RTX 3080 20GB Confirmed by Gigabyte?

I don't know how you think that it's a laughing stock seeing how the 3080, with it's laughing stock 10 gigs, is 90% faster than the 1080ti w/it's superior 11gigs at 4k.


I still don't get why people are saying 10 gigs isn't enough for 4K Gaming. I have a 2080 super with 8 gigs, and I am not getting any slow down in Doom Eternal at all Ultra Nightmare settings (yes, I get the same FPS with Nightmare textures as I do on Ultra Nightmare textures - I am running the latest 456.38 drivers though, which may account for the difference). I also don't experience any issues with any of my other games.

I'll record it on my damn phone soon if people keep saying that FUD.
 
these are just skus i wouldnt read much in to them

They are coming. The 10GB card was just to encourage good will to consumers with a low price after the 2080Ti price debacle. Nvidia wants people to buy $900 20GB cards not $700 10GB cards. If the highest end Big Navi has 16GB, then it's a no brainer that 20GB is set for a Oct/Nov release (not from the Nvidia storefront as the FE is a 10GB card only).
 
They are coming. The 10GB card was just to encourage good will to consumers with a low price after the 2080Ti price debacle. Nvidia wants people to buy $900 20GB cards not $700 10GB cards. If the highest end Big Navi has 16GB, then it's a no brainer that 20GB is set for a Oct/Nov release (not from the Nvidia storefront as the FE is a 10GB card only).
Well if RNDA2 16gb cards beat the 20gb 3080 cards, Nvidia would be pissing in the wind. If 20gb cards are so much better than the 10gb version then most will question what is wrong with the 10gb version for such a premium extra cost? For games it would be frankly overkill for the whole generation, a 12gb (the board has room for two more memory chips) would make a hell a lot more sense to me but then this is Nvidia.
 
Well if RNDA2 16gb cards beat the 20gb 3080 cards, Nvidia would be pissing in the wind. If 20gb cards are so much better than the 10gb version then most will question what is wrong with the 10gb version for such a premium extra cost? For games it would be frankly overkill for the whole generation, a 12gb (the board has room for two more memory chips) would make a hell a lot more sense to me but then this is Nvidia.

I don't think there will be many 10GB cards available after Big Navi and 20GB cards become available.
 
Have we not all accepted that the 10GB exists to hit that magical $699 price point? It’s still decent performance for that but with none available... $1199 20GB cards here we come!
 
Well if RNDA2 16gb cards beat the 20gb 3080 cards, Nvidia would be pissing in the wind. If 20gb cards are so much better than the 10gb version then most will question what is wrong with the 10gb version for such a premium extra cost? For games it would be frankly overkill for the whole generation, a 12gb (the board has room for two more memory chips) would make a hell a lot more sense to me but then this is Nvidia.

Not sure that 12GB is possible (unless the additional 2GB is only decoration...) as the memory size must be consistent with the bandwidth

Here is are a recap for NVIDIA
- 3090: 384-bit and 24GB of VRAM (12GB would be theoretically possible)
- 3080: 320-bit and 10GB of VRAM (20GB is possible)
- 2080 Ti: 352-bit and 11GB of VRAM
- 2080: 256-bit and 8GB of VRAM

My guess is that the 3080 Ti will use 352-bit of bandwidth, so the memory size should be either 11GB or 22GB (the latter seems much more likely)
 
For Flightsim and P3DV5 at 4K 10gb will be a little tight. Those are very unoptimized platforms.
 
Not sure that 12GB is possible (unless the additional 2GB is only decoration...) as the memory size must be consistent with the bandwidth

Here is are a recap for NVIDIA
- 3090: 384-bit and 24GB of VRAM (12GB would be theoretically possible)
- 3080: 320-bit and 10GB of VRAM (20GB is possible)
- 2080 Ti: 352-bit and 11GB of VRAM
- 2080: 256-bit and 8GB of VRAM

My guess is that the 3080 Ti will use 352-bit of bandwidth, so the memory size should be either 11GB or 22GB (the latter seems much more likely)
Just go with the 384bit bus, more usable bandwidth, 12gb and I would be more happy then a unnecessary amount of extra vram. 10gb is kinda pushing it for a long term card while 12gb will probably be OK for the next 3 years.
 
I think you guys are all talking about the Super revision, and what's shipping is what it is.

I'm having trouble with the idea a 2nd line of pcb's are ready to ship this gen when they haven't recouped whatever the cooler design and cutting down Ampere to consumer spec cost them.

Tensor and rtx cores will be upgraded.

I'd like a memory bandwidth step up, so the 3060 can be a constant problem at the low end.
 
Not sure that 12GB is possible (unless the additional 2GB is only decoration...) as the memory size must be consistent with the bandwidth

Here is are a recap for NVIDIA
- 3090: 384-bit and 24GB of VRAM (12GB would be theoretically possible)
- 3080: 320-bit and 10GB of VRAM (20GB is possible)
- 2080 Ti: 352-bit and 11GB of VRAM
- 2080: 256-bit and 8GB of VRAM

My guess is that the 3080 Ti will use 352-bit of bandwidth, so the memory size should be either 11GB or 22GB (the latter seems much more likely)

I feel like anything less than 16 GB and people would complain.
 
Have we not all accepted that the 10GB exists to hit that magical $699 price point? It’s still decent performance for that but with none available... $1199 20GB cards here we come!
I think you guys are all talking about the Super revision, and what's shipping is what it is.

I'm having trouble with the idea a 2nd line of pcb's are ready to ship this gen when they haven't recouped whatever the cooler design and cutting down Ampere to consumer spec cost them.

Tensor and rtx cores will be upgraded.

I'd like a memory bandwidth step up, so the 3060 can be a constant problem at the low end.

Here's your answer. 10GB is about hitting a price point. 20GB is coming sooner than you think.
 
I think you guys are all talking about the Super revision, and what's shipping is what it is.

I'm having trouble with the idea a 2nd line of pcb's are ready to ship this gen when they haven't recouped whatever the cooler design and cutting down Ampere to consumer spec cost them.

Tensor and rtx cores will be upgraded.

I'd like a memory bandwidth step up, so the 3060 can be a constant problem at the low end.

Oh I’m sure recouping cost isn’t very difficult when you only invested a tiny bit of money in very few units to begin with.

You’d have an argument if large numbers of people received the FE cards. They did not.
 
Last edited:
Oh I’m sure recouping cost isn’t very difficult when you only invested a tony bit of money in very few units to begin with.

You’d have an argument if large numbers of people received the FE cards. They did not.

Nvidia also isn't making FE 20GB cards. The FE is 10GB only. Nvidia doesn't have to design a 2nd PCB at all.
 
Back
Top