How to allocate whole memory

Accelerated Computing CUDA CUDA Programming and Performance
1

How is it possible to allocate global memory as much as possible?
Out of memory exceptions can be caught but after reducing the memory to allocate i always get a WIN32 Runtime exception!
Allocating the global memory of the Quadro FX4800 is working but the Tesla C1060 doesn’t.

SDK 2.0
Tesla C1060
Quadro FX4800

2

Problem seems to be fixed with the latest driver (181.20) External Image .