One of the tests that can maximize the potential of a GPU is the Octanebench benchmark, since it is used for rendering purposes and can maximize the performance of a graphics card, and almost no need to optimize the drivers. In the case of the NVIDIA RTX 4070 Ti, it has recently appeared in the Octanebench database and its scores beat the previous RTX 3090 Ti by 5%.
For many gamers, a 5% rendering improvement won't be much. But for professionals, 5% faster means less time rendering animations and scenes, a bit more speed and stability, and overall saving time and money. be marginally more efficient.
The RTX 4070 Ti already beats the 3090 Ti in a rendering benchmark for GPUs
The RTX 3090 Ti has 10.752 CUDA cores clocked at 1860 MHz, giving FP32 performance of 39,99 TFLOPs. The RTX 4070 Ti, meanwhile, has 7.680 CUDA cores clocked at 2.610 MHz, yielding an FP32 performance of 40,09 TFLOPs.
The Octanebench benchmark is also supposed to take into account the card's RT cores, of which the NVIDIA RTX 4070 Ti has 60 compared to the 84 Ti's 3090. The RTX 3090 Ti is very close, on paper, to the RTX 4070 Ti.
When comparing, the RTX 4070 Ti gets a score of 725,3 points, while the RTX 3090 Ti gets a score of 688,17 points, which is 5% faster for the new graphics card over the old one. the previous generation. They warn that it must be taken into account that this score is a combination of highly aggregated and weighted results. The RTX 4070 Ti has been noted to be much more powerful in the data channels test, probably due to the improved VRAM, where it scores 953 points, compared to 697 for the RTX 3090 Ti. But in the case of the Direct Lightning and Path Tracing tests, the score is almost the same with very little appreciable difference, 705 and 696, compared to 687 and 687 for the RTX 3090 Ti.
At first glance, it looks like a nice performance upgrade from NVIDIA and a good replacement for the RTX 3090 Ti. But they warn that in games it would be very different, since these are more sensitive to driver optimizations and have only a partial correlation with the theoretical maximum FP32 performance. They warn that it is not known if the GPU has been overclocked, which would greatly influence the results.
Source: Wccftech