Nvidia's new professional cards could signal end to HBM shortage

Nvidia's new professional cards could signal end to HBM shortage

Nvidia might not have shown off any new consumer graphics cards at this year's Games Developer Conference and GPU Technology Conference shows, but that doesn't mean it didn't have any interesting hardware. In fact, the cards it did show off and the GPU rendering rigs it had to debut, suggest that the long-running shortage of High Bandwidth Memor -- first and second generation -- may well be over.

The first graphics card that Nvidia showed off during GTC was the Quadro GV100, which is technically two cards in one. It's a pair of Titan Vs connected using the NVLink2. They feature a total of 10,240 CUDA cores offering 236 TFLOPS of raw power from its Tensor Cores.

The really interesting part though is that together they have a combined 64GB of HBM2 memory. Considering it's only typically been used in single digit gigabyte amounts before now, that's quite an upgrade. It was also one that Nvidia applied to the standalone Tesla V100 GPU, which now has 32GB of its own HBM2 to play with.

As if that wasn't enough, Nvidia also had a monster GPU server box called the DGX-2, which Nvidia called the "World's largest GPU." Priced at $340,000, it contained a whopping 16 Tesla V100 graphics cards -- you can see where this is going. Collectively they have access to a shared pool of 512GB of HBM2 memory.

That's a staggering amount of high-speed memory. Although such a piece of kit is likely best suited to rendering movies and CGI effects, you have to wonder what such a system could do when it comes to GPU cryptocurrency mining.

It might take you a while to recoup your investment, but we wouldn't be surprised to see someone try and mine a few coins using one of these.