On newer generations of Intel Gen GPUs we now have the ability to reset specific engines for a light weight engine reset approach when things go wrong inside the GPU. Instead of resorting to a full GPU reset mechanism from the get go when things go wrong there are less aggressive options that could help us achieve a faster and smoother recovery from a hung GPU. This talk will go over one such mechanism that includes a low latency engine based reset approach through the use of a GPU watchdog timeout. The talk will end by discussing a specific use case under Chrome OS that shows how all the pieces fit together (including how media drivers can make use of this mechanism from user space) and also how a compositor could take advantage of this approach and improve the user experience.