NVIDIA Readying Tesla V100 Successor? Unknown GPUs with 108, 118 Compute Units Spotted


_rogame has spotted two unknown NVIDIA hardware listings on GeekBench. The scores are impressive enough to suggest that they could be next-generation Tesla products – high-performance graphics cards that are geared toward the data center.

The first achieved an OpenCL score of 184096 with 118 compute units, a core clock of 1.11 GHz, and 24 GB of memory, while the second scored 141654 with 108 compute units, a core clock of 1 GHz, and 47 GB of memory.

That would equate to a CUDA core count of 7552 and 6912, respectively. (As VideoCardz notes, this figure can be derived by multiplying the amount of compute units by 64.) In comparison, the TITAN RTX and RTX 2080 Ti boast 4608 and 4352 CUDA cores, respectively.

These listings are actually a couple of months old, but the performance numbers seem to confirm that NVIDIA has transitioned to a 7 nm process. The low core clocks also lend credence to the theory that these are Tesla successors.

All of this bodes well for NVIDIA’s next round of GeForce cards, which the company is rumored to unveil at the Game Developers Conference or GPU Technology Conference in March. During a recent earnings call, CEO Jensen Huang remarked that attendees “won’t be disappointed.”

Recent News