NVIDIA, this is NOT ok!

[jakkuh_t]

 

[SavageNeo]

ok

[WhitetailAni]

I didn't know NVidia made GPU waterblocks.

Where do I buy one?

[Marinecoco]

I have had a 2080 TI for over a year do I need to change out the thermal paste on my GPU?

[SimplyChunk]

10 minutes ago, Marinecoco said:

I have had a 2080 TI for over a year do I need to change out the thermal paste on my GPU?

personally i'd wait until warranty is finished

[Craftyawesome]

Something that was glossed over is that the 95C is Tc, or case temperature (see the datasheet linked down this page). Junction temp is basically a hotspot temp and will be hotter.

 

As for what junction temp is safe, it's a little less clear.

 

Micron themselves have 100C Tj Max for GDDR5(X) and G6. Most seem to have a max Tcase of 95C, but G6X is not listed here. www.micron.com/-/media/client/global/documents/products/technical-note/dram/tn0008_thermal_apps.pdf

 

As mentioned in the video, Igor's Lab suggests 105C is a safe limit. This is also the limit of the G6 in the 5700xt.

 

And obviously, Nvidia is willing to sell you a card that throttles at 110C.

[AnonymousGuy]

I just modified a Bykski backplate with a DIMM waterblock.  Works great.

 

 

Also in typical EK fashion...that backplate they're using is hilariously thin and shitty compared to the Bykski one.  Going to warp like a mo-fo when you go to actually tighten it and it's not going to compress thermal pads.  And wouldn't ya know it, mine does a better job than theirs too:

 

 

This is running at unlimited power (about 700W from the wall) while mining.

[TO_Techy]

<pedantry rant>

"The electronic components in your PC release nearly 100% of the electricity they consume as heat."

The only component that releases as heat 100% (nearly) of the electricity it consumes is a resistive heater. If the electronics in your computer do this, it wouldn't be a computer; it'd be a space heater.

Heat is an undesirable waste product of "work", even a 90% loss of energy to heat would be absurd. I believe only 5-10% of the energy consumed by RAM results in heat, the other 90-95% going towards the actual "work" of flipping bit, or transiting gates.

</pedantry rant>

 

 

[riklaunim]

A Fancard with reversed fans and placed so that it blows on the backplate cloud also work to some extent - or dropping the backplate and using sticky mini radiators on each ram chip so that that fancard airflow is more efficient.

[Grabhanem]

2 hours ago, TO_Techy said:

<pedantry rant>

"The electronic components in your PC release nearly 100% of the electricity they consume as heat."

The only component that releases as heat 100% (nearly) of the electricity it consumes is a resistive heater. If the electronics in your computer do this, it wouldn't be a computer; it'd be a space heater.

Heat is an undesirable waste product of "work", even a 90% loss of energy to heat would be absurd. I believe only 5-10% of the energy consumed by RAM results in heat, the other 90-95% going towards the actual "work" of flipping bit, or transiting gates.

</pedantry rant>

 

 

<pedantry rant>

No. 100% of the electricity (neglecting your RGB lights) going into your computer comes out as heat. Information does not constitute "work"; the only reason it takes power to flip gates is because you have to overcome the electrical resistance of the silicon, which is what produces heat.

 

Imagine you built an opaque box around your computer, space heater, or other appliance. Without knowing what's happening in that box, every joule of energy going into that box must come out of that box in some form or be stored by the box, by conservation of energy. Since between the time you power your computer on and off it doesn't change state, it must be releasing that energy, either as light, heat, or some other form. (Your computer does technically store energy in things like capacitors, inductors, and fans, but it doesn't store it permanently - all the energy stored in a spinning fan will eventually be returned to heat by friction, etc.)

</pedantry rant>

[mon1ka]

6 hours ago, SavageNeo said:

ok

ok

5 hours ago, YouSirAreADudeSir said:

personally i'd wait until warranty is finished

depends on how long is left on the warranty IMO.

[Luscious]

"Cannot fit in the top slot LOL"

 

Can we all agree on the BACKWARDNESS of the ATX spec. This now 36 year old massively outdated and antiquated PC standard is what is holding innovation, development and progress back. <<<-- video topic Linus???

 

Because in the good old dusty wild west untamed pioneering days of Windows 95 you didn't have TRIPLE SLOT 350W GPU's nor did the single core processors even remotely need the type of cooling and power delivery today's 20+ core silicon behemoths require.

 

Why the heck can't they come up with a new ATX standard that addresses these issues??? It's just a redesign of the motherboard with more sensible placement of the slots, CPU VRM, and CPU socket that would be required, Many mobos already sacrifice that top PCI slot with zero effect on trace length - important for GPU performance. An extra 40mm of space at the top of the board beyond the IO back plate would allow better VRM placement for CPU's. It would also allow for better RAM slot placement in many scenarios where space on the mobo is otherwise sacrificed for other components.

 

Some case/chassis retooling may be required, but given the current sad state of affairs, the PC space NEEDS a reinvigorating anyway.

 

In the server space this is a non-issue, mostly because there they only need to adhere to a chassis height (1/2/3/4U) and a 17 inch width requirement. It simply works - partly because there's no ATX standard, and also because a lot of it is proprietary. That's how they can get away with FOUR CPU sockets and 32+ RAM slots on a single motherboard, or piggyback a daughterboard that takes 20 GPU's. And even there, water cooling is becoming more common with workarounds that get 1U radiators and fans in there.

 

As for the EK GPU backplate itself, I can see it being useful in very limited cases. This would be great for a four-way GPU build IF that top slot wasn't blocking components near the CPU or RAM slots. But even in that case with adequate cooling and with your loads under control I doubt this would help. A chunkier rad in push/pull would give a more tangible improvement in C degrees that would transfer to whatever is on the back of that single card. Given that you cannot even get your hands on the cards right now, it's kind of comical this option for cooling is even available.

[DZSlasher]

This video definitely got me worried about the cooling on my RTX 3080 FE. According to HWInfo, my memory got to 94C at it highest in a Enthoo Evolv ATX TG case. Making me really consider what the temps would be if I put it into an ITX case which would have few fans for cooling. 

[Helpful Tech Witch]

6 hours ago, Luscious said:

 

Can we all agree on the BACKWARDNESS of the ATX spec. This now 36 year old massively outdated and antiquated PC standard is what is holding innovation, development and progress back. <<<-- video topic Linus???

i relly dont see what needs to be innovates in the motherboard space. 

[Stahlmann]

7 hours ago, DZSlasher said:

This video definitely got me worried about the cooling on my RTX 3080 FE. According to HWInfo, my memory got to 94C at it highest in a Enthoo Evolv ATX TG case. Making me really consider what the temps would be if I put it into an ITX case which would have few fans for cooling. 

While 94°C is hot, it technically is in spec and you shouldn't see throttling or premature failure.

[Stahlmann]

Another quick hint: If anyone plans to move to a waterblock because you want to reduce a 3080's VRAM temps, stay away from the Alphacool block. The included thermal pads are too thin and don't proplery contact the coldplate. Get another waterblock if you want one. This issue came up to a few people on this forum already (including me) after the HWInfo64 update that shows VRAM temps.

[SFFDesigns]

10 hours ago, DZSlasher said:

Making me really consider what the temps would be if I put it into an ITX case which would have few fans for cooling. 

The same mostly. 96c in Ghost S1 w/o any additional case fans as well as choked flow through fan. Would be the same or better in a case that's better optimized for  flow through layout.

 

ITX cases have similar or better cooling perf in most cases haha, unless you add like 12 fans to an ATX case.

[ElementalCyclone]

Regarding this . . . is there any kind or form of GPU backplate that doubles as heatsink ? i'm imagining the one that looks like chipset heatsink in the old (Pentium, Pre/at Sandy Bridge era) desktop motherboard, that square heatsink with that not-so-dense fins.

I think with some space-constrained setup, a watercooling wouldn't fit the space

[Helpful Tech Witch]

26 minutes ago, James Evens said:

There is one thing: flip the GPU cooler orientation.

All of those company design nice looking cooler with RGB but all you see is the backplate.

 

From the feature set: Cheap notebooks/mainboards with build in remote access. Population get's older so the demand for remote access/support grows.

I think that first one should be on the gpu makers. The whole design would be flipped. You would have to the the motherboard on the other side of the case.

[RejZoR]

Checked my RTX 3080 and memory was at 81°C after playing Borderlands 3 for quite some time. My card does have metal backplate and it has thermal pads underneath contacting PCB where memory modules are. It's actually so hot you can't hold it for more than 2 seconds. Never had graphic card that soaks so much heat through backplate.