RTX 3090 FE - Maximum Junction Temperature (TJ Max or Tjunction Max) @ 106C Too much? (2024)

Noticed (without overclocks) that my Memory Temperature hits 106C while randomly gaming, is this enough to be concerned with on a FE - RTX 3090 card?

"104C is still crazy, crazy hot. This article from September sheds some light on what an acceptable/safe temperature might be:"

For reference, Micron rates its GDDR5, GDDR5X and GDDR6 memory chips with a Maximum Junction Temperature (TJ Max or Tjunction Max) of 100C (degrees Celsius). The typical recommended operating temperatures ranges between 0C to 95C. The reading materials on Micron's GDDR6X don't reveal the TJ Max for the new memory chips so there's still a bit of mystery to the topic. According to Igor, the general consent is around 120C before the GDDR6X chips suffer damage. This would mean that the Tjunction value should be set at 105C or up to 110C.

Click to expand...

---

put a noctua fan ont he back of the card and the memory still hot:

Many of the rtx cards were pumped out of production at such speeds that QC was almost non existent so many issues with thermal pads and such have come to light if your ok with doing so below is a link to jayztwocents recent video on fixing the issue.

the card will throttle at 110 C... i think high memory temps are pretty normal..

trog

ThaiTaffy said:

Many of the rtx cards were pumped out of production at such speeds that QC was almost non existent so many issues with thermal pads and such have come to light if your ok with doing so below is a link to jayztwocents recent video on fixing the issue.

Click to expand...

Thanks for posting this, I was going to post the same video.

@erek, watch the video

eidairaman1 said:

Thanks for posting this, I was going to post the same video.

@erek, watch the video

Click to expand...

i understand those and seen the video before, but i want to know if the temp is actually typical and expected for FE cards or if my card is way outside the range?

erek said:

See Also

World’s Biggest Ether Mining Firm Turns Off Servers After MergeBuying a Bitcoin Mining Rig? What You Need to Know | River Learn - Bitcoin MiningReasons to be cheerful: 'GPU mining is dead less than 24 hours after the merge'geforce-rtx-3090-8k-hdr-gaming

i understand those and seen the video before, but i want to know if the temp is actually typical and expected for FE cards or if my card is way outside the range?

Click to expand...

As jay stated in the video nvidia's use of their sub standard white thermal pads is a joke your card is overheating like the majority of all the 3080ti's and 3090's with ram chips on the rear of the card. No consumer electronics should run over 100°c.

erek said:

i understand those and seen the video before, but i want to know if the temp is actually typical and expected for FE cards or if my card is way outside the range?

Click to expand...

Every 3000 series FE model is absolute rubbish as far as VRAM cooling goes. So yes, it's within the expected range of being terrible.

Vya Domus said:

Every 3000 series FE model is absolute rubbish as far as VRAM cooling goes. So yes, it's within the expected range of being terrible.

Click to expand...

That's why ekwb seems to have focused all their new blocks on Fe cards I guess.

erek said:

i understand those and seen the video before, but i want to know if the temp is actually typical and expected for FE cards or if my card is way outside the range?

Click to expand...

Yes it's too hot.

You readily vaporize water over 100°C (212°F)

Do the mods in the video and press on

erek said:

Noticed (without overclocks) that my Memory Temperature hits 106C while randomly gaming, is this enough to be concerned with on a FE - RTX 3090 card?

"104C is still crazy, crazy hot. This article from September sheds some light on what an acceptable/safe temperature might be:"

---

put a noctua fan ont he back of the card and the memory still hot:

View attachment 219305

View attachment 219306

Click to expand...

You need to repad to fix that.

Thermalright TFX (full spread) on the core.
Follow this template for Gelid extremes for the front side. (one pack 1.5mm, one pack 2mm)

Backside, use Gelid Extreme OR Gelid ultimate 1.5mm (you'll need two packs).

sad

Here's an ideal use case for justifying a waterblock. Any decent waterblock with a standard backplate will keep junction temps below 90c.

My Gigabyte 3090, stays easily at 70-80 °C on gaming, barely using 4GB of VRAM. But on deep learning with all VRAM employed and it becomes an oven getting near 90-110 °C after 30-90 minutes of training c: I've read some post on s1 reducing the temps by using fans and heatsinks on the backplate... I should try this.

Edit: My VRAM temps are usally about 10 to 20 °C higher than the GPU temp itself, according to OHM.

I believe the actually rating is up to 105 °C for GDDR6X You can find it on Micron website. Personally over a 100 °C is a real concern for longevity as I noted in my RTX 3070 Ti, RTX 3080 and RTX 3080 Ti in the review coverage I have done.

Slap a waterblock on it, or fix the thermal pads yourself. Who knows how long it will last. GDDR6X has built in ECC so small errors will not pop up or crash the system, but it will give you a performance dip. if the game is memory bandwidth hungry.

ir_cow said:

I believe the actually rating is up to 105 °C for GDDR6X You can find it on Micron website. Personally over a 100 °C is a real concern for longevity as I noted in my RTX 3070 Ti, RTX 3080 and RTX 3080 Ti in the review coverage I have done.

Click to expand...

The temp rating for GDDR6X is Tcase, while the sensor is reporting Tjunction. Tjunction is around 15C hotter, that means GDDR6X can operate at full speed up to 110C Tjunction (correspond to 95C Tcase), passing 110C Tjunction it will thermal throttle and risk of degradation at 120C (Tjunction)

nguyen said:

The temp rating for GDDR6X is Tcase, while the sensor is reporting Tjunction. Tjunction is around 15C hotter, that means GDDR6X can operate at full speed up to 110C Tjunction (correspond to 95C Tcase), passing 110C Tjunction it will thermal throttle and risk of degradation at 120C (Tjunction)

Click to expand...

Hmm the date sheet suggests 95°C Tcase and 105 °C for the Tj. That would be 10°C not 15°C above.

ir_cow said:

Hmm the date sheet suggests 95°C Tcase and 105 °C for the Tj. That would be 10°C not 15°C above.

Click to expand...

Data sheet is quoting Tcase for both temp, 95C before thermal throttle and 105C before degradation (110C/120C for Tjunction)

Nvidia set 110C Tjunction as the thermal throttling point in BIOS, at 104C Tjunction the fans are commanded to max speed I think.

But yeah I would rather not see VRAM reach boiling temp at all, who knows the BGA on VRAM might just fail due to high thermal swing (happened to Turing already)

Simpleris said:

Could it be possible to underclock(/undervolt?) GPU memory so it reduces temps on it? has anybody tried ever?

Click to expand...

Yes under intensive workload that stress the memory (DL or mining), it's better to underclock the memory, reducing the power limit and increasing fan speeds as much as you can handle help too.

ThaiTaffy said:

As jay stated in the video nvidia's use of their sub standard white thermal pads is a joke your card is overheating like the majority of all the 3080ti's and 3090's with ram chips on the rear of the card. No consumer electronics should run over 100°c.

Click to expand...

@ThaiTaffy

so if i just replace the pads on the back that will be a big help?

erek said:

@ThaiTaffy

so if i just replace the pads on the back that will be a big help?

Click to expand...

It's become such a problem with these cards there has been an influx of aftermarket backplates now, personally I would take a look, Many can be had extremely cheap compared to the cost of the cards.

Replace the junk Nvidia pads for gelid or some other quality product and slap some form of cooling backplate on it should cost you less than $60 for the lot which is nothing if you had to replace the whole card if it went up in smoke.

ThaiTaffy said:

It's become such a problem with these cards there has been an influx of aftermarket backplates now, personally I would take a look, Many can be had extremely cheap compared to the cost of the cards.

Replace the junk Nvidia pads for gelid or some other quality product and slap some form of cooling backplate on it should cost you less than $60 for the lot which is nothing if you had to replace the whole card if it went up in smoke.

Click to expand...

i got it down to a max of 100C so far just be repadding the back VRAM only, re-used all the other thermal pads.

still way too hot, or within acceptable operating parameters at least? (dead of winter though)

erek said:

i got it down to a max of 100C so far just be repadding the back VRAM only, re-used all the other thermal pads.

View attachment 236856

still way too hot, or within acceptable operating parameters at least? (dead of winter though)

Click to expand...

Been 17 months since 3090 launched and I haven't heard of any large scale VRAM related failure despite many many people mine on their 3090, so i assume you would be fine with VRAM running at 100C.

Though if you want your 3090 to live 5+ years then maybe try to keep VRAM temp below 90C.

nguyen said:

Been 17 months since 3090 launched and I haven't heard of any large scale VRAM related failure despite many many people mine on their 3090, so i assume you would be fine with VRAM running at 100C.

Though if you want your 3090 to live 5+ years then maybe try to keep VRAM temp below 90C.

Click to expand...

5 years would be real extreme, I usually upgrade within 3, so we’ll see. Was incredibly hard to acquire this FE. Lucked out at MSRP via another forum