550 but was stable after a quick benchmark to heat up the ram, it's bizarre that the memory is so sensitive to temp this generation

 

[Jay-G25]

This helps add a couple of bins

A man after my own heart , love a bit of PC in garden benching I've not done mine since Jan 2021 so high time i got it outside again .... hopefully stays dry the next couple of nights

Must say it worked as got my best benchmark scores with the water at 5 deg ..

 

[Jay-G25]

@OC2000 what are your core to water deltas like with what TIM?
Did you try Liquid Metal like Jay?

I’m getting my gear ready on air at the moment for testing, so soon considering what the “best” block installation will be.

Thanks

@Colonel_Klinck changed from paste to Liquid metal and was surprised at the delta difference he got , it certainly works on these cards