apparently there are some scaling issues with Slingshot
Hardware on the floor != system running public science.
Thank you, kind sirs. I merely quoted the relevant bit for this thread.
apparently there are some scaling issues with Slingshot
Hardware on the floor != system running public science.
That's CDNA, really.
According to this leaker MI300 is 6nm
Wonder which changes it has if true
Not sure about the leaker's history, but in this tweet he/she is asking if it's possible for MI300 to be ported to 6nm. Literally: "is it possible for MI300 to be ported to 6nm?".
His history is pretty goodNot sure about the leaker's history, but in this tweet he/she is asking if it's possible for MI300 to be ported to 6nm. Literally: "is it possible for MI300 to be ported to 6nm?".
Wroooong.According to this leaker MI300 is 6nm
And other tiles.Will MI300 consists of 4GCDs?
Not designed to.Will it even fit to OAM module package?
It should fit SH5 socket.//So, 4GCDs on 6nm + HBMs. Will it even fit to OAM module package?
Bingo.SH5 socket.
It should fit SH5 socket.
That's not a rumor, someone at DOE did a big oopsie.My bad, I completely missed previous rumors
Was this public oopsie? Can't remember it, not that I doubt it eitherThat's not a rumor, someone at DOE did a big oopsie.
Fortunately no.Was this public oopsie?