News for hackers. Stuff that matters.
Re:Confidentiality Issues on a GPU in a Virtualized Environm... by iamsalman on Monday February 09, 2015 @02:29PM (#9023124)
Currently, only AWS offers GPU instances. There are other, smaller providers like Nimbix or Peer1 but they offer bare-metal instances. With AWS, they currently have Grid K520 GPUs which is targeted for public and private Clouds.

The gist of this paper is that non ECC memory does not zero the memory but it conveniently skips the fact that public or private Cloud uses either NVIDIA Tesla or Quadro series cards, both of which have ECC (one of the main reasons they're expensive than say GTX Titan which offers same single/double-precision performance but does not offer ECC). No one really uses GeForce series cards for anything serious and definitely not virtualization so even though this is a vulnerability, it may not effect more than 1e-10% of virtualized GPUs out there

Re:Confidentiality Issues on a GPU in a Virtualized Environm... by maccard on Monday February 09, 2015 @12:12PM (#9022267)
Noticed this with CUDA before. When I tried to render data that lay outside the buffers I was working on, I would get textures of my web browser appearing instead of what I expected.
Re:Confidentiality Issues on a GPU in a Virtualized Environm... by j_s on Monday February 09, 2015 @11:28AM (#9021986)
TL;DR: We found that the GPU global memory is zeroed only in some configurations.
Re:Confidentiality Issues on a GPU in a Virtualized Environm... by lmeyerov on Monday February 09, 2015 @11:57AM (#9022174)
I've been thinking increasingly that webcl's place is in the cloud, including because of all the work on multiuser/multitenant confidentiality+integrity+availability!