How do I get the GPU clock rate in CUDA 13?

I updated CUDA to version 13.
But it seems that cudaGetDeviceProperties has changed.

Instead of returning the cudaDeviceProp struct with clockRate, it returns a mutilated version thereof with deprecated fields missing.

struct __device_builtin__ cudaDeviceProp
{
    char         name[256];                  /**< ASCII string identifying device */
    cudaUUID_t   uuid;                       /**< 16-byte unique identifier */
    char         luid[8];                    /**< 8-byte locally unique identifier. Value is undefined on TCC and non-Windows platforms */
    unsigned int luidDeviceNodeMask;         /**< LUID device node mask. Value is undefined on TCC and non-Windows platforms */
    size_t       totalGlobalMem;             /**< Global memory available on device in bytes */
    size_t       sharedMemPerBlock;          /**< Shared memory available per block in bytes */
    int          regsPerBlock;               /**< 32-bit registers available per block */
    int          warpSize;                   /**< Warp size in threads */
    size_t       memPitch;                   /**< Maximum pitch in bytes allowed by memory copies */
    int          maxThreadsPerBlock;         /**< Maximum number of threads per block */
    int          maxThreadsDim[3];           /**< Maximum size of each dimension of a block */
    int          maxGridSize[3];             /**< Maximum size of each dimension of a grid */
    //int clockRate is now missing...
    size_t       totalConstMem;              /**< Constant memory available on device in bytes */

How do I get the clockRate from CUDA 13, now that NVidia has decided to remove these data members from the cudaDeviceProp struct?

Edit: looking at the old 12.3 data struct, I noticed that the missing fields were deprecated.

Solution

clockRate field was indeed depracated (used with cudaGetDeviceProperties) in CUDA 12.x, and removed in version 13.

In order to get the clock rate (in KHz) in version 13, you can do one of the following:

The simplest, as @paleonix commented, is to use cudaDeviceGetAttribute, with cudaDevAttrClockRate for the attr parameters.

(This is mentioned in the CUDA 13 release notes - see the section Removed Fields and Their Replacements):
```
int clockRateKHz;
cudaDeviceGetAttribute(&clockRateKHz, cudaDevAttrClockRate, 0);  // 0 here is the device number
// do what you need with clockRateKHz
```
^{Note:

I omitted checking the cudaError_t status result for simplicity.

This should be done in the actual code.}
Altenatively, you can use cuDeviceGetAttribute from the driver API (requires #include <cuda.h> and link with driver library - cuda.lib on Windows).

You need to use CU_DEVICE_ATTRIBUTE_CLOCK_RATE for the attrib parameter as shown below:
```
#include <cuda.h>

// ...

cuInit(0);
CUdevice device;
cuDeviceGet(&device, 0); // Or the index of the device you're interested in

int clockRateKHz;
cuDeviceGetAttribute(&clockRateKHz, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, device);
// do what you need with clockRateKHz
```
^{(again omitted checking the result of the API calls (with the CUresult status code) for simplicity).}