Hot News? More Like... HWInfo64 Now Shows Hot VRAM for RTX 3080/3090!

[AldiPrayogi]

And there's me uncomfortable pushing my 1660Ti to anywhere higher than low 70C lol

[Mling]

We need twin tower coolers for our GPUs

[Craftyawesome]

4 hours ago, linuxChips2600 said:

tbh I'm not comfortable running any of my electronics above 85 C for an extended period of time.  But I'd say that given your results posted you should not even consider running at power targets any higher than ~65%.

Keep in mind that as stated in my OP the minimum max operating temperature of GDDR6X memory chips is 95 C, so although Tjunction may be set at 110 C by Nvidia drivers for now, I would strongly caution against running the memory on an RTX 3080/3090 at 95 C or higher for any period of time, as once the transistors fail it's game over unless if you are tech and business savvy enough to find replacement GDDR6X memory chips and do the actual replacement yourself...

FWIW, the measured temperatures are junction temps (hot spots). The general advice for temps in the 80s being hot is not for junction temps. This also caused some confusion with Navi. "Operating at up to 110C Junction Temperature during typical gaming usage is expected and within spec."

 

I definitely agree about exercising at least some caution for something running more often and harder than typical gaming.

 

 

Edit:

 

Let's do some more comparisons to the 5700xt:

 

From the author of HWiNFO:

"I just clarified this with AMD and the situation is following.
This value on Navi GPUs is the memory temperature, but it's not the case temperature (T_case) that can be measured on the surface of the chip. It's the junction temperature (Tj) provided by the memory sensor.
Tj is always higher and the difference seen from Tcase can be up-to 18 C (according to measurements done by AMD). The maximum value of Tj here is 105 C.
So users that were concerned about seeing too high temperatures here should not worry about elevated values and not treat this as the surface temperature (which has a limit of 95 C)."

 

According to pcgamesn, the micron memory used is the 14 Gbps GDDR6 MT61K256M32.

https://www.micron.com/products/ultra-bandwidth-solutions/gddr6/part-catalog/mt61k256m32je-14

https://media-www.micron.com/-/media/client/global/documents/products/data-sheet/dram/gddr/gddr6/gddr6_sgram_8gb_brief.pdf?rev=cf98a21f67f64c33ba0fc7572213b50f

 

These both list a max temperature of 95 C, which is also what Martin referenced. The datasheet also specifies T_C, which should be T_case and not junction temperature

 

I wouldn't be too worried as a typical gamer.

Edited February 1, 2021 by Craftyawesome

[linuxChips2600]

2 hours ago, Craftyawesome said:

I definitely agree about exercising at least some caution for something running more often and harder than typical gaming.

This

The law of large numbers means that although one is unlikely to become extremely unlucky with getting crappy GDDR6X modules for their RTX 3080/3090, given that such cards are around USD 1K each (give or take a couple hundred bucks), I do not think that (unless if one absolutely doesn't care about accidentally trashing at least hundreds of dollars worth of VRAM chips) it is worth risking bricking a card's VRAM modules in the event that one receives a card with "still-in-spec" but abnormally low-quality GDDR6X chips.  Hope you follow.

[cocobeex]

Just trying to contribute some info about mining 

I am on Nicehash 

This is one of my 3080 Mining Rigs with 6x EVGA RTX 3080 FTW3 ULTRA,  

Core -150 

Mem +1200   (one of the card couldn't oc much, and one of the card hashrate jumps around if i am using it remotely.) 

TDP varies, from 62% to 80% 

MEM  Temp 90 to 96c  

 

Here are some screenshots 

 

[linuxChips2600]

4 hours ago, AldiPrayogi said:

And there's me uncomfortable pushing my 1660Ti to anywhere higher than low 70C lol

*Laughs in PowerColor Red Dragon Vega 56 hitting 95 C junction temps at +40% power limit + -25 mV undervolt + 3 - 8% overclocked "clock-boost-curve" under OpenCL stress tests*

[Moonzy]

11 hours ago, Nacht said:

Graphics cards aren't designed to mine on 24/7 try gaming on it.

Graphics cards aren't meant just for gaming either

 

The design should be capable of handling full load under any circumstances.

[Massey_Ferguson]

29 minutes ago, Moonzy said:

Graphics cards aren't meant just for gaming either

 

The design should be capable of handling full load under any circumstances.

And I think it's quite irresponsible from developer to create cards which run up to 110 C memory temp when mining (which many of these cards end up doing anyways, like it or not) without giving any warning about it and while the fan rolls at 50% cause the GPU temperature is at low 50C.

[Moonzy]

1 minute ago, Massey_Ferguson said:

And I think it's quite irresponsible from developer to create cards which run up to 110 C memory temp when mining (which many of these cards end up doing anyways, like it or not) without giving any warning about it and while the fan rolls at 50% cause the GPU temperature is at low 50C.

My MSI gaming x trio fan actually spun up to 100% with core temp in the 50s, I'm assuming it's to cool the VRAM

[linuxChips2600]

20 minutes ago, Massey_Ferguson said:

while the fan rolls at 50% cause the GPU temperature is at low 50C.

That honestly might be a VBIOS-specific or custom fan-curve issue (not to imply that it was necessarily your fault); as similar to what @Moonzy said, in my OP where I linked the Classical Technology vid, the narrator/host describes one of the first signs that something was wrong with his founder cards was that the fans were ramping up to 100% even when the core was nowhere near thermal throttling.

[linuxChips2600]

10 hours ago, linuxChips2600 said:

Does anyone know for sure that GDDR6X is a Micron-exclusive product (since it appears to be the case at least to me just from Micron's website)?

According to https://en.wikipedia.org/wiki/GDDR6_SDRAM#GDDR6X it does indeed look like GDDR6X is a Micron-exclusive tech/product, which makes this whole situation just that much worse.  I'll start a separate discussion soon under the "Graphics Cards" sub-forum for compiling a list of various 3080/3090 cards based on their max VRAM temp running a Blender Optix vs an Ethereum mining benchmark...

[Massey_Ferguson]

9 minutes ago, linuxChips2600 said:

That honestly might be a VBIOS-specific or custom fan-curve issue (not to imply that it was necessarily your fault); as similar to what @Moonzy said, in my OP where I linked the Classical Technology vid, the narrator/host describes one of the first signs that something was wrong with his founder cards was that the fans were ramping up to 100% even when the core was nowhere near thermal throttling.

Might of been my fault. I use MSI Afterburner for controlling the settings including fan speed. Not sure if that overrides the cards normal fan settings when left on "auto".

[linuxChips2600]

2 minutes ago, Massey_Ferguson said:

I use MSI Afterburner for controlling the settings including fan speed. Not sure if that overrides the cards normal fan settings when left on "auto".

Honestly I won't be surprised; as a tool which is basically supposed to support multiple generations of AMD and Nvidia graphics cards (including the latest-gen ones), it's not by any means a trivial task to ensure proper compatibility with each of the aforementioned cards.

I suppose that if you accidentally brick your VRAM using MSI Afterburner, you could christen your copy of Afterburner "MSI VRAM-burner"...

[Massey_Ferguson]

8 minutes ago, linuxChips2600 said:

I suppose that if you accidentally brick your VRAM using MSI Afterburner, you could christen your copy of Afterburner "MSI VRAM-burner"...

The book which i hold my hand on when swearing that "under no circumstances I have tampered with or misused this card" after returning my burned card to warranty.

[Moonzy]

25 minutes ago, Massey_Ferguson said:

Might of been my fault. I use MSI Afterburner for controlling the settings including fan speed. Not sure if that overrides the cards normal fan settings when left on "auto".

I use msi afterburner as well for my tuning, the fan ramps up to 100% on auto

 

Funny thing is that the fan ramps up... Trying to cool the VRAM on the back of the PCB for my 3090

I added a 120mm fan blowing on the back, did not bring down the temps but it doesn't throttle anymore at least

[Massey_Ferguson]

12 minutes ago, Moonzy said:

I use msi afterburner as well for my tuning, the fan ramps up to 100% on auto

 

Funny thing is that the fan ramps up... Trying to cool the VRAM on the back of the PCB for my 3090

I added a 120mm fan blowing on the back, did not bring down the temps but it doesn't throttle anymore at least

I was thinking about trying the same. Wonder if ramming some heatsinks on the backplate would help.

 

Looking at the youtube video where they take my card apart, it looks like the memory is connected to the backplate with heat tape.

 

Anyways, I don't want to open the card up to break warranty.

[HenrySalayne]

3 hours ago, Moonzy said:

The design should be capable of handling full load under any circumstances.

It's always a compromise. If you want "full load any circumstance" you'll get a product at least 25% worse in every aspect. A graphics card designed for gaming will be tuned to excel in gaming. Something like Furmark or mining is not the designed workload. If the VRAM gets to hot, it might be the wrong tool for the job.

[Moonzy]

29 minutes ago, HenrySalayne said:

It's always a compromise. If you want "full load any circumstance" you'll get a product at least 25% worse in every aspect.

how?

I don't mind it costing more

 

if they wanna be cheap about it, put a disclaimer that the it will overheat and throttle if it's fully loaded.

 

the spec says it's capable of 19GB/s, it better be able to achieve that

 

32 minutes ago, HenrySalayne said:

A graphics card designed for gaming will be tuned to excel in gaming. Something like Furmark or mining is not the designed workload. If the VRAM gets to hot, it might be the wrong tool for the job.

in other words, let's design this thing to be the bare minimum needed to do its task

[HenrySalayne]

5 minutes ago, Moonzy said:

in other words, let's design this thing to be the bare minimum needed to do its task

I wouldn't call it bare minimum, but every product has a purpose. RAM would be terrible to store your holiday pictures for the next 10 years. And additional cooling for VRAM under the backplate might not be necessary for gaming. If mining stresses the VRAM considerably more, you need to either reduce the power output (reducing performance) or change the cooling system. Boost clocks and thermal throttling are the same basic principle. The only difference is how it is communicated to the costumer.

[Moonzy]

5 minutes ago, HenrySalayne said:

If mining stresses the VRAM considerably more, you need to either reduce the power output (reducing performance) or change the cooling system.

if it's more than the listed spec, sure

but it wont go above the listed spec unless you specify it to, and my GPU is overheating on stock settings, so basically on spec

 

you dont say "my hard disk can run 7400rpm" but it'll overheat after 5 minutes and spin down to 5400, then blame customer for doing continuous file transfer for more than 5 minutes because it's out of the average file transfer of few seconds.

 

also, i wish we can control VRAM voltage/power, that would indeed be useful, but afaik, we can't

 

7 minutes ago, HenrySalayne said:

Boost clocks and thermal throttling are the same basic principle.

VRAM dont have boost clock, afaik

[HenrySalayne]

31 minutes ago, Moonzy said:

you dont say "my hard disk can run 7400rpm" but it'll overheat after 5 minutes and spin down to 5400, then blame customer for doing continuous file transfer for more than 5 minutes because it's out of the average file transfer of few seconds.

This is actually a pretty bad example for your cause. There are specific HDD models for continuous or tightly packed use. SSDs are the same: sequential speeds are not at all continuous speeds.

35 minutes ago, Moonzy said:

but it wont go above the listed spec unless you specify it to, and my GPU is overheating on stock settings, so basically on spec

 

While gaming or while doing something else?
If you use your graphics card for the advertised purpose (which is gaming for all 3000 models - despite the VRAM or price) and it overheats nevertheless, it is badly designed. In this case, I fully agree (if your case supplies an adequate amount of fresh air).

But if it overheats while mining, tough luck. The choice to put some VRAM onto the rear of the PCB is questionable, but as long as it works for the intended purpose, it's fine. That's all I'm trying to say.

[Moonzy]

1 minute ago, HenrySalayne said:

But if it overheats while mining, tough luck. The choice to put some VRAM onto the rear of the PCB is questionable, but as long as it works for the intended purpose, it's fine. That's all I'm trying to say.

if you advertise it to be 19Gbps capable, it should  be able to handle it without throttling

not just advertise it as 19Gbps but design around 8 or 9Gbps that games only needs, because that would be false advertising imo

[HenrySalayne]

2 minutes ago, Moonzy said:

if you advertise it to be 19Gbps capable, it should  be able to handle it without throttling

How long does it have to reach this speed? Forever? 1 s? 1/10 s? How long does your CPU hits its maximum boost clock while doing AVX workloads? Exactly.
 

[Moonzy]

21 minutes ago, HenrySalayne said:

How long does it have to reach this speed? Forever? 1 s? 1/10 s? How long does your CPU hits its maximum boost clock while doing AVX workloads? Exactly.

you're comparing boost clock to a rated operating speed

 

a better comparison would be intel rating their cpu at 4.0ghz base, then includes a cooler that thermal throttles it to 3ghz under all core load

but it does go to 4ghz on light use.

except this time you cant change the cooler, so you dont have any other choice

 

edit: I still think that a product should be able to perform as spec-ed under any load, not interested to debate about this further

[RejZoR]

I'm gonna try and place some fans on top of backplate to see if it affects VRAM temps since backplate is contacting PCB on other side of VRAMs. If there will be a measurable effect I might do something funky and stick some chipset heatsinks on top of backplate where it's contacting VRAMs.