Point of view of a dev (anonymised) from an IRC channel :
the scorpio has a dedicated hardware unit for sending draw calls to the gpu?
interesting
hmm, does it not have a unified memory architecture?
will it be akin to threading your rendering code, albeit on an entirely different processor?
cause if you have to feed that processor from the cpu, they didnt solve the bottleneck, only moved it
lol this digital foundry article
it claims that not being bound by draw calls will make the scorpio gpu perform on par with a 1070
uuuuuuuuuuuuuh there are ways to mitigate your draw calls, like combining vertex buffers into a large array
and once you get past starving your gpu, there are several other factors into performance. I very much doubt that claim.
reading further on this, its not something that would hinder backwards compatibility at all
its just like I figured, its running your rendering thread on a secondary processor
you basically are going to treat that hardware like a core
ms is bragging that their gpu commands are "based off of dx12."
well duh, thats how things like vulkan and dx12 work, they expose the intermediary language directly without an abstraction layer
you write your own spir-v translation
thats, like, the entire point of dx12 and vulkan
right on schedule, im seing some hilarious overreacting to "dx12 on da mf CPU!!!!1!1!11"
microsoft talking about dx12 being built into the gpu is akin to goodyear bragging that streets are made for the shape of their tires
yeah, their tires and everyone elses tires







