Nvidia Rectifies Error in One of Its Latest RTX 40 Super GPUs — GeForce RTX 4070 Super Features 48MB L2, Contrary to Earlier 36MB Report
Nvidia’s upcoming GeForce RTX 4070 Super is expected to be the top card; Early specifications show that the GPU will have 48 MB L2 cache. However, there is an error in the data published by Nvidia that the L2 cache is 36MB. Nvidia later clarified this and Wccftech reported that the actual spec was 48MB.
GeForce RTX 4070 Super is made of an AD104 graphics processor with 7,168 CUDA cores and a 192-bit memory interface. In particular, the GeForce RTX 4070 Ti, which has 7,680 CUDA cores, uses the AD104 GPU replacement. Despite the lower CUDA core count, the AD104 in the GeForce RTX 4070 Super provides the best possible memory bandwidth by maintaining the full 48 MB L2 cache.
Although the reason for the typo in Nvidia’s version information is unclear, the official website says that the GeForce RTX 4070 Super has a 48 MB L2 cache and a memory sub-disk capable of delivering 504 GB/s maximum bandwidth. system.
In terms of computing performance, the GeForce RTX 4070 Super and GeForce RTX 4070 Ti offer 35.5 FP32 TFLOPS compared to 40.09 FP32 TFLOPS. Integrated memory means new RTX 4070 Super motherboards should perform almost the same as the RTX 4070 Ti. The performance gap can be further narrowed by overclocking.
It’s worth noting that Nvidia and TSMC’s continued improvements in process technology can increase efficiency and reduce performance fluctuations, bringing greater overclocking capabilities to GPUs built today. Therefore, the overclocking capabilities of AD104 GPUs should now exceed those of GPUs launching in 2022 or early 2023.
What’s exciting is the expectation of what the AIB model of Nvidia’s dedicated GeForce RTX 4070 Super graphics cards will be like. to behave. Some motherboards are expected to push TGP (Total Graphics Power) above the 220W set by Nvidia, and they may also make GPUs more capable of overclocking.
