Something I've noticed for several years when installing Intel's inf drivers is that if you just double click the install, it doesn't usually install everything. What does get installed varies depending on hardware, OS and even between 32bit and 64bit versions of the same OS. For instance on my old MSI-GD65 i5-750 combo, pretty much everything would be installed in 32bit win7/Vista, and almost nothing would be installed when using the same OS but with a 64bit install.. I'd get around this by creating a shortcut on the desktop to the driver .exe and use the -overall switch that forces all drivers to be installed over the default Microsoft versions. On my most recent build I'm doing for a friend, an MSI H77 and 3770 combo, I found that when I first ran the creative suite from PCMark 8, I got a message saying the score couldn't be put on the board because of timing irregularities. This is an anti-cheat mechanism, here's Futuremark's description: "This message indicates funny business with the system clock during benchmark run. In most cases, this means that, no, you cannot cheat in 3DMark by adjusting Windows timers during the benchmark run or otherwise tampering with the measurements done by the benchmark. If this message persists and you have not done anything out of the ordinary, it may indicate a hardware issue with the real time clock in your system or the presence of a background program that somehow twists the time-space continuum of your operating system in such a way that this anti-cheat detection is tripped." Knowing I hadn't tried to cheat, or bend space and time (at least not intentionally), I was puzzled for a few seconds. I fired up LatencyMon and tried another run. It became obvious that there were lots of DPC latency issues with drivers. (I've had these before in win10, but only after upgrading from 7 to 10 because 10 keeps the same drivers and runs them in compatibility mode. A fresh install of 10 resolved all of this and more for me). So I set a system restore point and installed the drivers with the -Overall switch (had previously just used the .exe as anyone would be expected to). Not only did it pass, but the score went up from 50xx to 6208. Quite a big increase. I had LatencyMon in the background and could see there were far less latency spikes occurring. This means that hardware is communicating with each other much better. I've seen similar increases on a few builds I've tested in the past. For some reason I have a vague memory of a performance decrease too, but only once if at all. If you want to try for yourself, set a restore point so you can revert if needs be. There was still one noticeable spike related to the audio chain. Need to work out if it is AMD's audio driver or Realtek's.
