techpowerup The Reason Why NVIDIA's GeForce RTX 3080 GPU Uses 19 Gbps GDDR6X Memory and not Faster Variants

[axipher]

Quote

When NVIDIA announced its next-generation GeForce RTX 3080 and 3090 Ampere GPUs, it specified that the memory found in the new GPUs will be Micron's GDDR6X variant with 19 Gbps speed. However, being that there are faster GDDR6X modules already available in a 21 Gbps variant, everyone was left wondering why NVIDIA didn't just use the faster memory from Micron. That is exactly what Igor's Lab, a technology website, has been wondering as well. They have decided to conduct testing with an infrared camera that measures the heat produced. To check out the full testing setup and how they tested everything, you can go here and read it, including watching the video embedded.

Micron chips like GDDR5, GDDR5X, and GDDR6 are rated for the maximum junction temperature (TJ Max) of 100 degrees Celsius. It is recommended that these chips should run anywhere from 0C to 95C for the best results. However, when it comes to the new GDDR6X modules found in the new graphics cards, they are not yet any official specifications available to the public. Igor's Lab estimates that they can reach 120C before they become damaged, meaning that TJ Max should be 110C or 105C. When measuring the temperature of GDDR6X modules, Igor found out that the hottest chip ran at 104C, meaning that the chips are running pretty close to the TJ Max they are (supposedly) specified. It is NVIDIA's PCB design decisions that are leading up to this, as the hottest chips are running next to voltage regulators, which can get pretty hot on their own.

Source

 

It is a beautiful PCB, but I was a little skeptical myself on the density of the components, but I don't really have the means to test it, let alone get myself a 3080, I did try and was promptly met with broken web stores.

 

This is to be expected when you cram that many components in to a small PCB.  Hopefully this doesn't mean any long term damage of pre-mature failures of DRAM chips for the Founder's Cards.

[Bastiaan_NL]

That is some hot memory
Although I love the design, that is another reason not to go with it.
 

[axipher]

If AMD doesn't do something for H264 Encoding on Navi2, I will likely be looking at a 3080 for the better NVENC encoder for H264, definitely not a FE card though.

[ENTERPRISE]

Blimey, 104c, yeah that is hot. I love the PCB and the fact it is MUCH smaller than I thought it would be prior to us seeing them, I was expecting a really long PCB. As already mentioned though, cramming those components in will have a cost BUT assuming sufficient cooling, this should be fine. After all this is the FE edition and while the cooler is improved, it will not do as well as the AIB solutions.

[axipher]

2 hours ago, ENTERPRISE said:

Blimey, 104c, yeah that is hot. I love the PCB and the fact it is MUCH smaller than I thought it would be prior to us seeing them, I was expecting a really long PCB. As already mentioned though, cramming those components in will have a cost BUT assuming sufficient cooling, this should be fine. After all this is the FE edition and while the cooler is improved, it will not do as well as the AIB solutions.

From a few of the AIB solutions I've seen though, some are choosing a less filled out VRM solution compared to the FE.

 

Hopefully they have chosen RM components that will each be able to handle the increased load at full power and we don't have cards with failing power stages early in the life.

 

We have all this wonderful tech on the card, but with NVENC making progress, and CUDA + RT cores running while in games, there's a good chance that gaming loads might actually be the really high power stuff that maxes out a cards power stage this generation as more AAA games are coming out with good RT support and all the new gamers that want to stream from the same PC they are gaming with.

 

Going to be really interesting to see card temps for 3000 series in a few months.

[ENTERPRISE]

20 minutes ago, axipher said:

From a few of the AIB solutions I've seen though, some are choosing a less filled out VRM solution compared to the FE.

 

Hopefully they have chosen RM components that will each be able to handle the increased load at full power and we don't have cards with failing power stages early in the life.

 

We have all this wonderful tech on the card, but with NVENC making progress, and CUDA + RT cores running while in games, there's a good chance that gaming loads might actually be the really high power stuff that maxes out a cards power stage this generation as more AAA games are coming out with good RT support and all the new gamers that want to stream from the same PC they are gaming with.

 

Going to be really interesting to see card temps for 3000 series in a few months.

Yeah, I mean we hopefully will not see the same issues as we did with the 2000 series memory going faulty very early on, the same applies for the VRM's etc, hopefully they will be able to have a suitably long life. With everything though....time will tell I guess. Watercooling FTW I guess.

[axipher]

17 minutes ago, ENTERPRISE said:

Yeah, I mean we hopefully will not see the same issues as we did with the 2000 series memory going faulty very early on, the same applies for the VRM's etc, hopefully they will be able to have a suitably long life. With everything though....time will tell I guess. Watercooling FTW I guess.

I've been kind of happy with my Sapphire Pulse 5700 as the fans on it stay off most of the time, even while watching Twitch or YouTube and the Decode Engine is being used, fans normally stay off so my computer is near silent with just a few Gentle Typhoons on a 360mm AIO.

Going to enjoy going back to a full look though, but not sure if a 360mm will be enough for let's say a Ryzen 7 5800XT (future part speculation) and a 3080 under water.  Might need to either get a 360mm thick boi rad, or a dual 90mm radiator to put on the rear of the case (4U rackmount case only has 3x 120mm on the front and 2x 90mm on the back).

[ENTERPRISE]

3 hours ago, axipher said:

I've been kind of happy with my Sapphire Pulse 5700 as the fans on it stay off most of the time, even while watching Twitch or YouTube and the Decode Engine is being used, fans normally stay off so my computer is near silent with just a few Gentle Typhoons on a 360mm AIO.

Going to enjoy going back to a full look though, but not sure if a 360mm will be enough for let's say a Ryzen 7 5800XT (future part speculation) and a 3080 under water.  Might need to either get a 360mm thick boi rad, or a dual 90mm radiator to put on the rear of the case (4U rackmount case only has 3x 120mm on the front and 2x 90mm on the back).

Well for my WC setup I have Hardware Labs 360mm Fat boy RAD as well as a standard thickness 280mm RAD to the top of my case. This cools an AMD 3950X and will also cool the upcoming GPU (3080/3090). I have 2 D5 pumps as well, mainly for safety should one fail.

[axipher]

13 minutes ago, ENTERPRISE said:

Well for my WC setup I have Hardware Labs 360mm Fat boy RAD as well as a standard thickness 280mm RAD to the top of my case. This cools an AMD 3950X and will also cool the upcoming GPU (3080/3090). I have 2 D5 pumps as well, mainly for safety should one fail.

I might have to look at trying to source a Hardware Labs Black Ice SR2, I definitely have the free space in my PC case, the Fractal Design Celsius' CPU Pump Block had previously ran the included 360mm RAD + CPU Block + RX480 GPU block, I don't think the fatboy SR2 would provide significant more flow resistance so I think it might be fine.

[ENTERPRISE]

1 hour ago, axipher said:

I might have to look at trying to source a Hardware Labs Black Ice SR2, I definitely have the free space in my PC case, the Fractal Design Celsius' CPU Pump Block had previously ran the included 360mm RAD + CPU Block + RX480 GPU block, I don't think the fatboy SR2 would provide significant more flow resistance so I think it might be fine.

When it comes to RADS I cannot recommend the Hardware Labs kit enough. I believe my 360mm RAD is the Black Ice GTR. Bulky as hell, but it definitely gets the job done. Infact it is so bulky, including my res (which is mounted to the RAD) it actually limits me from having super log GPU's.

[axipher]

13 minutes ago, ENTERPRISE said:

When it comes to RADS I cannot recommend the Hardware Labs kit enough. I believe my 360mm RAD is the Black Ice GTR. Bulky as hell, but it definitely gets the job done. Infact it is so bulky, including my res (which is mounted to the RAD) it actually limits me from having super log GPU's.

Thanks for the glowing recommendation.

Aside form my cable mess from my temporary ThermalTake non-modular PSU, I have plenty of space on the front of my 4U case for a thicker rad and even push-pull if I needed to.

 

Edited September 23, 2020 by axipher

[ENTERPRISE]

On 9/23/2020 at 6:39 PM, axipher said:

Thanks for the glowing recommendation.

Aside form my cable mess from my temporary ThermalTake non-modular PSU, I have plenty of space on the front of my 4U case for a thicker rad and even push-pull if I needed to.

 

 

 

Yeah having a case that can take the thicker rads really helps. It would not be a problem in my case BUT as I wanted the res mounted in front and upright for accessibility, it does limit the length of add in cards but not by a huge amount. Always a balancing act...well not unless you have dual chamber case where you have a dedicated section for all the WC Gear. CaseLabs were great for that. 

[asunderw]

On 9/23/2020 at 7:23 PM, axipher said:

I might have to look at trying to source a Hardware Labs Black Ice SR2, I definitely have the free space in my PC case, the Fractal Design Celsius' CPU Pump Block had previously ran the included 360mm RAD + CPU Block + RX480 GPU block, I don't think the fatboy SR2 would provide significant more flow resistance so I think it might be fine.

Kind of off-topic (somewhat), but I have been wondering: how do you mount a SR2? I have a GTX and I would like moving that up top and getting a SR2 in the front, but the SR2's port/reservoir block sticks out massively due to the plugs (can't recall how they're called...indentations?) so it seems impossible to mount it flush. How do you deal with that on a flat surface like typical 120/140mm rails?