Hacker News Reader

The GPU => CPU memory bus is a major bottleneck for NVIDIA's growing Deep Neural Net driven adoption.

GPUs churn through data once it is across the bus.

A heirarchy of GPUs, outputs wired to inputs, mirroring the heirarchy of deep nets would be useful for real time robots & cars, NVIDIA's other big market.

View on HN

> A heirarchy of GPUs, outputs wired to inputs, mirroring the heirarchy of deep nets would be useful for real time robots & cars, NVIDIA's other big market.

Isn't that exactly what they're doing with NVLink?

https://devblogs.nvidia.com/parallelforall/inside-pascal (NVLink High Speed Interconnect section)

> heirarchy of GPUs, outputs wired to inputs

Nvidia introduced a new, faster SLI bridge for the new 1000 generation, aren't they used in GPGPU setups?

You just use the PCI bus. GPUDirect gives you DMA access to the other GPUs on the same bus. If they're in PCI express x16 slots, it's relatively speedy.

16 GByte/second sounds fast, but it's less than the bandwidth of dual channel DDR2 RAM. Never mind modern GDDR5X/DDR4/HBM memory.

The SLI bridge is quite slow though, even the updated SLI bridge is just 2GB/s.

Ouch, I'd have expected more.

Reply on news.ycombinator.com