NVIDIA’s Ada Lovelace GPUs will hit gamers later this year in the next generation of GeForce RTX graphics cards. These cards are expected to rock at some extreme specifications and some extreme GPU requirements to function as described in detail Kopite7kimi in its latest round of leaks.
NVIDIA’s next generation flagship GeForce graphics card detailed: AD102 GPU, 24 GB & 21 Gbps memory, up to 600W TGP
There may be one new rumor around the naming convention used by the next generation of GeForce graphics cards, and some are now pointing to a name change to the GeForce RTX 50 series instead of the previously expected GeForce RTX 40 series. In the end, whatever NVIDIA decides to call the next GeForce series is not as important as its underlying specifications.
** 90 PG137 / 139-SKU330, AD102-300, 24G 21Gbps, 600W
– kopite7kimi (@ kopite7kimi) April 11, 2022
NVIDIA GeForce RTX 4090 ‘Ada Lovelace’ graphics card with GA102-300 GPU ‘Rumored’ specifications
We know from the leaked NVIDIA Ada Lovelace GPU configurations, which show that the flagship GA102 Gaming chip will rock a total of 144 SMs, a 71% increase over the existing GA102 GPU, and house a massive number of CUDA cores of 18,432. The chip used by the flagship GeForce graphics card will shake the AD102-300 GPU, and we can not say for sure if this will be the full-fat SKU, as the RTX 3090 Ti with GA102-350 rocks the full configuration, while the GA102 -300 SKU on NVIDIA GeForce RTX 3090 is a cut SKU.
According to Kopite7kimi, the flagship GeForce graphics card will be based on the PG137 / 139-SKU30 PCB. These are both reference and Founders Edition designs. The reference design is provided to AIBs as a reference for their custom models, while the FE design is exclusive to the Founders Edition reference model.
The graphics card is said to retain 24 GB of memory capacity while offering a 21 Gbps memory speed across a 384-bit bus interface. It is the same memory design as the flagship Ampere GeForce, GeForce RTX 3090 Ti, and should rock the same amount of bandwidth at 1004 Gb / s. As for the power input, the card will have a single PCIe Gen 5 connector with a TGP rating of up to 600W. You will need a 4x 8-pin to 1x 16-pin adapter if you are running an older PSU or a single Gen 5 to Gen 5 connector if you have a compatible power supply.
NVIDIA Ada Lovelace ‘GeForce RTX 40’ GPU configurations
GPU name | GPCs / TPCs | SMs per TPC / total | CUDA colors | L2 cache | Memory bus |
---|---|---|---|---|---|
AD102 | 12/6 | 2/144 | 18432 | 96 MB | 384-bit |
AD103 | 7/6 | 2/84 | 10752 | 64 MB | 256-bit |
AD104 | 5/6 | 2/60 | 7680 | 48 MB | 192-bit |
AD106 | 3/6 | 2/36 | 4608 | 32 MB | 128-bit |
AD107 | 3/4 | 2/24 | 3072 | 32 MB | 128-bit |
NVIDIA Ada Lovelace & Ampere GPU Comparison
There is a Lovelace GPU | SMS | CUDA colors | Top SKU | Memory bus | Ampere GPU | SMS | CUDA colors | Top SKU | Memory bus | SM increase (% above amperes) |
---|---|---|---|---|---|---|---|---|---|---|
AD102 | 144 | 18432 | RTX 4090? | 384-bit | GA102 | 84 | 10752 | RTX 3090 Ti | 384-bit | +71% |
AD103 | 84 | 10752 | RTX 4070? | 256-bit | GA103S | 60 | 7680 | RTX 3080 Ti | 256-bit | +40% |
AD104 | 60 | 7680 | RTX 4060? | 192-bit | GA104 | 48 | 6144 | RTX 3070 Ti | 256-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 | +20% |
The NVIDIA Ada Lovelace GPU family is expected to bring a generational leap similar to Maxwell to Pascal. It is expected to launch in the second half of 2022, but expects supply and prices to match current cards despite NVIDIA spending billions of dollars on acquiring the good, good TSMC 5nm wafers.
NVIDIA CUDA GPU (RUMORED) Preliminary:
GPU | TU102 | GA102 | AD102 |
---|---|---|---|
Architecture | Turing | Ampere | Ada Lovelace |
Treat | TSMC 12nm NFF | Samsung 8nm | TSMC 5nm |
Die Size | 754 mm2 | 628 mm2 | ~ 600mm2 |
Graphics Processing Clusters (GPC) | 6 | 7 | 12 |
Texture Processing Clusters (TPC) | 36 | 42 | 72 |
Streaming multiprocessors (SM) | 72 | 84 | 144 |
CUDA colors | 4608 | 10752 | 18432 |
L2 cache | 6 MB | 6 MB | 96 MB |
Theoretical TFLOPs | 16.1 | 37.6 | ~ 90 TFLOPs? |
Memory type | GDDR6 | GDDR6X | GDDR6X |
Memory bus | 384-bit | 384-bit | 384-bit |
Memory capacity | 11 GB (2080 Ti) | 24 GB (3090) | 24 GB (4090?) |
Flagship SKU | RTX 2080 Ti | RTX 3090 | RTX 4090? |
TGP | 250W | 350W | 450-850W? |
Release | September 2018 | Sept. 20 | 2H 2022 (TBC) |