I’m sorry if this is a stupid question, but I just don’t know who this is for. It seems like it’s for general desktop use, and graphic programs like Creative Cloud and applications like that. But I feel like it would be really expensive, and a company might as well give out high powered laptops instead. Don’t get me wrong, this is cool and it works great, I just don’t get what it’s for. Would it make sense for one man companies or home use?
Thanks for your time!
Fair question, the answer is anyone looking for acceleration in their enterprise virtualization. With virtualization we ask the CPU to run the host, be vCPU’s for the guests, manage all resources like storage and networking, and then with VDI we also ask it to be a GPU. This last is arguably the worst as GPU’s are massively parallel and CPU’s serial in nature, so very different. As a result, VDI has suffered since its onset with a less then optimal end user experience and GRID is answering that demand. Giving admins the ability to share or dedicate GPU in a virtualized architecture adds a powerful resource for them to change the end user experience. Even with Knowledge workers there are commons apps such as Google Earth that run far better with GPU, and Office 2013 is leveraging graphics acceleration now. Maybe more exciting, GRID allows the virtualization of applications that before simply could not be, making GRID a game changer for 3D and similar intensely graphical applications that demand GPU. By virtualizing those desktops and applications you unchain that critical user from a very expensive physical desktop and give them mobility, BYOD, while allowing the admins to centralize data, keep it secure. The benefits don’t end there, as an example, adding GPU changes your density math as it takes workload off the CPU. To your question about making sense for one person companies or home use, enough so that DaaS offerings are adding GPU to their available resources. As you said, it works great, and we are only moving towards a more graphical experience and GRID brings flexibility and freedom to those graphics.
I setup the first fully virtual Design & Engineering consultancy on the back of the launch of GRID and Citrix Xen7. I have ran traditional engineering consultancies but always been restricted by the team I could attract geographically located to my office and also the cost of having thousands of square feet of office space to house everyone in.
We launched in the start of 2014 and have been evangelically showing customers the way we deliver design solutions using CAD and Analysis engineers and software, from a team I have across the UK, gives them the best results at the best price as I haven’t got to pass on huge overheads.
This is not a solution for a “one man” operation as the investment and ongoing costs would be too great, however the cost to implement this way of working is not Blue Chip territory either. The money I save each month on offices, desks, coffee etc – not to mention power consumption (yew we are an environmentally friendly business now too) goes a long way towards the ongoing development.
My business is built on GRID technology – I wasn’t disappointed.
GRID and GPU pass through – dependence on graphic card
Hello,
I‘m looking for a solution, virtualizing several VMs which contain different GPU consuming applications. In contrast to the general focus of this grid concept, in our setup only one VM is running at a time and we could plug the monitor directly to the graphic card. In addition we have some space, heating and power restrictions, so the Grid K1 or K2 is no option for our approach.
I red in some forums that the GPU pass through is possible also for the nVidia Quadro but it seems not to be straight forward. I am somehow new to this topic and so I have several questions:
- According to the release note of the nVidia driver, GPU passthrough is also possible for Quadro K2000, K4000, etc. Can I rely on that?
- Is there dependence between the used virtualization solution (XenServer, VMWare, etc) and the graphic card?
- Is the grid technology also possible for KVM?