[SOLVED] GIGABYTE Vega 64, advice/teardown change thermal paste

Zeyke62

Prominent
Oct 25, 2019
10
0
510
Hello everyone

I buying recently a new PC with a GIGABYTE Vega 64. After 7 years with my old friend GTX 660 TI so I'm not too familiar with high-end GPUs.
Im already aware of the bad review of this GPU but for 250 bucks this was a pretty good deal, and surprisingly it's not as bad as I expected, It still looks like a new old stock.

In stock setting with no UV/OC (Maybe later)
In idle (Freq.Proc/Freq.Mem 27mhz/167mhz 3W) with 900 rpm fan the temp was pretty low 25-26c
Looking a video (Freq.Proc/Freq.Mem 55-80mhz/167-500mhz 4-8W) with 900 rpm fan the temp 26-28c

In games that are not very demanding like Code Vein with vsync at 75 FPS (Freq.Proc/Freq.Mem 1080mhz/800mhz 45-80w)
with 900 rpm the temp was pretty reasonable (a guess ?) 44-50c

In games that are really demanding like the witcher 3 highest setting with no vsync or locker FPS (Freq.Proc/Freq.Mem 1580-1630mhz/945mhz 215-244W)

But the temp seem pretty high to me in this scenario and I can hear coil whine when the GPU goes up to more than 100 watts but it's quite acceptable but the higher the GPU goes up in the watts the more horrible the coil whine is but is not the problem and that don't bother me that much.

The problem (or not?) GPU temp rise to 80-85c it seem pretty high to me and the thermal paste doesn't seem to be doing its job properly (?) with 1700-1850 rpm Fan the sound is pretty reasonable i can rise up the fan to 3500-4000 rpm max but the sound is too loud.

So I intended to dismantle the GPU and change the thermal paste with grizzly kryonaut for that I looked on the internet for a tutorial to dismantle a GIGABYTE vega 64 but without success, I could still see that the GPU has thermal pads and I don't intend to go that far and change them but I'm scared of Removed the thermal pads which seems quite fragile or something else since I have nothing to refer.

hence my post here to look for people who have this GPU and know the best way to access the processor without too much mishandling the thermal pads or the GPU itself, if it's a good idea to do this or if I should leave it like this and not take any risks

thank you to the people who will take the time to read and help me
(Sorry for my bad english)
 
Last edited by a moderator:
Solution
If the GPU is causing your other components to show signs of their own thermal throttling, you need to increase airflow.

Locking the FPS will certainly help.
Lowering the settings on its own will not, the card will just produce more frames.
In conjunction with locking the FPS, yes, it will reduce card load.

Power draw is directly related to the amount of processing to be done, yes.

I'm afraid this is not that new of a technology. Power saving really took off with the Energy Star program where limits on standby power draw and idle power draw were codified. Since then pretty much every electronic product has had this. The difference today is that the end user has direct access through drivers and applications, which has been around...

Eximo

Titan
Ambassador
GPUs are designed to go as fast as possible. They will thermally control themselves. 85C is a good sign. If it is hovering around there and the clock speeds are still relatively high then it is doing its job. A cool GPU suggests it is not being utilized. An overly hot GPU will crash or throttle and be very noticeable that it isn't performing.

As for actually re-pasting. That is complicated. Vega 56 and 64 have HBM memory modules in close proximity to the GPU die, sometimes they are uneven and there were even two main design runs/manufacturers that produced slightly different results.

Check Gamer's Nexus teardowns and review videos for the relevant information on the subject. They were quite thorough in their investigation.

Vega 64 only got a bad reputation for its price/performance compared to the competition. Otherwise it is a fine card. Lacks much room for overclocking improvement, which wasn't super helpful in its adoption by enthusiasts.
 

Zeyke62

Prominent
Oct 25, 2019
10
0
510
Thank you for your answer i appreciate it

So about 85c is pretty normal at full power?
The problem is that 85c in a small but well ventilated box still makes all the components inside heated well.

So to avoid reaching such a temperature I avoid pushing the GPU to full power by lock my FPS and lowering the settings a little.

The advantage (at least I don't know if it's normal) is that the GPU seems to use the power it needs to run the game in good condition (if I decrease the settings and lock the FPS the GPU will detect the necessary power and decrease the frequencies and consumption, sorry if this was obvious to you to me it's pretty new and damn welcome in this case) when the parameters in WATTMAN are in automatic.
I don't know if this will be the case if I touch the parameters?
If I want for example UV/OC by changing the settings so by disabling the automatic settings of WATTMAN, will the GPU run all the time at full power in games and no longer be able to automatically detect the necessary power required and ajust it automatically? (or i didn't understand how this technology works and also how WATTMAN works which is definitely possible)


My other problem is that I really feel like I have to at least change the thermal pastes since the GPU looks like an old stock (a musty smell when I opened the box and on the GPU) which makes me think that the thermal pastes are dry and no longer give everything they have.

But on the other hand, the performance is not that bad except in full power, hence the dilemma, I'm really afraid to dismantle my GPU to damage the thermal pads or anything else and see the temperature increase from before, I read on the forums, messages from people who had dismantled their GPU to precisely change the thermal paste get temperature increases after that and performance decreases, so I get a little paranoid

(in the end the temperatures are really honest when I don't push the GPU to hard but I really feel like it can be really better if I just change the thermal paste for a better and new one)

i'm pretty new and naive with this technology i'm really sorry if i ask somes awkwards and obvious question
 

Eximo

Titan
Ambassador
If the GPU is causing your other components to show signs of their own thermal throttling, you need to increase airflow.

Locking the FPS will certainly help.
Lowering the settings on its own will not, the card will just produce more frames.
In conjunction with locking the FPS, yes, it will reduce card load.

Power draw is directly related to the amount of processing to be done, yes.

I'm afraid this is not that new of a technology. Power saving really took off with the Energy Star program where limits on standby power draw and idle power draw were codified. Since then pretty much every electronic product has had this. The difference today is that the end user has direct access through drivers and applications, which has been around for about ten years now.

Again, I would look at Gamer's Nexus teardown videos for Vega 56 and 64. They went into detail and what they could do to make improvements and what could go wrong. I'm just saying it isn't quite like a standard GPU paste re-do. Easy to chip exposed silicon either way, so care must be taken. Thermal pads are somewhat reuseable, just try to not get your skin oils on them.

I haven't really seen your explanation for bad performance at full load. Is it overheating and throttling and causing stutters or something, or you just don't like that the fans are ramped up to full speed to keep it cool?

I'm not super familiar with Wattman, but I do know that undervolting Vega often leads to better results. And most gains come from overclocking the memory. Clock speed is basically already maxed out, which is one of the reasons the power draw is as high as it is.
 
Solution
85ºC is right around the temp where it will start to throttle heavy. Around 80ºC max is fine. If you plan to OC you may want a better aftermarket cooler. After I put an aftermarket cooler on my Sapphire Vega 64, it performs as good as a heavily OCd GTX 1080. Even better in some cases (at 1440p). https://forums.tomshardware.com/threads/vega-64-issue.3552567/#post-21455631

I haven't seen a teardown of Gigabyte's version of the Vega64. It is advisable to replace the thermal pads (if there are any). These pads are fragile and conform to the chips they are up against once used. They can re-conform but the fit probably won't be as good as it was originally. Just like thermal paste, you shouldn't reuse thermal pads.
 
  • Like
Reactions: long john silver

TJ Hooker

Titan
Ambassador
So about 85c is pretty normal at full power?
The problem is that 85c in a small but well ventilated box still makes all the components inside heated well.

So to avoid reaching such a temperature I avoid pushing the GPU to full power by lock my FPS and lowering the settings a little.
If you're worried about it heating up your case, it's power draw you want to be concerned about, not GPU temperature. Whether it's at 85C or you repaste and get it down to 75C, if it's drawing 250W it'll have the same effect on the rest of your components.

However, the things you've been doing to reduce to reduce GPU temperature will also reduce power draw. I'd also recommend looking into undervolting, as mentioned above.

https://www.tomshardware.com/reviews/amd-radeon-rx-vega-64-undervolting,5234.html