Vega owners club

[MarkJohnson]

I just dismantled it. Looks like the hbm contact is not good enough. I think thermal pads cause it. But i don't know if it needs thicker or thinner thermal pads.

Definitely thinner pad, the grease even looks thick on the outer support ring of the die. pads are too think and stopping contact, even when compressed.

 

[Caring1]

Yep, thinner pads would help.

 

[redatak]

Ok guys,problem solved. I put thinner pads 1.5mm+1mm(before was 2mm+1.5mm) coolermaster pro v2 paste,and i put some 2.5mm on the back between backplate and pcb. Also a non sharp knife i pushed some paste in the gap between hmb and gpu and i heat the whole thing with a hairdryer.
Before my vega was throttling(105c hotspot) in superposition 8k benchmark somewhere between 9-10th scene of the benchmark(total 17scenes). Just check my temps now with 52%fans and 280w power consumption. thanks everyone who help and spend time.

https://imgur.com/a/dpdBRBg

 

[delshay]

Yep, thinner pads would help.

That depends.

All thermal pads work better when compress. More compression "in the same space" better the transfer of heat. What you are actually doing is squashing the pad material so tightly together the heat will travel though it faster.
On most GFX card the space between the hot device & the cooling plate does not increase so thicker pads should work better as long as the gap between the device & cooling plate does not increase.

Read PDF doc on your thermal pads you have. You will find at a certain pressure the pads will work at they very best.

In-short, the more compact the material "in a given space", the better the transfer of heat.

 

[redatak]

That depends.

All thermal pads work better when compress. More compression "in the same space" better the transfer of heat. What you are actually doing is squashing the pad material so tightly together the heat will travel though it faster.
On most GFX card the space between the hot device & the cooling plate does not increase so thicker pads should work better as long as the gap between the device & cooling plate does not increase.

Read PDF doc on your thermal pads you have. You will find at a certain pressure the pads will work at they very best.

In-short, the more compact the material "in a given space", the better the transfer of heat.

thats true,before i had 6-8c better vrm temps but hotspot was 100+ when gaming heavy titles. so i didnt have choice.

 

[TheoneandonlyMrK]

That big a gap needs a thermal pad, not TIM.

It didn't come with one, why bother , it's hard enough returning it to as it was.
Just get thinner thermal pads and retry.

 

[Flyordie]

Well, after all this time.. I am still rocking my RX Vega64. -_- Does all I need it too and it still clocks above 1800Mhz on its core. I got $2500 in cash sittin in the bank though if AMD decides to surprise us with a liquid cooled version of the big Navi. If so, I'll still keep this Vega64 in service but relegate it to my bottom X16 slot.

 

[delshay]

thats true,before i had 6-8c better vrm temps but hotspot was 100+ when gaming heavy titles. so i didnt have choice.

Your improvement in cooling is more likely due to better quality of pad & not being thinner. Do a search & Read the PDF docs on your pads, new & old.

 

[redatak]

Your improvement in cooling is more likely due to better quality of pad & not being thinner. Do a search & Read the PDF docs on your pads, new & old.

I got same pads since the first time i opened the gpu. Thermal grizzly minus 8

 

[delshay]

I got same pads since the first time I opened the gpu. Thermal grizzly minus 8

Both my cards use Fujipoly 17w mk with extreme pressure (Maximum Force) for ultra fast heat transfer.

Read the PDF docs on this website to understand the difference what pressure does to all different types of thermal pads.

Fujipoly - Thermal Interface Materials - XR-Um

www.fujipoly.com

.

 

[TheoneandonlyMrK]

That depends.

All thermal pads work better when compress. More compression "in the same space" better the transfer of heat. What you are actually doing is squashing the pad material so tightly together the heat will travel though it faster.
On most GFX card the space between the hot device & the cooling plate does not increase so thicker pads should work better as long as the gap between the device & cooling plate does not increase.

Read PDF doc on your thermal pads you have. You will find at a certain pressure the pads will work at they very best.

In-short, the more compact the material "in a given space", the better the transfer of heat.

No the first set of replacement pads were too thick, simple.

 

[redatak]

No the first set of replacement pads were too thick, simple.

But you say the same thing. You say about gpu die temps he says about vrm temps. Both you are right but i had to sacrifice few degrees on vrm to calm down hotspot.

 

[TheoneandonlyMrK]

But you say the same thing. You say about gpu die temps he says about vrm temps. Both you are right but i had to sacrifice few degrees on vrm to calm down hotspot.

To be clear, you replaced the stock thermal pads with too thick a pad such that even with compression there was no pressure at all ie a gap, with the right thickness the hotspot has a heatsink attached to it now.

 

[redatak]

To be clear, you replaced the stock thermal pads with too thick a pad such that even with compression there was no pressure at all ie a gap, with the right thickness the hotspot has a heatsink attached to it now.

I measured Stock thermal pads was also 1.5mm+2mm but was ultra soft and compressed, tg minus8 1.5+2 was not so soft,then i replaced them with 1+1.5mm. so i suppose the pressure was not perfect because minis8 couldnt compressed as much as stock.

 

[delshay]

No the first set of replacement pads were too thick, simple.

Yes you can go overboard with too thick a pads but read what I said. As long as the gap between the hot device & the coldplate "does not increase". ..If you fit pads that are way too thick & you don't have the mounting force to compress it, then the distance between the hot device & the cold plate has gone up, which will result in worse performance.

There are a lot of different type of pads out there & they all require different PSI to get the best out of them. This why II said read the PDF Docs on the type of pads you are dealing with as they all compress differently.

 

[laitauriz]

Hi guys. I have an issue with my vega64 Nitro+ bought two years ago and a half. I OCed it to 1700 mhz /1140 mhz and played gladly with it. Max hotspot temp ( when power usage was near 350 w) was 87°/90 °maximum during gaming.
But now,since some 8/9 months, i noticed the hotspot temp was gradually ramping. To the point that today it reaches 108 ° max . I tried to up the fans but it does nothing. It's purely link to power usage. If i go over 300 W it starts to ramp up like crazy. The HBM and core temp have not changed at all since the beginning and are still 60° ish at max.

Any idea of what to do? I clean dust regularly on the card, may have been doing shit with my duster blower? Or does it just need thermal paste replacement ?

Thanks for your advices

 

[redatak]

Hi guys. I have an issue with my vega64 Nitro+ bought two years ago and a half. I OCed it to 1700 mhz /1140 mhz and played gladly with it. Max hotspot temp ( when power usage was near 350 w) was 87°/90 °maximum during gaming.
But now,since some 8/9 months, i noticed the hotspot temp was gradually ramping. To the point that today it reaches 108 ° max . I tried to up the fans but it does nothing. It's purely link to power usage. If i go over 300 W it starts to ramp up like crazy. The HBM and core temp have not changed at all since the beginning and are still 60° ish at max.

Any idea of what to do? I clean dust regularly on the card, may have been doing shit with my duster blower? Or does it just need thermal paste replacement ?

Thanks for your advices

300w+ is too much. How much voltage?

 

[TheoneandonlyMrK]

Yes you can go overboard with too thick a pads but read what I said. As long as the gap between the hot device & the coldplate "does not increase". ..If you fit pads that are way too thick & you don't have the mounting force to compress it, then the distance between the hot device & the cold plate has gone up, which will result in worse performance.

There are a lot of different type of pads out there & they all require different PSI to get the best out of them. This why II said read the PDF Docs on the type of pads you are dealing with as they all compress differently.

Did you look at his pictures, the Gap increased.

I'm not arguing against your other points ,all true but you can only apply pressure with the parts your given on your card , increasing it is tricky and beyond what the guy with the issue needed to do.

Hi guys. I have an issue with my vega64 Nitro+ bought two years ago and a half. I OCed it to 1700 mhz /1140 mhz and played gladly with it. Max hotspot temp ( when power usage was near 350 w) was 87°/90 °maximum during gaming.
But now,since some 8/9 months, i noticed the hotspot temp was gradually ramping. To the point that today it reaches 108 ° max . I tried to up the fans but it does nothing. It's purely link to power usage. If i go over 300 W it starts to ramp up like crazy. The HBM and core temp have not changed at all since the beginning and are still 60° ish at max.

Any idea of what to do? I clean dust regularly on the card, may have been doing shit with my duster blower? Or does it just need thermal paste replacement ?

Thanks for your advices

You have probably dried out, to a degree the thermal paste and pads, I would try replacing those first.

 

[laitauriz]

300w+ is too much. How much voltage?

1120 mv for p7 .

Did you look at his pictures, the Gap increased.

I'm not arguing against your other points ,all true but you can only apply pressure with the parts your given on your card , increasing it is tricky and beyond what the guy with the issue needed to do.

You have probably dried out, to a degree the thermal paste and pads, I would try replacing those first.

Okay i will try that. But its that not weird that only the hotspot is affected andnot the hbm2 or GPU core temp ?

 

[TheoneandonlyMrK]

1120 mv for p7 .




Okay i will try that. But its that not weird that only the hotspot is affected andnot the hbm2 or GPU core temp ?

Something has to dry out, f up first, mine needed replacing for the same thing , with waterblockes due to 24/7 folding.

 

[dragontamer5788]

So I love programming my Vega64. Except last night, I discovered that its not doing hipMemcpy correctly for sizes larger than 1GB. I'm wondering if there's a PCIe bug in my system somehow, or if my Vega64 is somehow defective. I rented an instance from GPUEater.com, and the GPUEater.com instance of Vega56 runs my code correctly (same ROCm drivers and everything).

This is a bad time to figure out a hardware error as a AMD GPU-programming fan. Navi doesn't work in ROCm reliably yet, while Navi2x isn't released yet.

It could be the motherboard though (ASRock Taichi x399), because hipMemcpy is a PCIe 3.0 traversal. Or it could be the card itself. I don't think its my CPU or RAM, because those parts have survived a suite of other tests already.

--------------

I probably can continue to make small test programs on my home computer, then rent out GPUEater.com instances whenever I'm doing a "production run", since they charge per-second of instance time. So its not a big deal to spin up instances for a few minutes and shut them down.

 

[Flyordie]

Did you look at his pictures, the Gap increased.

I'm not arguing against your other points ,all true but you can only apply pressure with the parts your given on your card , increasing it is tricky and beyond what the guy with the issue needed to do.

You have probably dried out, to a degree the thermal paste and pads, I would try replacing those first.

I have worried about that as my card gets older but it appears, based on temps.. my card hasn't budged outside of margin of error since it was new on either. VRMs have never been above 70C. Hot Spot Temp has never been above 80C. Must be Saph used cheap pads and paste?

 

[redatak]

I have worried about that as my card gets older but it appears, based on temps.. my card hasn't budged outside of margin of error since it was new on either. VRMs have never been above 70C. Hot Spot Temp has never been above 80C. Must be Saph used cheap pads and paste?

Nah... Just your overclock is very aggressive

 

[TheoneandonlyMrK]

I have worried about that as my card gets older but it appears, based on temps.. my card hasn't budged outside of margin of error since it was new on either. VRMs have never been above 70C. Hot Spot Temp has never been above 80C. Must be Saph used cheap pads and paste?

It's easy to assume similar use cases but you shouldn't, my card has had a very hard life, very hard for a consumer card, Sapphire are very good but being a reference AMD spec'd it not them.

 

[Flyordie]

It's easy to assume similar use cases but you shouldn't, my card has had a very hard life, very hard for a consumer card, Sapphire are very good but being a reference AMD spec'd it not them.

Its been running F@H for over 3/4 its life. (I only stop during the summer months.. Late May-Mid-Sept.) Rest of the time it folds away all day, every day. Even while gaming in some cases. I can play World of Warships @ 2K and be 61FPS+ and it never have an issue.

What I am saying is- With the Sapphire card being a premium aftermarket card, you'd think they would go 100% premium everything. So seeing a premium card suffer that, it worries me that it could happen down the line to mine being a "Built by AMD" card and not a "Powered by AMD" card.

My guess is the same as yours though, the thermal paste has dried on his card. I've had it happen on my old Radeon 5770 (2009). I just hope his is the molded die so its safer to pull apart and reapply a good paste.