Soldato
Got some piccies and info here.
The tensor cores will be open for use by anyone for AI work or RTX. NVlink also allows faster and more native memory unification between cards but multi GPU still needs to be explicitly coded for, no auto free meal as it seemed.
Quadro RTX 6000 uses the TU104 die with 3072 cuda cores and has 4 DP and one usb-c virtual link port.
The tensor cores will be open for use by anyone for AI work or RTX. NVlink also allows faster and more native memory unification between cards but multi GPU still needs to be explicitly coded for, no auto free meal as it seemed.
Quadro RTX 6000 uses the TU104 die with 3072 cuda cores and has 4 DP and one usb-c virtual link port.
Last edited: