A method and apparatus for collecting debug and crash information are described. In one embodiment, a system comprises one or more compute engines an external interface; a non-volatile memory coupled to the external interface and operable to store captured information, wherein the captured information comprises one or both of debug information and crash information; a first trace aggregator coupled to the non-volatile memory and the one or more compute engines to capture the one or both of debug information and crash information from at least one of the one or more compute engines in response to a crash of the system; and a controller, coupled to the non-volatile memory and the first trace aggregator, to cause captured information to be sent from the first trace aggregator to the non-volatile memory and to subsequently control transfer of the captured information stored in the non-volatile memory to the external interface.