Monday, October 4th 2021
NVIDIA Rumored To Launch RTX 3090 SUPER in January 2022
NVIDIA has been preparing a new flagship graphics card with their RTX 3090 SUPER featuring a fully unlocked GA102 GPU. The RTX 3090 SUPER will include 10,752 CUDA cores, 336 Tensor cores, and 84 RT cores paired with 24 GB of GDDR6X memory on a 384-bit wide memory bus. These speed improvements are set to increase the cards TDP to 400 W compared to 350 W for the base RTX 3090. We have recently seen a new rumor from @hongxing2020 that claims NVIDIA will be launching the RTX 3090 SUPER, a refreshed RTX 3070 Ti with 16 GB memory, and the RTX 2060 12 GB in January 2022. The RTX 3090 SUPER and RTX 2060 12 GB launches have been expected for some time but this is the first mention we have seen of an RTX 3070 TI 16 GB.
Source:
@hongxing2020 (via VideoCardz)

40 Comments on NVIDIA Rumored To Launch RTX 3090 SUPER in January 2022
Yeah only cool if you have a bestbuy that sells them near by too lol
Otherwise you'll be on a road trip camp out.
A power target of 71% (125W TDP instead of 175) gives me around -7% performance compared to stock, which is barely noticeable even with an on-screen fps counter.
A power target of 114% (200W TDP instead of 175) gives me no observable or measurable performance gain, only more heat and noise.
This makes me question manufacturers' choice of TDP. Does our new PC gaming world really revolve around that extra 1% so much? :confused:
And much more so of the competitor has that 1%.
If you want a specific recent example, look at 6900XT "XTXH" cards. - extra 30% and 100W over 6900XT for 5-7% performance.
Clocking and targeting GPUs beyond the nice place on the efficiency curve has been AMD thing for a long while. With Turing and especially with Ampere, Nvidia followed suit.
As with your example lowering power target shows that quite clearly.
Undervolting today is a much more complex question. Generally undervolting guides for something like Ampere simply set the maximum voltage and frequency and are done with that. Actual undervolting seems to be a smaller part of the process.
You might as well buy upcoming coins and hope they explode instead of wasting electricity. That's what I did. Did not mine for one minute, yet earned alot of money from bitcoin, ethereum and dogecoin so far. However it's a gamble and there's many ways to make money.
Overclocking used to be a lot of fun, getting 15..20...25%++ out of CPUs and GPU's in the past with little effort. Nowadays the vendor pushes them pretty much to the limit of their capability, where you're fairly lucky to get 10-12% extra under non-extreme conditions like LN2.
So I scratch my tweaking / overclocking / optimizing itch by undervolting, keeping the most performance I can (perhaps even a few % points more) while drastically reducing power draw and pumping up efficiency to fun heights.
I'd love to see a comparison of Pascal v Turing v Ampere, all stock vs all UV'd to get a glimpse of stock efficiency vs UV's efficiency gains gen-to-gen. Would be bloody hard to pull off though, silicone lottery and all.
For example, at 1440p with the stock power limit, my core clocks is at 2145mhz/1.0V, but @4K and stock power limit, the core end up at 1920mhz/850mV, at this V/F point the Perf still scale almost linearly with power and it's a waste not to give the core some more power to work with.