The Chiphell forums have allegedly leaked performance benchmarks for an NVIDIA GeForce RT 40 graphics card based on the AD106 GPU. According to Chiphell, the name of this GeForce RTX 40 GPU has not yet been finalised, but it does feature the full AD106 GPU die with 4608 CUDA cores and packs 8 GB memory, implying a 128-bit bus interface. The GPU also has 32 MB of L2 cache, 144 TMUs, and 48 ROPs.
AIDA64 and various 3DMark benchmarks were used to test the card. The card delivers up to 26 TFLOPs of compute power in FP64, which is 13% more than the RTX 3070 Ti. This gives us a boost of around 2.8 GHz, which is higher than the other Ada cards.
The AD106 GPU scores 10% slower to 3% faster than the RTX 3070 Ti, with a 25% increase in the DXR Feature Test demonstrating its raytracing prowess.
The most important number is the power consumption, with the card drawing only 180W at peak, which is 27% less than the RTX 3070 Ti and 23% less than the older TU104 GPUs. According to rumours, the NVIDIA AD106 GPU will be included in the RTX 4060 graphics card, but with different specifications. The RTX 4070 Laptop GPU has similar specifications, but it does not have a boost clock this high, nor does it have a board power that exceeds 115W.
In comparison to the RTX 3060 Ti, which had a TGP of 200W, the RTX 4060 Ti appears to have a 20% lower TGP, which is consistent with what we’ve heard about the card.
The GeForce RTX 4060 Ti is expected to use the AD106-350-A1 GPU core, a scaled-down version of the full AD106 GPU, with 34 SMs or 4352 CUDA cores, an 8 GB GDDR6 memory running at 18 Gbps across a 128-bit bus interface, and 288 GB/s of bandwidth. The GPU also has 32 MB of L2 cache, which is an 8x increase over the GeForce RTX 3060 Ti.
Also Read:
