XFX 7900XT merc 310 high junction temperature

Quickce

New member
Just installed the bykski block with ptm 7950 on my XFX 7900 xtx. Results are attached for anybody interested. First pic is after 20 minutes of idle, 2nd pic is after 15 minute stress test on 3d mark. Before I installed the block I was getting 60c idle with 70c hotspot temps and 80c+ with 100c hotspot temps under load.

View attachment 5932
View attachment 5933
Out of curiosity, what was the board power setting that you used? Stock? +10% +15%? Which 3D Mark Stress test did you do?

I just recently installed a TG Phasesheet on my Alphacool block on my Merc 310 and I am experiencing very good results. in RAGE Bios mode, (403W) and when I do, I get a peak hotspot temp around 67-69C after a Steel Legend 20-pass stress test. I'm very heppy with these results. I'd just like to compare apples to apple with your test. Thanks!
 
Last edited:

Foomonchoo

New member
Hello,
Found this place after acquiring an Eisblock for reference 7900XTX off ebay and struggling with it and I would like to share my findings. First some caveats:
I'm installing this on the Yeston Sakura Sugar card. It isn't documented anywhere but since all the holes line up I can only assume it is using the reference board design.
This card runs aims for a 350W power target.
The Eisblock came from ebay without the 3mm pads for the backplate.
This is my first time ever attempting water cooling.

Initial install was abysmal. I spread my paste evenly as recommended by many and tightened all screws until they stopped against the board. I did not install the backplate since I didn't have pads. Average temps in the 50s and hotspot on the GCD was pegged at 110C in Furmark.

I noticed the bending in the board and backed off all the screws until it was flat against a ruler. This was marginally better, with the GCD hovering around 100C and bouncing off 110C and average temps in the 50s.

After finding reading through this thread I attempted to stack cheap thermal pads against the back of the core and installed the backplate. This was marginally better yet with the GCD hovering in the mid 90s, but still spiking to 105C occasionally. Others have noted that their cards maintain full boost clock at those temps, but this Yeston card seems to start throttling earlier and I could only maintain about 1500mHz in Furmark.

I took the whole thing apart one more time. I reapplied paste and this time just by laying dots on the MCDs and a fat log across the GCD. I reinstalled it using the spring bracket from the original air cooler that goes across the back of the core but with the eisblock screws. Tightened the screws until the bracket was touching the board on all four corners. Did not install the eisblock backplate.
Computer didn't post. Debug light for VGA problem. Sweating, backed off the four bracket screws by one full turn.
Computer boots, card now has an average temp in the mid 60s and a hotspot of about 85C. Raising the power limit to 380W It will now max the hotspot at 90C. Maintains ~2000mHz in the furmark test at these temperatures.

tl;dr, the eisblock for the 7900XTX reference is unable to apply sufficient pressure on the core. Attempting to tighten the four main screws further risks damage to the card. The 1Nm torque recommendation for those screws is absurd. Just use the spring bracket from your air cooler.
 

Lettuce

New member
My hotspot with the stock cooler on my xtx merc was 75c, I had it from 1st release and have never pulled it apart, I have a brutally aggressive fan curve though. Came here to decide what block/pads and whatnot to get, Thank you all for the detailed trials you've had to go through.
 

ckinfos

New member
I am experiencing the same bad thermals to a sapphire nitro + rx7900xt. the hot spot goes way up to 115c! even the 105c isnt supposed t ohappen with water cooling! i have tried repasting with thermal pads in 1.5mm (wrong size but had totest!). changes didnt happen. i used the correct ones then 1mm. The dgrres fell upto 113C. i have refilled the whole waterloop to be sure no air is blocking. but alas. as a water cooling on cpu and gpu it should have been down to at least 80celcius for the gpu and some 95-100 celius for the hot spot. i think something is not correct in the waterblock structure. the airblock i changed it due to lessen the sound of the fans. i think i go back to the air to test last time. please acool retest the waterblock. there seems to be either inadequate water passing in the block or too thin aluminium metal. the back side is ok with the given pads from your company. the card is really hot not on the gpu die but on the left side with the many r15 chokes. there seems to be the hotspot reading from.if the thermals seem better than the water cooling...pitty i wil lgo air and wont recommend the waterblock. let me mention i even changed pump to put a stronger flow...no. pump is on 100% speed.
 
Top