So there are reports that the newest Nvidia driver actually slows down scientific computations. This from someone over at Einstein.

There is also notes from Nvidia stating it causes issues with people running Motherboards with PLX chips and multiple GPUs.

So a word of warning not to be anxious in adopting the newest drivers just yet.
____________

Last time I have changed MOBO and CPU + RAM + system and have installed newest drivers for 1060 3GB. As far as I may observe (different CPU and system) times are equal to drivers version 382,53.
At least here and in PG sieve.

I am using nVidia driver 384.111 on my Linux box with SuSE Leap 42.3. GTX 750 Ti.
Tullio

The slowdown was 1-5% depending on the type of work. Not huge but when you are only a small percentage faster than others, it all adds up lol.

As for the issues with PLX chip motherboards. I'll see if I can find the post and copy it here.

Some more general issues remain unfixed with the 411.70 update. Users of GTX 1060 cards connected to AV receivers will find that those devices will switch to two-channel stereo mode if they allow audio output to remain idle for five seconds. GTX 1080 Ti cards might exhibit random DPC watchdog violations when used as members of a multiple-GPU setup on motherboards with PLX switches. Cursor corruption might appear in Firefox when a user hovers over certain links.

Does anyone have any performance data with different driver versions? And if so, which driver versions are quickest? For windows, Geforce 10x0 series.

Does anyone have any performance data with different driver versions? And if so, which driver versions are quickest? For windows, Geforce 10x0 series.

Does anyone have any performance data with different driver versions? And if so, which driver versions are quickest? For windows, Geforce 10x0 series.

From what I've heard anything 400 and above is reported to be slower than anything 399 and below. I don't have data on this, just what people have been reporting.