Zotac Trinity RTX 3090 Throttling at 72 degrees- should I check the thermal pads?

[GamerDad_]

Hey Guys,

 

I've had the card for about 7 months, I got lucky to find one used at MSRP of 1500. However, I never really dove deep into testing out through benchmarks. I was getting about what I expected out of games, maybe a bit less. But yesterday I started to benchmark while watching HWinfo and GPUz. My GPU never went over 72 degrees. I always thought that was pretty normal as I keep my fan curve agressive and keep my side panel off when gaming on my case. I had it overclocked by 100mhz this whole time on the core, and 1000 on the memory. 

 

I ran Time Spy and was getting about 15-20 % less than the average score on graphics, and overall for that matter. I thought this was a fluke. But multiple runs and multiple restarts, same thing. I never had RTSS on before, so I enabled that. I saw that my gpu once it hit 72 was throttling the clock down from 1850 or so all the way down to 1300 on some runs. I thought that was absurdly low, which it certainly is. But 72 is not hot, and that should not be happening. My memory seems to be getting hotter faster in the card. I thought this was my overclock, so i sent that back to 0. Same problem. Ran it at stock core clock. Same problem. 

 

Google searches turn up some instances of bad thermal pad and paste application. OK, but that surely has to be rare....right? What kind of shitty manufacture would take a halo flagship card and apply the damn paste or pads incorrectly on them? But I digress....I built my PC myself, and have become pretty comfortable with that. I have upgraded several parts, including my CPU and installed an AIO. I have ordered thermal pads, and they are here now. But opening up my 1500 dollar card is a bit daunting and nerve racking to be honest. It looks simple enough. But still, very nerve racking. 

 

My question is this; is there ANYTHING else I should consider before doing this? Does anyone else have any experience with the trinity 3090? I ordered both 1mm and 2 mm thermal pads, because I couldn't find solid information anywhere about which one I'll need. If I am on the right track, great...is there ANYTHING that anyone can tell me to avoid or be careful of? anything I need to do to it while it is opened up?

 

All help and tips are welcomed and appreciated. Sorry for the long story. I talk a lot when I'm nervous I guess.

 

[Bombastinator]

Such things don’t count just the card, and because it’s a 3090 it’s people who paid more for the card to get things like SLI so you may be looking at a skewed collection of test results.  Without knowing the rest of the rig there’s no telling what might be the actual issue though.  Could be some weak component or simply a misconfigured one, or maybe they’re all OCing their memory or something and you’re not.  That’s the problem with those competitive results things.  It’s all really open.  You may not WANT to be in that top group if you find out what it takes to be there.  It sounds to me like the thing is gaming fine to your expectations so it’s not like there is a massive problem.  There may be people willing to help you wring the last frame out of your machine though.

[coreygibson]

check what the "GPU Memory Junction Temperature" is in HWinfo before you do anything. If it get's to 110c the card will start throttling. a lot of 3090s have this problem when running heavy workloads

 

 

2 hours ago, Bombastinator said:

Such things don’t count just the card, and because it’s a 3090 it’s people who paid more for the card to get things like SLI so you may be looking at a skewed collection of test results.  Without knowing the rest of the rig there’s no telling what might be the actual issue though.  Could be some weak component or simply a misconfigured one, or maybe they’re all OCing their memory or something and you’re not.  That’s the problem with those competitive results things.  It’s all really open.  You may not WANT to be in that top group if you find out what it takes to be there.  It sounds to me like the thing is gaming fine to your expectations so it’s not like there is a massive problem.  There may be people willing to help you wring the last frame out of your machine though.

he's defiantly throttling, those clocks are too low

[Falkentyne]

4 hours ago, GamerDad_ said:

Hey Guys,

 

I've had the card for about 7 months, I got lucky to find one used at MSRP of 1500. However, I never really dove deep into testing out through benchmarks. I was getting about what I expected out of games, maybe a bit less. But yesterday I started to benchmark while watching HWinfo and GPUz. My GPU never went over 72 degrees. I always thought that was pretty normal as I keep my fan curve agressive and keep my side panel off when gaming on my case. I had it overclocked by 100mhz this whole time on the core, and 1000 on the memory. 

 

I ran Time Spy and was getting about 15-20 % less than the average score on graphics, and overall for that matter. I thought this was a fluke. But multiple runs and multiple restarts, same thing. I never had RTSS on before, so I enabled that. I saw that my gpu once it hit 72 was throttling the clock down from 1850 or so all the way down to 1300 on some runs. I thought that was absurdly low, which it certainly is. But 72 is not hot, and that should not be happening. My memory seems to be getting hotter faster in the card. I thought this was my overclock, so i sent that back to 0. Same problem. Ran it at stock core clock. Same problem. 

 

Google searches turn up some instances of bad thermal pad and paste application. OK, but that surely has to be rare....right? What kind of shitty manufacture would take a halo flagship card and apply the damn paste or pads incorrectly on them? But I digress....I built my PC myself, and have become pretty comfortable with that. I have upgraded several parts, including my CPU and installed an AIO. I have ordered thermal pads, and they are here now. But opening up my 1500 dollar card is a bit daunting and nerve racking to be honest. It looks simple enough. But still, very nerve racking. 

 

My question is this; is there ANYTHING else I should consider before doing this? Does anyone else have any experience with the trinity 3090? I ordered both 1mm and 2 mm thermal pads, because I couldn't find solid information anywhere about which one I'll need. If I am on the right track, great...is there ANYTHING that anyone can tell me to avoid or be careful of? anything I need to do to it while it is opened up?

 

All help and tips are welcomed and appreciated. Sorry for the long story. I talk a lot when I'm nervous I guess.

 

Check the very latest version of hwinfo64 (you can also install the current beta version of it but only over the latest full version) for the core, hotspot and memory junction temperature.

If you're reaching over 108C on memory it's going to throttle hard.

Also you can have GPU-Z open at the same time and look for any purple / magenta bar "Thermal" flags during this time.

If you do re-pad, make sure you have the exact measurements of the original pads, unless you already got that information from Zotac or another user.  There are often more than one thickness of pad used on one side of the card.

You can get a cheap digital micrometer (caliper) from Amazon for $10 to measure the originals also.

 

The best pads to replace on the front (core) side of the card are Gelid Extremes (Not Thermalright Odysseys nor Gelid Ultimates).  But these are not cheap pads so make sure your measurements are proper.  Another option that many people are doing successfully--even more so if you are NOT sure of the thickness of the old pads, is TG thermal Putty.

 

https://www.digikey.com/en/products/detail/t-global-technology/TG-PP10-50/6204863

 

[Bombastinator]

55 minutes ago, coreygibson said:

check what the "GPU Memory Junction Temperature" is in HWinfo before you do anything. If it get's to 110c the card will start throttling. a lot of 3090s have this problem when running heavy workloads

 

 

he's defiantly throttling, those clocks are too low

Throttling implies heat though.  I don’t know what max heat is for a 3090. There have been cards that needed to be below 75c.  I suppose it’s possible the 3090 is amongst them.  

[coreygibson]

3 minutes ago, Bombastinator said:

Throttling implies heat though.  I don’t know what max heat is for a 3090. There have been cards that needed to be below 75c.  I suppose it’s possible the 3090 is amongst them.  

There are other ways for a card to throttle than GPU core temps (and cpu's as well). Not getting enough power is one way. Another way, which is what is probably happening here, is the memory junction temperature. The 3090s are notorious for hot memory causing them to throttle, the gpu core temp not having any factor.

[Bombastinator]

21 minutes ago, coreygibson said:

There are other ways for a card to throttle than GPU core temps (and cpu's as well). Not getting enough power is one way. Another way, which is what is probably happening here, is the memory junction temperature. The 3090s are notorious for hot memory causing them to throttle, the gpu core temp not having any factor.

Not enough power often means no go rather than slow go. 3090s may be special.  The junction temperature IS heat.  I take it there is no junction temp probe or something?

[coreygibson]

7 minutes ago, Bombastinator said:

Not enough power often means no go rather than slow go.

Normally yes, especially with older hardware. the most common case of throttling due to a lack of power is a high-end cpu paired with a low end motherboard. But there are other components that will clock back due to lack of power to keep themselves alive.

7 minutes ago, Bombastinator said:

3090s may be special.  The junction temperature IS heat.  I take it there is no junction temp probe or something?

there is most new cards, most utilities wont read it though. That's why I told him to use HWinfo.

[GamerDad_]

Just wanted to update. First, thank you for your advice and opinions. I appreciate the time you took to respond. 

 

My card was not power throttling. HWinfo and GPUz showed thermal as my reasoning for performance cap. When you dip that low, something is definitely wrong. The GPU never went above 72. I checked so my software first to be sure that wasn't limiting me, which it wasn't.

 

I ordered both 1mm and 2mm thick, because I didn't know what the thickness was from any research. For anyone wondering it was 2mm for all of the pads inside. 

 

taking it apart was simple. ZOTAC was absoulte garbage at applying both the thermal paste and the pads. The quality of the pads were so bad they were crumbling apart when I tried to remove them. The thermal paste wasn't covering the entirety of the GPU, but pretty close. However it was very dry and crusted, hardly elastic at all.

 

I applied my thermal pads. Which was tedious and terrifying, and I ran out of 2mm pads so I had to use stacked 1mm for two of the rows of memory. I applied a generous helping of kryonaut paste to the chip. I kind of felt like the Verge...but hey I didn't want to risk anything and not have enough.

 

I did the memory pads on the back first, then screwed it back on and did the front, forgot to mention that. 

 

after over an hour of testing and benchmarking, my memory is down over 15 degrees. My GPU was getting close to 80 at peak, but it never throttled. I was staying at 1835-1865 on the core, with turbos going up to 2055. 

 

I no longer have thermal issues or throttling issues. In fact, I was able to get a better overclock stabilized to 125 Mhz core and 1050 9n the memory. Which is up about 25 and 50 respectively. 

 

I am seeing a performance improvement of 10-15% depending on the game or program, and my 3d Mark scores are in line.

 

further testing is still required as it's only been a few hours, but things look to be normal now. It's a huge relief. For anyone about to do it, it's not as bad as you think.

[coreygibson]

4 hours ago, GamerDad_ said:

 

after over an hour of testing and benchmarking, my memory is down over 15 degrees. My GPU was getting close to 80 at peak, but it never throttled. I was staying at 1835-1865 on the core, with turbos going up to 2055. 

 

Dont worry about your gpu temp going up. now that you have good thermal contact with the memory, the cooler has a bit more heat to get rid of.