XFX 7900XT merc 310 high junction temperature

[DirtyAlpaca]

Quote of Arran

Is that hotspot temp particularly bad considering the +15% power draw? theres a limit on what the heatplate can do to certain spikes, and the average temperature (if thats what the GPU temperature is) seems low enough that the waterblock is doing its job just fine and thatr the Hot spot is just going to spike very high when overclocked. My XT sees 77c in games with only 5% power limit increase, 320watts. with 50c on the GPU temp. Is there anyone who can give more reference to their XTX with max power limit? as i know it would easily hit 110c on air. Just the RT cores alone seem to run incredible hot, i drop 25C by disabling RT in games. Id argue youre woying about nrrothing here from a temperature standpoint. The bow in the backplate may be a aesthetic over sight but doesnt seem cause any performance issue. You mentioned the rapid swings in temperature but that's normally a positive, it means the cooler is doing its job quickly, dissipating heat after a quick rise or spike.

Hard to tell who you are quoting here, since many people posted their complaints.
I'm trying to find the throughline in your post, but I cannot spot it, so I'll split some things up here for my own sake.

My XT sees 77c in games with only 5% power limit increase, 320watts.

I think you own a reference card since you are talking about "+5% PL and end up with 320 Watt of TBP.
The MERC310 is actually a reference card with minor changes, but flashed with a different performance bios. So we are slightly discussing a different story here.

Just the RT cores alone seem to run incredible hot, i drop 25C by disabling RT in games. Id argue youre worrying about nothing here from a temperature standpoint.

When I posted that Cyberpunk picture it shows [1440p/ultra settings] 144FPS (capped). It's impossible to run CP2077 on 144FPS with RayTracing turned on. So that was with the setting turned OFF.

Is there anyone who can give more reference to their XTX with max power limit? as i know it would easily hit 110c on air.

Yea, maybe.

The bow in the backplate may be a aesthetic over sight but doesnt seem cause any performance issue.

I eventually came to the same conclusion. The GPU is supported by the sub frame. The impact of the PCB bent is neglectable and I agree with you about the aesthetic thing. The block disperses a vast amount of heat, luckily my radiator setup is beyond regular to extract it fast enough. Not everyone has these 'fat' rads.

Quote of Arran about 30% overclocking

7900XT boost clock as standard is 2450mhz, the base clock is 2050, youre running 2751mhz at 350w and are wondering why its running a bit warm, thats a 30% overclock. the hot spot is the single hottest sensor and its still 13c below max. i have no idea what your issue is here.

So many numbers in this thread it's becoming dazzling.
AMD's reference card has a boostclock of 2400MHz. The performance bios on the MERC310 Black runs on about 2736MHz. That's a 14% overclock. AMD's reference TBP of 315W. The XFX MERC on approximately 349W. That's a 4.5% increase.

REFERENCE BIOS BENCHMARK
You forced me to run a benchmark with the reference bios on the MERC310 Black edition.
2400MHz boost clock with only 309W TBP. Not convinced yet?



Some of us place waterblocks for aesthetics, some for improved cooling performance and some for silent builds. I know performance is a variable thing, but check the following pictures.

AIRCOOLER PERFORMANCE
This was screenshotted with the factory heatsink (aircooler). Left side 35%ish fanspeed, right 100% fans.


----------

WEBSITE INFO
I purchased for improved cooling, so RDNA3 can do it's thing and clock better with lower voltages. Like written on the product page.





So where did you pull the 30% overclock from? Who are you trying to defend here?


Apologize me for the boring lecture
I have some parts on the way, like a modified block to close this once and for all.
There is this theory in my mind where I have an assumption that the GCD is placed slightly lower than the memory complex dies.


Stay tuned for mid September.

 

[Deralique]

I can see from those photos your thermal paste has terrible coverage this will massively effect the hot spot temps.
My time spy extreme temps on the hot spot sit around 70C when the water temp is 35C. This is on the sapphire reference 7900XT with a +5% power limit, 1030 voltage. If yours is performing worse than the air cooler and is the reference block id have to say youve installed it incorrectly to reach those temps. Also in games my hot spot on the stock cooler would easily reach 100 if i was using +10% power limit

Really appreciate your input Arran - I'll definitely be re-installing the GPU block tonight with a thicker application of thermal paste. Method before was a thin application manually spread across the entire die. I'll be much more liberal this time - hopefully that helps. Just FYI - the temp I was seeing was with no OC, no increased power limit - just default settings in adrenaline . . . again if the the block isn't making good contact with the die it's all irrelevant - temps will suck regardless. I'll repost tonight after reinstalling and refilling the loop. Fingers crossed . . .

 

[Deralique]

Repasted and reinstalled the block last night. Used a much more liberal application of paste - see image.

Fixed the issue with the bowed back plate - happily admit user error here. The outer most thermal pads were placed in the wrong position. Red squares indicate where the pads were - fixed now with card/block exhibiting negligible bow.

Some testing

PC has 3 x 360 rads (EKWB PE) with Phanteks T30's
5800X with EKWB Velocity Block
D5 Pump at 60%, Fans ramp to max 60%

All temps recorded by HWiNFO

XFX 7900XTX Marc with Alphacool Eisblock Water block
Adrenaline Stock settings - Ambient 20c
Idle GPU - 25c, GPU Hotspot - 30c

3dMark Time Spy stress Test
GPU - 53c, GPU Hotspot - 110c

RDR2 - Ultra Settings 4K
GPU - 54c GPU Hotspot 111c

Tested a few more games but very similar results and stopped there - . . . a 50c delta between edge and hotspot on load is way too high . . . and it's still performing worse than the stock cooler even when its contending with a +15% power limit and OC'ed.

Not enthused with these results and not looking forward to deconstructing my PC and putting back in a 360 AIO and the stock cooler. I just can't justify the cash on a block with performance like this . . .

 

[Deralique]

Quote of Arran

Hard to tell who you are quoting here, since many people posted their complaints.
I'm trying to find the throughline in your post, but I cannot spot it, so I'll split some things up here for my own sake.


I think you own a reference card since you are talking about "+5% PL and end up with 320 Watt of TBP.
The MERC310 is actually a reference card with minor changes, but flashed with a different performance bios. So we are slightly discussing a different story here.


When I posted that Cyberpunk picture it shows [1440p/ultra settings] 144FPS (capped). It's impossible to run CP2077 on 144FPS with RayTracing turned on. So that was with the setting turned OFF.


Yea, maybe.


I eventually came to the same conclusion. The GPU is supported by the sub frame. The impact of the PCB bent is neglectable and I agree with you about the aesthetic thing. The block disperses a vast amount of heat, luckily my radiator setup is beyond regular to extract it fast enough. Not everyone has these 'fat' rads.

Quote of Arran about 30% overclocking

So many numbers in this thread it's becoming dazzling.
AMD's reference card has a boostclock of 2400MHz. The performance bios on the MERC310 Black runs on about 2736MHz. That's a 14% overclock. AMD's reference TBP of 315W. The XFX MERC on approximately 349W. That's a 4.5% increase.

REFERENCE BIOS BENCHMARK
You forced me to run a benchmark with the reference bios on the MERC310 Black edition.
2400MHz boost clock with only 309W TBP. Not convinced yet?
View attachment 5485


Some of us place waterblocks for aesthetics, some for improved cooling performance and some for silent builds. I know performance is a variable thing, but check the following pictures.

AIRCOOLER PERFORMANCE
This was screenshotted with the factory heatsink (aircooler). Left side 35%ish fanspeed, right 100% fans.

View attachment 5483
----------

WEBSITE INFO
I purchased for improved cooling, so RDNA3 can do it's thing and clock better with lower voltages. Like written on the product page.

View attachment 5484



So where did you pull the 30% overclock from? Who are you trying to defend here?


Apologize me for the boring lecture
I have some parts on the way, like a modified block to close this once and for all.
There is this theory in my mind where I have an assumption that the GCD is placed slightly lower than the memory complex dies.
View attachment 5486

Stay tuned for mid September.

Is this modified block being provided by Alphacool?

 

[DirtyAlpaca]

Is this modified block being provided by Alphacool?

Thank you for sharing your latest test.
No, that block is not provided by Alphacool.

Details will follow-up, either here or on r/watercooling.

 

[juanpcm]

I still have the same problem, my loop is hard tube so i just cant keep doing this, its pretty annoying... i bought in a store in USA im probably stuck with it, I need to find a way to fix this... to the guy defending this product because of an OC, I had better temps with the stock cooler, on air, like i think most of us here... i have mine -10 power to keep temps at check. The other guy with 85 temps, although not as bad, still pretty bad temps for a water block.. how is alphacool not doing anything about all this? its clear that somethings wrong with this product, ive been thinking about shaving a bit the washers but i dont want to crack my $1000u GPU.. Alphacool testing the cards with furmark its useless, furmark wont move mine pass 70s.. any game will take it to 110 stock... they need to make a video showing what are "we doing wrong"

 

[Deralique]

Anyone have the “core” version of this water block? Similar results? Looks basically the same beside the inlets . . .

 

[Pifke87]

Anyone have the “core” version of this water block? Similar results? Looks basically the same beside the inlets . . .

Yes, received it last week for my XFX 7900xt and have the same issues as described here...

My first mount had temps at 110 within seconds, after remounting and pasting the temperatures went down to ~max 90° while the overall GPU temp is at low to mid 50s. This delta is not acceptable, since the stock air cooler never reached such temps.

I remounted it several times now, with different tourques and different methods of applying paste, nothing changed. Something seems to be uneven, because when i look how the paste has spreaded after removing the block, i see that the center of the DIE has the most pressure while on the edges, little to no spreading happens. So either the pads for the memory are to big, or the washers on the edges do not fit or the block itself is uneven, i cannot say, but its a huge dissapointment to me, to spend 200€ for a waterblock, performing worse than the stock air cooler. Hope Alphacool comes up with something, i'm happily admitting some user error if this means i have a working waterblock.

 

[unknowncow]

Just dropping in to say that I have the same issues with this block. Started out with very high hot spot temps, close to 100°C. Remounted the block, same issue. Another remount and tightened the screws more, got the temps (hot spot) down to about 70-75°C. Still a bit high delta but I deemed it OK. This was 6 weeks ago. Now the temps are back to 102-106°C. Overall temp stays in mid 50s. Really don't want to drain and remount every month.

 

[Pifke87]

Little Update here..

in my desperation, i removed all of the thermal pads from the memory chips and it improved at least the DIE termperatures by a lot. From previous 85 to 88° in a timespy run with a delta of 30+ to 72 to 75° and a delta of slighty above 20° even without applying new thermal paste. Of course it cannot stay this way, because the memory junction was at about 90° during the run, so not so healthy over time, but i think i'll replace the thermal pads with softer ones and see how it goes.

btw. the removal of the thermal pads had no impact to the performance, at least for the short period of a timespy run

 

[Horushka]

Little Update here..

in my desperation, i removed all of the thermal pads from the memory chips and it improved at least the DIE termperatures by a lot. From previous 85 to 88° in a timespy run with a delta of 30+ to 72 to 75° and a delta of slighty above 20° even without applying new thermal paste. Of course it cannot stay this way, because the memory junction was at about 90° during the run, so not so healthy over time, but i think i'll replace the thermal pads with softer ones and see how it goes.

btw. the removal of the thermal pads had no impact to the performance, at least for the short period of a timespy run

That is some nice info and testing, thank you! I also thought about mem thermal pads being too thick/hard, preventing gpu die from good cotact with waterblock.

 

[Samweb86]

Finally I found this thread..
Alphacool Eisblock Aurora Acryl RX 7900XTX/XT Nitro with Sapphire AMD Radeon RX 7900 XTX PULSE OC

I used up more than 10 grams of thermal paste and two sets of thermal pads (not including the complete ones)..
I have reinstalled the water block over a dozen times..
Мy best temperature result is 45-92С*(-hot spot temp).. it's a huge delta! and this result is worse than an air cooler!
I'm just tired..

 

[Samweb86]

Compatibility of this water block and video card is questionable..(?)

 

[Deralique]

Have a refund request in now with Ali Express - hopefully they’ll approve. Any experience with the bykski block?

 

[Samweb86]

Have a refund request in now with Ali Express - hopefully they’ll approve. Any experience with the bykski block?

I have Bykski waterblock for Sapphire Radeon RX 7900 XTX. Bykski showed himself even worse. I hoped that alphacool would change the situation for the better..

 

[Luckystricker]

Hi

I also got the card myself now, the Black Edition Version A.2. I tested the card with the default settings and the Furmark bench test.

Of course, first with the original cooler for half an hour, here are the values. Fan speed fixed at 1800 rpm.


After the conversion to the water cooler, the temperatures were these. The fans on the radiators are set to a fixed 750 rpm.


And also a 10 minute stress test of the AMD software itself with the water cooler.



As a small note, this was only the third graphics card for me that I converted to a water cooler. And since I'm also following this thread here because of this, I also took the tip from @Vanzin war so that the screws on the cooler should only be tightened with 1 nm. That's why I also got myself a torque wrench of 0.4 - 2 nm. And the screws were only tightened with 0.4 nm, because I personally felt that was enough.

I only applied the thermal paste as a sausage to the GCD, with the six outer MCDs a small blob each. Thermal paste and pads are of course the ones that come with the cooler. I did that because I came across this post here from Igor'sLab.

Blob, sausage or solid? Apply thermal paste to the GPU correctly - we measured! | Reminder | Page 2 | igor´sLAB

Since people keep asking: Today’s test will once again show in measurement series how to properly apply a purpose-built thermal paste on large GPUs. The selection of a suitable paste and the…

www.igorslab.de

I hope this helps one or the other to solve his temperature problem.
Best Regards

 

[slader]

So I have been watercooling since my 9800gx2 days, and mounted more cards then I can count because I also do this for friends.

Since I was also getting very high temps and deltas in the 50c to the point of the card crashing and not showing any output at times so I decided to tear it apart.

I was getting a very slight bowing on the card but was nothing near the 1nm to 1.5 nm torque spec recommended in this topic... I tried getting near that spec but would not recommend it.

Since there is speculation about the thermal pads I also dry mounted the block and measured the gaps to the memory etc with a measuring gauge. The gaps fluctuate between 0.7 and 0.8mm, I am unable to measure the gap to the gpu itself, but seeing these clearances you won't be able to make 0.5mm thermal pads work so you really need 1mm pads.

That being said I do feel that the alphacool pads are very stiff. So I do see where the higher spec recommendation comes from I guess.

For me personally the thermal pads got destroyed so I'm waiting for new ones to rebuild.

On a side note the little o rings are questionable.
And pics are coming since the forum finds them too big.

 

[slader]

this is my setup: (in short I have a 280mm rad purely for the GPU)




the GPU out of the case:



checking for bowing:


the impressions from the paste and pads:


and the state of the washers:


(before and after temps will come later...since those files are on the pc)

Last edit: tested 1nm on an old 1660gt, it's strips the heads of the Philipsbolts, so please don't attempt.

 

[Rydeon]

same problem here! the card seems slightly bent to me, the hotspot is much higher than the temperature of the gpu. I look forward to some suggestions

 

[DirtyAlpaca]

same problem here! the card seems slightly bent to me, the hotspot is much higher than the temperature of the gpu. I look forward to some suggestions

Hello Rydeon. It looks like your card and waterblock run fine. The 7900xt(x) disperses a vast amount of heat and according to your screenshot there's no reason to feel troubled.

We are mainly discussing hitting the 105/110°C temperature wall on stock (and even below) speeds.

Enjoy your setup.