The API is open, It seems just to call the method NVAPI_INTERFACE NvAPI_GPU_GetPstates20   (   __in NvPhysicalGpuHandle    hPhysicalGpu, __inout NV_GPU_PERF_PSTATES20_INFO *    pPstatesInfo

http://docs.nvidia.com/gameworks/content/gameworkslibrary/coresdk/nvapi/group__gpupstate.html#gaeffe0838ca9850b9984fa9be117f637e

Detailed Description

The GPU performance state APIs are used to get and set various performance levels on a per-GPU basis. P-States are GPU active/executing performance capability and power consumption states.

P-States range from P0 to P15, with P0 being the highest performance/power state, and P15 being the lowest performance/power state. Each P-State maps to a performance level. Not all P-States are available on a given system. The definition of each P-States are currently as follows:

P0/P1 - Maximum 3D performance
P2/P3 - Balanced 3D performance-power
P8 - Basic HD video playback
P10 - DVD playback
P12 - Minimum idle power consumption


..
Here is the method to call:


NVAPI_INTERFACE NvAPI_GPU_GetPstates20   (   __in NvPhysicalGpuHandle    hPhysicalGpu,
__inout NV_GPU_PERF_PSTATES20_INFO *    pPstatesInfo
)      
DESCRIPTION: This API retrieves all performance states (P-States) 2.0 information.

P-States are GPU active/executing performance capability states. They range from P0 to P15, with P0 being the highest performance state, and P15 being the lowest performance state. Each P-State, if available, maps to a performance level. Not all P-States are available on a given system. The definition of each P-States are currently as follow:

P0/P1 - Maximum 3D performance
P2/P3 - Balanced 3D performance-power
P8 - Basic HD video playback
P10 - DVD playback
P12 - Minimum idle power consumption
TCC_SUPPORTED

Since:
Release: 295
SUPPORTED OS: Windows XP and higher

Parameters:
[in]   hPhysicalGPU   GPU selection
[out]   pPstatesInfo   P-States information retrieved, as documented in declaration above
Returns:
This API can return any of the error codes enumerated in NvAPI_Status. If there are return error codes with specific meaning for this API, they are listed below.

still this is hot, I have 2 gtx980 in my tower (750 and 780 outside) and it doesn't go that high...
but fans run higher too (~90% for the 980), but for the 750ti this is really anormal, I never saw that card going higher than 60°C (outside or inside the tower).
You should add some autofan to ccminer (or use msi ab)

My 970 Gigabyte G1's are never colder than 71 while mining.  That's not normal?  I have plenty of fans, custom water cooling on CPU, and fans are set to 75-100% most of the time.  I used to run with the side panel off so it would stay around 75C, but it only goes up to 78C with it on so I just set the temp limit in Precision X to 77C(95% power) and stick with that.

Even with a large fan blowing directly at them it stays around 70.  Pretty sure those temps are completely normal.

G1's have the custom Windforce setup which should be plenty of cooling. I'm sure it would go down to maybe 65 degrees if they weren't inside the case and set up for using SLI sometimes, but I can't imagine they'd ever be cooler than that.  All of my 750s and 750ti's run 50-61 most of the time so it seems normal to me that the 970 would be about 10-20C more.

Is there and AMD version of this? I skimmed through the other post and didn't see an answer. 

thank you Epsylon3 for this miner, i'm testing 1.5-git and found 2 issues:
while mining.set_extranonce feature is working fine with nicehash and has no side effects with trademybit.com, it breaks mining on yaamp.com (after the first share no more yays).
second, the hash rate rate has decreased ~200khs on all x## algos compared with version 1.4.9 on sm_30 (gtx660), do you think it's caused by changes made in the calculation of the average hash rate or by recent sm_50 optimizations?

3be08ad9445ddc7e594c916e5cc85bef2b58af66

was looking into blake256, what exactly do crcsum ?
Looks to me as some useless calculation...

pdata changes only when you request a new block

regarding share/block issue, I don't know... A block is the same as a share except that the difficulty is higher
(meaning more digits should match...)
So assuming the algo is tuned to find shares, you should find also blocks.
(may-be you should try on testnet and see if it happens)

edit: but yeah there is this strange thing when you swap h[6] and not h[7] may-be it explains a little the problem.