Compute Capability 2.0 Texture Size Limits

We have an app that uses 3D textures for volume processing. We want to use very large textures with more than 2048 elements in depth. According to the docs compute capability 2.0 cards support 3D textures of size 4096x4096x4096. We have a Tesla C2050 and unfortunately it is not reporting the increased 3D texture size. It is reporting the old 1.3 capability of 2048x2048x2048. We are using the latest cuda (3.1) with the latest driver (257.21) on a windows 7 machine. We are using the driver API. Here is what is reported about the card:

Tesla C2050
total memory = 3181969408
clock rate = 1147000
driverVersion = 3010
computeCapability = 2.0
maxThreadsPerBlock = 1024
maxThreadsDim = (1024,1024,64)
maxGridSize = (65535,65535,1)
sharedMemPerBlock = 49152
totalConstantMemory = 65536
SIMDWidth = 32
memPitch = 2147483647
regsPerBlock = 32768
textureAlign = 512
peakClockRate = 1147000
multiprocessorCount = 14
maximumTexture2dWidth = 65536
maximumTexture2dHeight = 65536
maximumTexture3dWidth = 2048
maximumTexture3dHeight = 2048
maximumTexture3dDepth = 2048
concurrentKernels = 1
kernelExecTimeout = 1
eccEnabled = 0

Is this a driver bug? Or is the documentation wrong?
-Derek

See:
http://forums.nvidia.com/index.php?showtop…p;#entry1071814

See:
http://forums.nvidia.com/index.php?showtop…p;#entry1071814

Thanks for this link. I really did try to search the forums :*)

When you say “future release” does that mean 3.2 or is it not determined? This is an important feature for our product (medical imaging of large volume datasets).

-Derek

Thanks for this link. I really did try to search the forums :*)

When you say “future release” does that mean 3.2 or is it not determined? This is an important feature for our product (medical imaging of large volume datasets).

-Derek

Hopefully for 3.2, but no promises!

If this is important for your product, please file a bug to help us prioritize this feature.

Hopefully for 3.2, but no promises!

If this is important for your product, please file a bug to help us prioritize this feature.