AEG AEG F 44860…

Here are the answers:

“Does it work on a virtual Linux system via VMWare in a server environment?”

To achieve their high perfomance, our parallel algorithms designed for GPUs have to run on actual GPU hardware. Since usually a VM does not have (exclusive) control over the hardware it runs on, it cannot guarantee that the GPU is available for a specific programm.
So, no, even if the software may “work” on a virtual machine, you won’t get the GPU speedup.

“Does is work on an office PC with a low-end graphics card with passive cooling?”

In principle, yes. However, there are some minimum hardware requirements, depending in part on your OLAP model:
First of all, at present we only support Nvidia GPUs that are CUDA-capable (which sort of means that non-graphics algorithms run on them). They also need to have “compute capability 1.3″, which means they can handle double precision floating point arithmetics.
Finally, the total graphics memory on the card(s) needs to be large enough for the data to be processed. We do hold (part of) the cube data in graphics memory. Hence, the larger your cubes, the bigger your required GPU memory.
What does this mean in practice? — We are currently doing a lot of our own development and tests on cards of the GTX-2xx series, which come with up to 2 GB graphics memory and start as low as EUR 200,-. This may not be low-end, but it’s what most people nowadays have in their home computers. (You may have to beg to get it in your office desktop, though. So it’s a valid question.)

“Does it work on a terminal PC via web-interface, either virtual
desktop or browser-based?”

Yes, if you were asking about client machines. Since our current parallel algorithms are only part of the SERVER software, it does not matter what front-end you use to access the server, be it web-based, via Excel spreadsheet, or otherwise.