"graphics memory" is probably not the right term here. While the Pro SSG does have 2TB of storage on board, that comes in the form of a NAND flash, PCIe SSD (from Samsung IIRC). While fast, this SSD provides single digit GB/s bandwidth at most, compared to the hundreds of GB/s of bandwidth that the DRAM-based GDDR5 or HBM provide. When we say "graphics memory," we're almost always referring to the latter, and never see more than maybe 32 GB of DRAM based memory on board.

Yeah, that seems like false advertising. I believe that much true HBM/GDDR5 would have to cost tens of thousands of dollars. So I was skeptical. That said: if I could get a card with that much real GDDR5 on it, I'd consider paying whatever it took.

More curious about the latency and how that impacts performance, even with DMA going over the PCIe bus to a NVMe SSD is a longer trek.

GPUs are generally not latency optimized anyway - if an app is sensitive to tens of microseconds then it belongs on the CPU, broadly speaking.

AMD's advantage here is getting drop-in throughput, by switching the SSD directly to the GPU. Thus, the onboard SSD doesn't waste host PCIe lanes (4 lanes, typically) that are getting hard to come by in regular desktop computing systems.

Isn't the "2TB" NVMe SSD (in this case) directly interfaced to the card, so not going over the system PCIe bus?

It is, which is why I'm curious how this improves performance compared to exactly the scenario of hitting NVMe storage over the bus.

Indeed, the non-Pro variant only gets you 512GB of mechanical memory

"having 2TB of solid state graphics memory"

That made me chuckle. What else would it be? Vacuum tube graphics memory?

Dayum. Now if they would just port Thrust to OpenCL I'd be in business.

I am surprised this claim is buried in the article with no highlight whatsoever.