Hi Andy, I re-checked the logic, cqp behaviour is affected patch#1. By re-testing cqp, I confirmed that force to submit two consecutive single job logic should be applied to cqp as well, otherwise we have a chance to miss a flush or end up with sending last p and next i frame together which may protentially cause corruption. Nice catch! Solution is simple, as you already tried, remove the "if statement" to apply new logic to cqp as well. New patches can be found at: https://lists.freedesktop.org/archives/mesa-dev/2016-November/136860.html https://lists.freedesktop.org/archives/mesa-dev/2016-November/136855.html I will push them to upstream once you confirmed the fix. :-) Regards, Boyuan