Hot Chips: Xbone Silicon (5b transistors, 363mm2, 8GB Flash, 1.31TFlops, 204GB/s peak BW)

[DJEVOLVE]

MB1025 said:

ethomaz said:

The quote I made was not that... this one is about the PS4 onion bus lol

http://www.neogaf.com/forum/showpost.php?p=78403725&postcount=104

My understanding of hUMA is it both gives coherency to CPU/GPU acess of shared memory pool but also removes software defined partition to the memory space so both gpu and cpu can access all portions of the shared main memory space.

Onion looks to offer the GPU acess to information in the CPU l2 cache, which I misstated, but that doesnt seem to me what hUMA does.

 

What I posted before this was a direct response your post got on GAF. So how does it not have to dow ith each other. What I posted above was another clarification from the same poster. You nor I have any idea what this is yet you are just posting things from GAF without clarification and then trying to play off that. Then  you tell people they are wrong because your post you posted from GAF says oher wise.

Yea something like that.....