mcm amd costs is 59% that of a monolithic design... intel cant

>>62048297
Who covered the EPYC hotchips presentation? Seeing nothing on anand

Glue

>>62048324
No one did.
No one also covered RTG presentation and NVIDIA one.

>>62048324
servethehome

>>62049164
Hello Brian. Friendly reminder to lube before shareholders meeting ^_^

>>62048297
do all the ccx share the pcie lanes, or are certain pcie lanes assigned to a specific ccx die?

>>62049164
What the fuck is EMIB all about then?

>>62049204
I think it works the same as the memory controller. Each die has a dedicated controller but it's shared between all of them.

>>62049207
I-intel glue is different...

>>62049237
It is.
It's still da glue.

>>62049204
Lanes are shared.

>>62049181
they talked about navi there, damn, what did they say?
nvidia talked about goyworks, who cares

>>62049296
>they talked about navi there, damn, what did they say?
They did?

>>62049296
No, AMD talked Vega, NVIDIA talked Volta & Volta programming.

>>62049335
it said next generation gpu on schedule so I thought vega is not next generation anymore
still, would be nice to know what both said then

>>62049359
>muh NCU
>muh SR-IOV
Versus
>muh L1$ and unified memory
On the NVIDIA side.
Nothing exactly interesting.

>>62048297
now do this to gpus. I wanna see mcm gpus connected with infinity fabric.

File: 1502488836418.jpg (24KB, 267x297px) Image search: [Google]

24KB, 267x297px

>>62048297
NOOO INTEL WAS SUPPOSED TO BE BETTER