To be precise, the GeekBench specifications don’t list down the CUDA core count for both GPUs, but rather the respective Compute Units (CU) for each model. For the uninitiated, though, CUDA cores are essentially the CU count multiplied by 64. The first NVIDIA GPU, specifically, lists down an OpenCL score of 184096 points with 24GB of graphics memory clocked at 1.11GHz. Along with 118 Compute Units. One thing to note is that the Geekbench listing doesn’t specify what sort of memory format the GPU is running on. Seeing how the current both NVIDIA’s Turing and AMD’s Navi GPU architecture both run on GDDR6, it’s not wrong to deduce that the unnamed GPUs are doing the same.
7552 Cuda cores (118 CUs)> 1.11GHz core clock> 24GB of memory GB5 Compute score: 184096 (Open CL)https://t.co/rUFWACaxrg 6912 Cuda cores (108 CUs)> 1.01GHz core clock> 47GB of memory GB5 Compute score: 141654 (Open CL)https://t.co/C1jnm3sZ0D pic.twitter.com/3JrW8LrnFj — _rogame (@_rogame) February 28, 2020 As for the other GPU, it has a lower OpenCL score of 141654 points, as well as fewer CUDA cores at 6912 cores. However, the GPU seems to be running a whopping 47GB of graphics memory at a clockspeed of 1.01GHz. At the moment, it’s unclear when NVIDIA intends to make the existence of these alleged GPUs public, considering its policy of not talking about unreleased and unfinished products. That said, GTC 2020 is just around the corner, so we’re kind of hoping we’ll get some sort of information about the GPUs then. (Source; GeekBench [1] [2] via _rogame@Twitter, Videocardz)
