OK, I have the latest nvidia cuda SDK 8.0, latest Visual Studio Free, Windows 10 with the latest updates, not Insider, drivers for GPU are updated as well. Cumputer has been formatted a couple days ago, and nothing besides a Firefox is installed. I am telling all this because I am not sure what am I supposed to mention. I have 2 projects, on the first one I am using nvidia cuda VS project, the second one is Console application. In the second (console) project, I am allocating with malloc 12500x12500 float elements array, and since my CPU has 8 threads (I7 4790k), I start 8 threads, in each thread I am filling the array (of course each thread fills every 8th row) with ones (1), but the value is not the problem, can be anything else other than 1. In the first (nvidia cuda) project, I am doing the same on the GPU (GTX970), and with the CPU with threads, but the first project crashes when it gets to the CPU threads part, but GPU part runs normally. Only thing is, when I reduce the array to only size less than or equal to 10000x10000 floats, both CPU and GPU are working in the first (cuda) project. And As soon as I increase the size, CPU part is crashing, although, literally the same code running in console application is working (I have created the console application project just to verify if it works). One more thing to mention, if I write something like write to allocated memory only if first dimension is greater than 5300 and the second is greater than 2000, it works, the problem are the smaller values (when I try to do it with more than 10000x10000 elements), but if it is less than this, I do not need to write only with previous statement, it works for all elements. Do I need to setup anything in Visual Studio in order for this to work? Thank you very much in advance.