Filed under: AD102, AD102 GPU, Featured, GeForce, GeForce GPU, GeForce Graphics Card, GeForce RTX, GeForce RTX 4090, Hardware, News, nVidia, NVIDIA 4090, Nvidia GeForce, NVIDIA GeForce RTX 4090, Nvidia GPU, Nvidia graphics card, NVIDIA RTX 4090, NVVIDIA GeForce RTX, RTX 4090, RTX Graphics Card, Rumor, Sticky

NVIDIA Allegedly Begins Testing Its Fastest Next-Gen GPU, The AD102,…

NVIDIA's Following-Gen Flagship GeForce Graphics Card Specific: AD102 GPU, 24 GB & 21 Gbps Memory, Up To 600W TGP

NVIDIA’s following-gen AD102 GPU is at the moment getting examined for the GeForce RTX 4090 graphics card, as reported by Kopite7kimi.

NVIDIA GeForce RTX 4090 Rumors: Flagship AD102 GPU Enters Check Section, 24 Gbps GDDR6X Memory All set

There has been speculation that NVIDIA could possibly go for the GeForce RTX fifty sequence naming scheme alternatively of the predicted GeForce RTX forty sequence branding but it looks like Kopite7kimi has said NVIDIA’s decision to adhere to the forty series naming plan. Other than that, one particular major milestone is that NVIDIA may perhaps by now have started out screening and assessing its flagship Ada Lovelace GPU, the AD102, which will power a sequence of graphics cards these types of as the RTX 4090 and the RTX 4080.

Now we do not know if NVIDIA has moved Ada Lovelace GPUs to mass generation yet but that could be taking place in the course of the mid of 2022 as hinted beforehand. So the chips staying tested below could be early samples. But if these chips are indeed out in the wild, then we might probably see some leaks in the coming months. Of course, the efficiency figures will never glance that good at 1st due to the fact these GPUs are however to get finalized motorists but it may possibly give us an idea of the type of clock speeds we can expect in ultimate variants.

A different appealing issue that Kopite7kimi states are that NVIDIA is going to use 24 Gbps memory dies on its AD102 GPU. The leaker only mentions it for the RTX 4090 and we know that the 3090 is the only series that received the fastest GDDR6X memory answers so considerably. Regardless of whether NVIDIA chooses 24 or lower pin speeds for the relaxation of the lineup is some thing that continues to be to be noticed. A single matter is for guaranteed we will be finding about one TB/s memory speeds on client gaming graphics cards which are likely to give us a 15% boost in bandwidth quantities and that’s without having accounting for the brand new compression systems that will be onboard the Ada Lovelace architecture.

NVIDIA Ada Lovelace & Ampere GPU Comparison

Ada Lovelace GPU SMs CUDA Cores Top rated SKU Memory Bus Ampere GPU SMs CUDA Cores Top rated SKU Memory Bus SM Raise (% Around Ampere)
AD102 one hundred forty four 18432 RTX 4090? 384-bit GA102 84 10752 RTX 3090 Ti 384-little bit +seventy one%
AD103 eighty four 10752 RTX 4070? 256-bit GA103S 60 7680 RTX 3080 Ti 256-bit +40%
AD104 sixty 7680 RTX 4060? 192-little bit GA104 forty eight 6144 RTX 3070 Ti 256-little bit +25%
AD106 36 4608 RTX 4050 Ti? 128-bit GA106 30 3840 RTX 3060 192-bit +20%
AD107 24 3072 RTX 4050? 128-bit GA107 20 2560 RTX 3050 128-bit +twenty%

NVIDIA GeForce RTX 4090 ‘AD102 GPU’ Graphics Card PCB Style and design – What We Know

The NVIDIA GeForce RTX forty sequence graphics cards outfitted with the AD102 GPUs are expected to offer TDPs of up to 600W. That is at the very least what the present BIOS shipping and delivery to board associates is rated at so the rumors about 450-600W TDPs may be genuine but we haven’t nonetheless observed the last figures. The ability ratings are usually on the substantial facet through the tests period so people could be optimized when the playing cards essentially start. The playing cards will be outfitted with the PCIe Gen 5 connectors and will ship with a four x eight-Pin to 1 x sixteen-pin adapter to support the massive electrical power draw. The future GeForce RTX 3090 Ti itself will be delivery with a 3 x eight-Pin to one x sixteen-Pin adapter.


The leaked PCB style talked about the memory which will attribute twelve solder factors on the PCB and all are compatible with Micron’s GDDR6X memory. Increased-stop cards might go with solitary-sided and twin-ability memory considering that that provides the best electric power/temperature balance and function up to 24 GB capacities but at greater speeds (Up To 24 Gbps). As for the mainstream section, we are very likely to see 20 Gbps+ models but in eight GB and up to 16 GB flavors which can assist decrease electric power considering the fact that the electricity regulation will be dropped to 3 VRMs for the memory.

As for cooling these monster PCBs, Intel is reportedly going to reuse their triple-slot BFGPU style and design while board companions are likely to make use of 3.5 and even quad-slot cooling methods weighing in excess of 2 kg. Most AIBs may possibly just end up making use of AIO and Hybrid cooling models, some thing that you will be seeing in the RTX 3090 Ti. The playing cards are anticipated to attribute up to a 28 section VRM design on the flagship NVIDIA GeForce RTX 4090 graphics card so all that more cooling will be put to good use.


GPU TU102 GA102 AD102
Architecture Turing Ampere Ada Lovelace
Method TSMC 12nm NFF Samsung 8nm TSMC 5nm
Die Dimensions 754mm2 628mm2 ~600mm2
Graphics Processing Clusters (GPC) six 7 twelve
Texture Processing Clusters (TPC) 36 42 72
Streaming Multiprocessors (SM) 72 84 144
CUDA Cores 4608 10752 18432
L2 Cache six MB 6 MB 96 MB
Theoretical TFLOPs 16.one 37.six ~ninety TFLOPs?
Memory Bus 384-bit 384-bit 384-bit
Memory Potential eleven GB (2080 Ti) 24 GB (3090) 24 GB (4090?)
Flagship SKU RTX 2080 Ti RTX 3090 RTX 4090?
TGP 250W 350W 450-850W?
Launch Sep. 2018 Sept. twenty 2H 2022 (TBC)

The NVIDIA Ada Lovelace GPU loved ones is predicted to bring a generational soar related to Maxwell to Pascal. It is envisioned to launch in the next half of 2022 but expects offer and pricing to be similar to latest cards inspite of NVIDIA spending billions of dollars to obtain all those very good good TSMC 5nm wafers.

Which next-technology GPUs are you looking ahead to the most?

The write-up NVIDIA Allegedly Starts Testing Its Speediest Future-Gen GPU, The AD102, For GeForce RTX 4090 Graphics Card, Attributes 24 Gbps GDDR6X Memory by Hassan Mujtaba appeared to start with on Wccftech.