Posts: 5,185
Threads: 230
Joined: Apr 2010
Hey Guys,
just wanted to inform you about a funny new message that the NVCC spits out after updateing to CUDA 6.5:
Quote:nvcc warning : The 'compute_11', 'compute_12', 'compute_13', 'sm_11', 'sm_12', and 'sm_13' architectures are deprecated, and may be removed in a future release.
That means NVidia will drop CUDA support for (example) the old GTX series like my gtx285. They will no longer be supported in a future CUDA versions.
Actually it is a lot of GPU's that are using the sm_1* model. See here to find out which SM_* version is yours:
http://developer.nvidia.com/cuda-gpus
--
atom
Posts: 2,936
Threads: 12
Joined: May 2012
If you're not sure what atom means by "SM_*", this is what is called "Compute Capability" on that page atom linked to. So if your GPU has a "Compute Capability" of 1.x, then you will need to buy a new GPU to continue using oclHashcat.
Posts: 247
Threads: 59
Joined: Mar 2011
Do you need to update to each new CUDA SDK?
Posts: 5,185
Threads: 230
Joined: Apr 2010
Yeah kernel compiler is part of the SDK.
Posts: 247
Threads: 59
Joined: Mar 2011
So to keep up with newer cards, you need the newer CUDA SDKs, and can't stay with an older version?
Is that a marketing decision, to force people to buy newer cards, or is there a valid engineering reason?
Posts: 5,185
Threads: 230
Joined: Apr 2010
Huh? I'm not sure if I understand you correctly. I'm not involved in NV's decision making nor do I know why they do the things they do.
Posts: 247
Threads: 59
Joined: Mar 2011
Do you have to use CUDA 6.5, or can you stay with older SDKs?
Or can you use both older and newer?
Posts: 344
Threads: 2
Joined: Aug 2011
To do as you want atom would have to maintenance two different cuda versions.
Posts: 247
Threads: 59
Joined: Mar 2011
So he has no choice but to abandon the older cards?
Which means owners of older cards have to buy newer cards, which sounds like something that NVIDIA's marketing department would like.
Anyway, they are getting better, with more "bang for the buck."
Posts: 2,936
Threads: 12
Joined: May 2012
The cards in question are 6-8 years old and are from the CUDA 1.0 days. There always comes a point where the technology moves beyond a certain point, and you have to deprecate the old stuff.
You've been around awhile, so you likely recall when AMD users experienced this exact issue last year when AMD dropped support for the old Radeon HD 4000 series from the same era.
Time to upgrade.
