And just like that out of nowhere, in comes the leaks about the GeForce RTX 3090 Ti...
ZOTAC's new FireStorm OC software update includes support for some of NVIDIA's unreleased graphics cards, including the GeForce RTX 3090 Ti, GeForce RTX 3080 Ti, and the GeForce RTX 3070 Ti. VideoCardz reported the story sourcing Japanese PC Mania, adding that they have not heard anything about the GeForce RTX 3090 Ti yet, and neither have I.
The references to the new GeForce RTX 3090 Ti graphics card were found inside of the latest FireStorm inside of its own "resources" directory. In there, you'll find the RTX 30 series Ti logos, created on April 21 -- ZOTAC has the GeForce RTX 3090 Ti, GeForce RTX 3080 Ti, and GeForce RTX 3070 Ti listed in FireStorm.
What we do know about the GeForce RTX 3090 Ti: The new NVIDIA GeForce RTX 3090 Ti, if it is indeed real, would probably use the GA100 GPU -- the full-fat Ampere GPU that sits above the GA102 that powers the current flagship GeForce RTX 3090, the GeForce RTX 3080, and the upcoming GeForce RTX 3080 Ti.
GA100 is a behemoth of a GPU, with a die size of 826mm2 and packs 54.2 billion transistors, and made on TSMC's 7nm process. But the thing is, that GA100 doesn't support DirectX -- so it's not for gamers, with NVIDIA aiming its powerful GA100 GPU at professionals and GPU compute applications.
- Read more: NVIDIA taps Samsung again for new Ampere GPU production on 8nm in 2021
- Read more: NVIDIA will shift over to TSMC for new 7nm Ampere GPUs in 2021
A tweaked GA100 could be a big deal inside of a new flagship GeForce RTX 3090 Ti graphics card, acting like the new TITAN RTX -- if the GeForce RTX 3090 wasn't already TITAN RTX enough with its 24GB of GDDR6X memory -- and battle whatever GPU beast AMD is building in its labs with the RDNA 3 architecture coming in 2022.
NVIDIA's new GeForce RTX 3090 Ti would most likely pack a monstrous 48GB of ultra-fast GDDR6X memory. Yeah, 48 gigabytes of freaking VRAM.
We also know that NVIDIA is shifting over to TSMC -- away from Samsung where it co-developed its 8nm node for the Ampere GPUs (everything other than GA100 is made at Samsung on 8nm, while GA100 is made at TSMC on 7nm) for its new Ampere GPUs in 2021. That back in October 2020, and received more confirmation of this news since.
A couple of industry sources have now told me that NVIDIA's relationship with Samsung will only continue moving forward and that once this first wave of Ampere GPUs is made on the current 8nm node at Samsung that the next GPU on Samsung's new node will be when the fun really begins.
NVIDIA itself has confirmed that its codename Ampere Next GPU is coming in 2022 and then Ampere Next Next drops in 2024, which I think Ampere Next = Lovelace GPU while Ampere Next Next = Hopper. NVIDIA's new Hopper GPU architecture should be their first MCM (multi-chip module) design which will be similar to AMD and its use of chiplets in its Zen-based range of CPUs.
- Read more: NVIDIA Hopper GPU reportedly delayed, Lovelace GPU coming soon on 5nm
- Read more: NVIDIA next-gen GPUs: Ampere on 7nm, Hopper on 5nm
- Read more: NVIDIA's next-gen AD102 GPU: RTX 4090 twice as fast as the RTX 3090?!
Right now, Samsung's current 8nm node is not very good... but their new 5nm node from what I'm hearing, is extremely good and will bring them up to the level of TSMC. TSMC is already straining with supply and AMD Radeon RX 6000 series cards are not easy to buy, neither are the Xbox Series X/S and PlayStation 5 consoles.
Samsung would be serving NVIDIA and its next flock of GPUs in what should be the Lovelace GPU architecture, which should be powering the next-gen GeForce RTX 40 series graphics cards. I'm sure we'll be hearing plenty about this throughout 2021 and into 2022.
But for now, could we see GA100 power the GeForce RTX 3090 Ti in some way?
- > NEXT STORY: Securities fraud lawsuit against CD Projekt Group moves forward
- < PREVIOUS STORY: Sony patent teases Bitcoin and crypto-focused esports betting platform