You could try DISM scan.I did sfc scan of windows the day before yesterday as well which found errors and fixed them but the issue persists.
146796.660µs is 0.14 seconds so, setting any delay of 1 or higher should be ok, try 5 or 10. It may or may not make any difference at all. I think I have only had one gpu in the past where it helped, I was running 2 x gtx 295's so quad SLI, and one of the cards ran fine on like 35% of drivers and the others it needed the the TDR limit changed. with it changed, I noticed no problems, could overclock etc everything fine. I put it down to dodgy drivers at the time/SLI issues and didnt care. But with a single GPU, if new, I'd probably just RMA it, especially now there is availability so it wont be waiting a month etc."increasing the drive timeout delay via registry" - how do i do that?
maybe a red herring, it downclock speed when GPU is not under loadMeant to say that when I startup with lags, it shows 1
Thing is, on a normal startup i never shows one, even if not under load. Startups that do have a lag, it never goes higher than 3.maybe a red herring, it downclock speed when GPU is not under load
also may hint that pcie is not the suspect for lags
next suspect... Infinity Fabric? Up SoC voltage by 0.1-0.2V?