According to Kopite7kimi, NVIDIA’s next-gen AD102 GPU is now being tested for the GeForce RTX 4090 graphics card. There was speculation that NVIDIA will use the GeForce RTX 50 series branding instead of the expected GeForce RTX 40 series branding.
But it appears that Kopite7kimi has confirmed NVIDIA’s choice to continue with the 40 series branding. Aside from that, NVIDIA may have already begun testing and evaluating its flagship Ada Lovelace GPU, the AD102, which will power a series of graphics cards including the RTX 4090 and RTX 4080.
The AD102 GPUs in the NVIDIA GeForce RTX 40 series graphics cards are predicted to have TDPs of up to 600W. During the testing phase, power ratings are normally on the high side, so they can be tweaked once the cards are released. To support the massive power draw, the cards will be equipped with PCIe Gen 5 connectors and a 4 x 8-Pin to 1 x 16-Pin converter. A 3 x 8-Pin to 1 x 16-Pin adaptor will be included with the future GeForce RTX 3090 Ti.
We can expect some more action from RTX 4090 in the coming months
It’s unclear whether NVIDIA has begun mass production of Ada Lovelace GPUs, although as previously said, this might happen in the middle of 2022. As a result, the chips being tested here may be early prototypes. However, if these chips are actually in the wild, we may witness some leakage in the coming months.
Of course, because these GPUs haven’t yet received completed drivers, the performance stats won’t appear fantastic at first, but they should give us an idea of the clock rates we can expect in final models.
Another noteworthy detail revealed by Kopite7kimi is that NVIDIA’s AD102 GPU will feature 24 Gbps memory dies. The RTX 4090 is the only one mentioned by the leaker, and we already know that the 3090 is the sole series with the fastest GDDR6X memory solutions.
It remains to be seen whether NVIDIA picks 24 or lower pin speeds for the rest of the range. One thing is certain: consumer gaming graphics cards will have memory speeds of over 1 TB/s, resulting in a 15% increase in bandwidth, and this is without accounting for the brand new compression methods that will be built into the Ada Lovelace design.
The memory will include 12 solder spots on the PCB, all of which are compatible with Micron’s GDDR6X memory, according to the leaked PCB design. Single-sided and dual-capacity memory may be used on higher-end cards since it provides the best power/temperature balance and can hold up to 24 GB of data but at faster rates (Up To 24 Gbps). In the mainstream sector, we’re likely to see 20 Gbps+ architectures in 8 GB and up to 16 GB flavours, which will assist cut power consumption because the memory’s power regulation will be reduced to three VRMs.
NVIDIA CUDA GPU (RUMORED) Preliminary Specs:
| GPU | TU102 | GA102 | AD102 |
| Architecture | Turing | Ampere | Ada Lovelace |
| Process | TSMC 12nm NFF | Samsung 8nm | TSMC 5nm |
| Die Size | 754mm2 | 628mm2 | ~600mm2 |
| Graphics Processing Clusters (GPC) | 6 | 7 | 12 |
| Texture Processing Clusters (TPC) | 36 | 42 | 72 |
| Streaming Multiprocessors (SM) | 72 | 84 | 144 |
| CUDA Cores | 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 | Sep. 2018 | Sept. 20 | 2H 2022 (TBC) |
also read:
AMD to release its Ryzen 7000 ‘Zen 4’ CPUs With 3D V-Cache for Desktops in 2023
