Possible NVIDIA GeForce RTX AD106 GPU benchmarks surfaced

Presumably, the NVIDIA GeForce RTX 40 GPU with AD106 GPU offers almost the performance of a 3070 Ti with 26% lower power consumption. The name of this particular GeForce RTX 40 GPU hasn’t been decided yet, according to Chiphell, but it features the full AD106 GPU die with 4608 CUDA cores and 8 GB of memory, which means it should offer a 128-bit bus interface. The GPU also has 32 MB of L2 cache, a TMU count of 144 and 48 ROPs.

The card was tested in AIDA64 and various 3DMark benchmarks. In FP64, the card delivers up to 26 TFLOPs of processing power, which is 13% more than the RTX 3070 Ti. This results in a boost of about 2.8 GHz, which is higher than what we have seen with other Ada cards. The AD106 is between 10% slower and 3% faster than the RTX 3070 Ti in the standard benchmarks in 3DMark and 25% higher in the DXR Feature Test, which shows its ray tracing capabilities.

The most significant number, however, is the power consumption, where the card runs at a peak power of just 180W, which is 27% lower than the RTX 3070 Ti and 23% lower than the older TU104 GPUs. The NVIDIA AD106 GPU is expected to end up in the RTX 4060 graphics card, but with different specs as far as the rumors are concerned. The interesting thing is that the RTX 4070 laptop GPU has similar specs, but it doesn’t have such a high boost clock or a power of more than 115W.

AD106 vs GA104 vs TU104 3DMark score

AD106 will be RTX 4060 8GB?

According to post AD106 is..
– 4608 CUDA
– 2535MHz Boost
– 128bit D6 18Gbps, 288GB/s
– 32M L2
– 180W https://t.co/U4DXPE6BLX pic.twitter.com/FNo5j2CV7B

— 포시포시 (@harukaze5719) February 4, 2023

Compared to the RTX 3060 Ti, which had a TBP of 200W, the RTX 4060 Ti seems to have a 20% lower TBP – which, by the way, is also in line with what we’ve already heard about the card. The GeForce RTX 4060 Ti is expected to use the AD106-350-A1, which is a cut-down version of the AD106 graphics chip. It is rumored to feature 34 SMs or 4352 CUDA cores, an 8 GB GDDR6 memory running at 18 Gbps over a 128-bit bus interface, giving the card a bandwidth of 288 GB/s.

There is also 32 MB of L2 cache on board the GPU, which should be an 8x increase over the GeForce RTX 3060 Ti. This particular AD106 GPU from NVIDIA will likely target pure 1080p gamers who want high frame rates, while 1440p gaming would also be possible with DLSS techniques. NVIDIA thus has a lot of new mainstream products up its sleeve, which should hit the shelves in the second half of 2023.

NVIDIA GeForce RTX 40 Series graphics card lineup (rumored):

Source: Harukaze5719, Olrak