aaron spink (aaronspink.delete@this.notearthlink.net) on July 27, 2012 7:33 pm wrote:
> jp (jipe4153.delete@this.gmail.com) on July 27, 2012 3:55 pm wrote:
>
> >
>
> > Local memory in OpenCL speak is what is known as
> > shared memory in
> CUDA, which can be thought of as a fast on-chip user managed
> > cache.
> >
>
> > With PCIe 3.0 cards the max PCI express bandwidth is now 16
> >
> GB/s.
> >
> Another reason to not like GPUs, they cannot use already well
> established terms. My personal favorite is that Nvidia cannot agree with Nvidia
> how many cores a 2090 has!
>
> Global node bandwidth is currently at 51.2 for
> trailing gen. 16 is a small fraction of that.

Aha, so your using personal bias against an architecture?

And your mixing different system hardware components and want to enforce the same naming convention.

Btw I would be careful using the term "cores" and instead talk about SIMD units and FPUs.

Global node bandwidth is cute, but a majority of applications benefit from a high localized bandwidth. 51.2 is a small fraction of 250+ .