GeForce-RTX-4000 Ada Lovelace - A Titan with 48 GB at 900 Watt?
Nvidia's AD102 GPU alone can produce 450W, but speculation of a new Titan indicates that it can utilize up to 900W, a product paired with 48GB of DDR6X memory.
The GeForce RTX 4090 was rumored to have a TDP of up to 850 watts as early as February. Nvidia's forthcoming Ada generation is expected to have a TDP of 450 to 600 watts, depending on the model. Twitter user @kopite7kimi, had asked for information regarding a test board that might be used for a future Titan RTX.
The AD102 graphics processor is claimed to be the biggest GPU in the GeForce RTX 4000 series, maintains a conventional monolithic architecture, and is expected to be utilized in the GeForce RTX 4090 and a speculative Titan RTX. Meanwhile, the GeForce RTX 4080 is rumored to feature 16 GB of GDDDR6X memory and a TDP equivalent to the current GA102 GPU (up to 450 watts), while the GeForce RTX 4070 is supposed to have 12 GB of GDDR6 and a TDP of roughly 300 watts.
There are three 16-pin power connections on the test board that, according to the specs, could feed 1800 watts, so it's possible that this isn't an accurate representation of what the final product would look like.
The forecast for the Ada AD102 GPU
- TSMC N5 or N4 (custom process)
- up to 18432 CUDA cores
- up to 48GB GDDR6X at 24Gbps @ 384-bit
- up to 2× 16-pin (12+4-pin) PCIe 5.0 ("12VHPWR")
- up to 12 graphics processing clusters
- up to 144 streaming multiprocessors
- up to 72 texture processing clusters
- up to 600 mm² die size
- up to 900 watts TDP
Senior Member
Posts: 543
Joined: 2018-12-12
https://videocardz.com/newz/amd-rdna3-gpus-get-new-toned-down-specifications-navi-31-with-up-to-12288-stream-processors
Senior Member
Posts: 18980
Joined: 2008-08-28
I guess it was too good to be true. Still rumors are just that rumors.
Senior Member
Posts: 5135
Joined: 2020-08-03
hmmmmm,I was really interested in the single die rdna3 on 5nm,when the specs said it's basically a 5100 shader 6900xt made on 5nm.
let's see how it develops.
smaller dies needed in order to maximize the benefits of mcm ? 3x 4096sp, vs. 2x 5100, if 3x 5100 is not feasible.
whatever die(s) the cards pack,amd and nvidia have a way of intentionally crippling them with a smaller bus or pci-e lanes so that they don't look so enticing for higer resolution which forces you to buy up.
theoretically,a 256-bit g6 card on pcie16x wouldn't even slow limit 4K performance that much.so they will be looking for ways to make it appealing only for whatever market they advertise it for.
somehow I feel that even afer the next gen launch, 3080 12G and 6800xt are going to be very desireable products if they sell at msrp.
Senior Member
Posts: 3993
Joined: 2017-11-23
That's why i said if the differences are the same as between the 3070 and 3060 ti. Value wise the 3080 and 3060 ti made way more sense than the rest of the lineup in the 3000 series.