问题
On compilation of the CUDA SDK, I'm getting a nvcc fatal : Unsupported gpu architecture 'compute_20' My toolkit is 2.3 and on a shared system (i.e cant really upgrade) and the driver version is also 2.3, running on 4 Tesla C1060s
If it helps, the problem is being called in radixsort.
It appears that a few people online have had this problem but i havent found anywhere that actually gives a solution.
回答1:
I believe compute_20 is targeting Fermi hardware, which you do not have. Also, Cuda 2.3 doesn't support Fermi features or compute_20. Try using compute_13 instead.
回答2:
You are compiling the 3.x SDK with a 2.3 Toolkit, that won't work since as you've now seen compute_20 is not supported.
The simplest solution is to either upgrade your Toolkit or downgrade your SDK. You could try changing the makefile (common.mk in C/common) on Linux or changing the Build Properties on Windows to select only compute_13, but there may be other problems so it may be simpler to align the SDK/Toolkit versions.
Note that if you don't have administrator priviledges then you can install a local version of the Toolkit alongside the SDK but you would need admi to upgrade the driver.
来源：https://stackoverflow.com/questions/3047909/nvidia-cuda-sdk-examples-compilation-unsupported-architecture-compute-20