Vega 56 Throtling after changing thermal paste and thermal pads

[LizardTa]

As the title suggests I am having problems with my Vega 56 red dragon after changing out the old thermal pads and paste for better quality ones. I have applied Thermal grizzly konductonaut to the HBM and chip dies and used minus 8 pads on the VRM components.

However the card now throttles extremely heavly as soon as I put it under any work load and on some occasions crashes. The temps of the chip and HBM are fine. High 50,s to low 60,s on the core and 70 ish on the HBM. The VRM temps looks fine too as observed through HWinfo.

However as I said it throttles quickly when put under load reducing quickly from 1550 to around 1400 and then further reducing as time goes on. The power draw sits at only 160W under load when it used to draw 240W when undervolted so the card is restricting power but I can't see why looking at the temps.

The only thing I can see wrong is the Hot spot temperature jumps to over 100 very quickly when the card goes to work. And I have seen it hit 110 which seems ridiculously high. Could this be throttling card? I have no idea where the sensor is so I am having trouble working out the problem area.

I have screwed the cooler back on as hard as I can get it and all of the components that require direct cooling from the thermal pads are making good contact so I don't understand what's going on.

Can anyone advise a solution or is this a case of a damaged component during the job do we think? (Which I hope to god it isn't)

Cheers Frank

 

[eidairaman1]

I hope you read the warning on the conductonaut about corrosion

 

[LizardTa]

Sorry I meant Kryonaught not Conductonaught.

 

[DeathtoGnomes]

I remember a review that there was a height difference somewhere and that it can cause your problem. You may or may not have proper contact or you over tighten the cooler or you used the wrong thermal pads too thin or too thick. Maybe left the backing on one of them too.

 

[LizardTa]

I have defiantly taken the backing off the thermal pads I know that much and as for the thickness I matched like for like so they should be good. I might have a go at reduce the clamp torque though see if that helps. I feel as if this is a contact issue but as far as I can see when I remove the heat sink the TP spread and coverage looks good.

 

[Dinnercore]

You write that you screwed it back on as hard as you could? That could be a little bit too much and bend the pcb to hard so that some areas have not enough contact. I put a waterblock on vega and the instructions explicitly mentioned to never tighten down with full force, however a slight bend is normal.

The hot-spot temp is on the back of the card right behind the package. I think the sensor is right in or on the pcb with power planes beneath it. It shoots very high, tightly connected to the power draw of the card. It´s hard to cool there since that part has no direct contact to anything. I heard some airflow over the back of the card can help with that.

EDIT: I think I remember that the max. temp or throttle point for hot-spot was at 125°C.

 

[eidairaman1]

I remember a review that there was a height difference somewhere and that it can cause your problem. You may or may not have proper contact or you over tighten the cooler or you used the wrong thermal pads too thin or too thick. Maybe left the backing on one of them too.

Yes some of some cards the diet self was molded or unmolded if he happens to have on that is not molded then he's got the one with the height Gap

 

[LizardTa]

You write that you screwed it back on as hard as you could? That could be a little bit too much and bend the pcb to hard so that some areas have not enough contact. I put a waterblock on vega and the instructions explicitly mentioned to never tighten down with full force, however a slight bend is normal.

The hot-spot temp is on the back of the card right behind the package. I think the sensor is right in or on the pcb with power planes beneath it. It shoots very high, tightly connected to the power draw of the card. It´s hard to cool there since that part has no direct contact to anything. I heard some airflow over the back of the card can help with that.

EDIT: I think I remember that the max. temp or throttle point for hot-spot was at 125°C.

In that case I am beginning to think this may be that the tension bracket is over tightened. I will loosen it off and see if that helps. If the Hot stop temp is 125 for throttling then that's not the issue because 110 is the absolute highest I saw it go and it normally sits at about 103.

 

[dgianstefani]

Sell the 56 buy a 1170 or 1080, problem solved.

 

[Eric3988]

Were you having throttling issues at all before the repaste? If not, why not try to repaste and remount again? I suspect operator error if your issues emerged only after you did something to the card. If you had some issues before and that's why you repasted, then maybe you just have a poor performing card. Might try an RMA if that's the case.

 

[TheoneandonlyMrK]

Be more generous with your tim next time ,i'd retim personally it's seams ckear and reasonable that you have an un timed patch betwixt Soc and Hsf.
Gamersnexus debunked the pea adequately.

I have been stuck below 50° on mine since binning the fan for a Wb.

 

[LizardTa]

Were you having throttling issues at all before the repaste? If not, why not try to repaste and remount again? I suspect operator error if your issues emerged only after you did something to the card. If you had some issues before and that's why you repasted, then maybe you just have a poor performing card. Might try an RMA if that's the case.

Yeah I have come to the conclusion that I have not re-mounted the cooler properly and I have improper contact between the contact plate and the chip. I have had to order some new thermal pads and will give it another shot when they arrive.

 

[DeathtoGnomes]

Yeah I have come to the conclusion that I have not re-mounted the cooler properly and I have improper contact between the contact plate and the chip. I have had to order some new thermal pads and will give it another shot when they arrive.

keep us updated.

 

[silkstone]

Yeah I have come to the conclusion that I have not re-mounted the cooler properly and I have improper contact between the contact plate and the chip. I have had to order some new thermal pads and will give it another shot when they arrive.

Subbing to see how this turns out. I once tried replacing thermal pads on my 7870XT with much (supposedly) better ones. The temps shot up by quite a few degrees! Luckily, I'd kept the stock pads and could just switch them back, which I did, and temps went down. This wasn't with any sort of special full coverage heatsink, just the standard long aluminium strip with pegs on each end. Since then, I don't trust those thermal pads to be any better than the ones that come with the card. TIM, on the other hand, I've only ever seen temps come down upon replacement.

 

[LizardTa]

Subbing to see how this turns out. I once tried replacing thermal pads on my 7870XT with much (supposedly) better ones. The temps shot up by quite a few degrees! Luckily, I'd kept the stock pads and could just switch them back, which I did, and temps went down. This wasn't with any sort of special full coverage heatsink, just the standard long aluminium strip with pegs on each end. Since then, I don't trust those thermal pads to be any better than the ones that come with the card. TIM, on the other hand, I've only ever seen temps come down upon replacement.

Yeah they are coming tomorrow the new pads. I have for with the same Thermal grizzly minus 8 pads. I will see what's happens. I am beginning to think I might just have to get it as good as I can. Been a bit of a disasters all round this endeavour.

 

[LizardTa]

Right ok I have had another crack at 're seating the cooler and have tried a different technique that involves tighenting the tension bracket up first then securing the rest of the screws for the back plate.

This seems to have helped with the boost clocks now holding at around 1480 to 1500 but it's still not back to where it was. I believe that if I continue to 're seat the cooler I will be able to get the proper position and be able to get back to where I was at.

The moral of this story for me is that unless you are removing the stick cooler to install a water block then it really isbt worth the hassle of removing a blown heatsink design just to change thermal pads and paste. Mileage may vary for other of course.

 

[TheoneandonlyMrK]

Right ok I have had another crack at 're seating the cooler and have tried a different technique that involves tighenting the tension bracket up first then securing the rest of the screws for the back plate.

This seems to have helped with the boost clocks now holding at around 1480 to 1500 but it's still not back to where it was. I believe that if I continue to 're seat the cooler I will be able to get the proper position and be able to get back to where I was at.

The moral of this story for me is that unless you are removing the stick cooler to install a water block then it really isbt worth the hassle of removing a blown heatsink design just to change thermal pads and paste. Mileage may vary for other of course.

You also learnt something everyone should know ,if it's a option, always clamp the main die first ie gpu not vrms or heatsink but the best advice is to fit a few turns ,all the screws you can loose ,then tighten in diagonals inside to out and gpu first across the whole card, I've rebuilt a countless amount this way.

 

[GorbazTheDragon]

I've always repasted my GPUs and laptops but I've never really had any problems like this.

With the first time doing some part it is always good practice to do a test mount just to understand how the contact is and also check how the thing fits together.

 

[Apocalypsee]

Just wanted to add to this, I got a used Vega56 reference yesterday and thinking of repasted it since the temps are over 75C even under light load like GPU-Z 3D test. Upon inspection the previous owner did repasted it like he said but the contact is sub-par, it doesn't make full contact with the die or HBM at all, its tilted to the HBM side more than the die itself.

The card I have is unmolded ones. The best thing to do for stock cooler is first to dismantle everything and install the VRM heatsink first but DON'T tighten them first, just screw them to align the heatsink into place. Then you insert the main heatsink/vapor chamber to the hole and wiggle it around till all the 4 mounting pins are through the holes on the PCB. Only then you must first tighten the main heatsink. After that then you tighten all VRM heatsink screw. It is a bit of hassle to get the main heatsink to align properly.

And from unmolded dies, the HBM temps are always higher, mine is about 3-5C higher under load. Some from reddit got a lot higher like 10C difference. I didn't measure the hot spot area though but its about 10C higher I think.

 

[LizardTa]

Yeah thanks for the input on this.

I am about to undertake a 're paste and realignment again this weekend to see if I can get it right.

One of my issues seems to be that the origanal thermal pads that came with the cards seem a lot more soft than the ones I have replaced them with it. I don't know if that is potentially impacting my VRM temps.

I also suspect that I haven't used enough thermal paste on the CPU HBM as when I first took it off it was really glooped on so gonna try that.

Also the last time I reapplied the heatsink I tigented the tension bracket tight and this also helped a lot. I will report back (yet again) after I have done this over the weekend.

 

[kastriot]

Well i hope you will get card to previous temp states

 

[LizardTa]

Hello again!

It's been a while since I last posted in this thread but I am happy to say there has been a positive development in this story!

I have lost count of the number of time I have taken this card to bits now and reapplied the Thermal Grizzly Kryonaut compound and was beginning to think I had somehow wrapped the board indefinatly by mistake.

However I happened to go on the website that I got the paste off as I was running out of it and was thinking about changing my CPU paste out. I noticed it had a 1 star review so I clicked on it. And there it was! The exact same story as mine.

The guy said he couldn't spread it and that it spread thick which wasn't right. I just thought that's the way it was suppose to be like as it's a lot newer than any TIM I have ever used (Been using Artic Silver for years). His review was my exact experience.

So I bought a tube of Master Gel Maker and thought I will give it one last go. And what do you know! As soon as I turned it on idle was 7c less and now I can boost clock when OC runs back at 1580 no problem! Actually using less watts and genetating less heat than before which is nice as that was the origanal goal!.

So there you have it. Looks a bad batch of Thermal Grizzly kryonaut Knocking around out there! Don't know why I didn't think to try this sooner but there you go!

Cheers for everyone's input it's been much appriated! Now I am off to finally finish Far Cry 5 on my 4k TV .

 

[RejZoR]

You generally cannot overtighten GPU coolers as they have a fixed screw point and all the pressure is done by the springs on the screws. You can tighten them to the max and all you do is tighten the screws as far as they can go. There is still gap in between and springs ensure the required pressure. I don't think Vega is any different despite slightly different construction (GPU on an interposer opposed to being directly on PCB).

Theoretically, if pads are too thick and those mentioned springs don't apply enough pressure to fully squeeze them, preventing cooler from contacting GPU die properly...

 

[Deleted member 178884]

Sell the 56 buy a 1170 or 1080, problem solved.

Or fix it? That's what a retard would do.

 

[Vayra86]

Sorry I meant Kryonaught not Conductonaught.

Kryo-naught wouldn't do much good either

Sorry, I couldn't resist.