I hope this encourages @intel to start launching desktop chips without so much GPU die space and with much higher core counts.
-
-
-
And higher end CPUs have *more* die space wasted on a GPU. Nobody who buys those chips don't already have a discrete gfx card!
End of conversation
New conversation -
-
-
could unlock interesting proc gen possibilities
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Also, this is available right now: Cavium 48-core ARM64.pic.twitter.com/dl3MKYpltn
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Sounds like a nineteen-core architecture, then?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
WIDER MEMORY! MORE CACHE! L5 CACHE WILL SAVE US!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
FWIW,
#Erlang has a very neat model, where each micro-process has its own GC-ed heap that's generally tiny. It would parallellize just fine!Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Yes, but so will anything that actually uses memory. The point is that the micro-heap can be collected in L1/L2 and thus "stream."
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Calling malloc()/free() will also evict cache lines and add to throughput! The only safe-ish space is the local stack :-)
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
"Again"? They never stopped being useful!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.