AMD Unveils World's First Hardware-Based Virtualized GPU Solution at VMworld

[btarunr]

AMD today at VMworld 2015 demonstrated the world's first hardware-based GPU virtualization solution, the AMD Multiuser GPU. This new solution from AMD enables a virtualized workstation-class experience with full ISV certifications and local desktop-like performance. With the AMD Multiuser GPU, IT pros can easily configure these solutions to allow up to 15 users on a single AMD GPU. Demonstrations of AMD virtualization solutions can be found at VMworld 2015 booth 447.

"The AMD graphics cards are uniquely equipped with AMD Multiuser GPU technology embedded into the GPU delivering consistent and predictable performance," said Sean Burke, AMD corporate vice president and general manager, Professional Graphics. "When these AMD GPUs are appropriately configured to the needs of an organization, end users get the same access to the GPU no matter their workload. Each user is provided with the virtualized performance to design, create and execute their workflows without any one user tying up the entire GPU."

Built around industry standard SR-IOV (Single Root I/O Virtualization) technology, the AMD Multiuser GPU continues AMD's embracement of non-proprietary open standards. SR-IOV is a specification developed by the PCI SIG, and provides a standardized way for devices to expose hardware virtualization. The AMD Multiuser GPU is designed to preserve and support graphics- and compute-accelerated features for design and manufacturing or media and entertainment applications. The AMD Multiuser GPU addresses limitations of current virtualized GPU solutions that may not provide predictable performance for CAD/CAE, Media and Entertainment, and general enterprise GPU needs.

Created for GPU-accelerated workflows such as GPU compute and OpenCL, the AMD Multiuser GPU is designed to overcome the limitations of software-based virtualization such as reduced end-user performance. Users have access to native AMD display drivers for OpenGL, DirectX and OpenCL acceleration, enabling work without restrictions.

Additional AMD Multiuser GPU features include:

• Compute capabilities based on OpenCL supported by industry leading GPU virtualization
• Full feature set support including DirectX 12 and OpenGL 4.4
• OpenCL 2.0 acceleration support
• Stable, predictable performance
• User-owned share of local memory providing additional security
• 15 maximum users per physical GPU

AMD Multiuser GPU is designed to work on environments using VMware vSphere/ESXi 5.5 and up, with support for remote protocols such as Horizon View.
AMD strives to provide world-class professional graphics hardware for companies desiring to implement virtualized environments. To help IT managers with the installation process and ongoing support, there are no additional per-seat software costs from AMD beyond the purchase of the AMD Multiuser GPU.

View at TechPowerUp Main Site

 

[RejZoR]

So, if I understand this right, it allows multiple users share graphic workload on a single GPU (or a cluster of them)?

 

[the54thvoid]

Is this the same sort of thingy ma bob?

http://www.techpowerup.com/215650/nvidia-grid-2-0-launches-with-broad-industry-support.html

 

[RejZoR]

I'd say yes.

 

[Musaab]

Every Time nVidia make something new (Really New) AMD start digging in the open standards (New and Old) and try to rebrand these standards just to take small share of nVidia's success. then they tell the world don't use what nVidia offer take ours because we are the open standard. Some times they give what they do for free like the FirePro w10000 in servers hopping some big fish will take the bate and leave Tesla. Let us see!

 

[ZoneDymo]

Every Time nVidia make something new (Really New) AMD start digging in the open standards (New and Old) and try to rebrand these standards just to take small share of nVidia's success. then they tell the world don't use what nVidia offer take ours because we are the open standard. Some times they give what they do for free like the FirePro w10000 in servers hopping some big fish will take the bate and leave Tesla. Let us see!

what "really new" things would those be?

 

[ne6togadno]

Every Time nVidia make something new (Really New) AMD start digging in the open standards (New and Old) and try to rebrand these standards just to take small share of nVidia's success. then they tell the world don't use what nVidia offer take ours because we are the open standard. Some times they give what they do for free like the FirePro w10000 in servers hopping some big fish will take the bate and leave Tesla. Let us see!

nvidia pr from 30.08.2015. amd pr from 01.09.2015
those amd guys has to dig with light speed to come up in 2 thays with their solution.
most likely job is done by VMware with amd and nv parthering with them for hardware compability.
rest is just nv marketing rushing pr on sunday so they can be first on news train

 

[Musaab]

Sorry I was talking about GRID 1 GRID 2 is an update. So we are talking about one generation not one day.

nVi

what "really new" things would those be?

NVIDIA holds 70% of graphics IPs so there should be some new. Or maybe they are rebranding just like free sync.

 

[Cybrnook2002]

Would be pretty cool for my server at home.

 

[Disparia]

AMD is touting their "hardware" solution as more secure and a better experience because of how they implement vGPU. That each user gets a dedicated fraction of the resources and all features are exposed (OpenGL, OpenCL, etc) as if you had a local GPU. Many of the comparisons were done against GRID 1 so we'll have to wait for future reviews to see how AMD Multiuser GPU stacks up against GRID 2.

 

[Cybrnook2002]

Man, it's just nice to have options.

 

[Musaab]

AMD is touting their "hardware" solution as more secure and a better experience because of how they implement vGPU. That each user gets a dedicated fraction of the resources and all features are exposed (OpenGL, OpenCL, etc) as if you had a local GPU. Many of the comparisons were done against GRID 1 so we'll have to wait for future reviews to see how AMD Multiuser GPU stacks up against GRID 2.

nVidia GRID has been built around Kepler GPUs more than two years ago and it was first dedicated unit for vGPU so it's good to know that AMD is less than three years behind nVidia . That keep some hope that AMD can keep fight for another day, Hopefully days, months or even years.

 

[Steevo]

nVidia GRID has been built around Kepler GPUs more than two years ago and it was first dedicated unit for vGPU so it's good to know that AMD is less than three years behind nVidia . That keep some hope that AMD can keep fight for another day, Hopefully days, months or even years.

I think the difference here is the number of users, the use of hardware to do the work instead of an abstraction layer running on the CPU, complete access to the GPU resources instead of merely allocating it at the driver level, and the complete access to GPU (DX, GPGPU, OPENcl, etc...) instead of just a few walled garden interfaces.

 

[Musaab]

I think the difference here is the number of users, the use of hardware to do the work instead of an abstraction layer running on the CPU, complete access to the GPU resources instead of merely allocating it at the driver level, and the complete access to GPU (DX, GPGPU, OPENcl, etc...) instead of just a few walled garden interfaces.

nVidia call their shaders (CUDA Cores), Why? Because their GPU is two layers (hardware and software) so their driver is a software layer not an interface like AMD drivers.

 

[Steevo]

nVidia call their shaders (CUDA Cores), Why? Because their GPU is two layers (hardware and software) so their driver is a software layer not an interface like AMD drivers.

Dafuq did I just read?

No GPU is two layers, its hardware with microcode, and either a direct hardware programmable interface that exposes hardware state and allows software to directly access the hardware. OR its hardware with microcode, and a software driver that acts as interface between the hardware and the software attempting to access it.


Nvidia and AMD hardware is very similar at this point.

 

[truth teller]

after yesterdays events, nbadia had to recruit some more people to shill on the public forums so they can damage control for them.

how is that working out for you @Musaab? won't they ask you why havent you used an older account to shill so you don't stand out so much like that (instead of going at it with a fresh account)? will they still pay you after this rookie mistake? btw how much are they paying you to "promote" their stuff?

 

[Musaab]

It's pity to hear some thing like that from kids. By any way if you are a man worked in the middle east in construction and manufacturing there is a chance you worked for me once. No need for your money but if it's of any use go buy yourself some manners if there is still some in the market. If you don't have money I'll talk to NVIDIA to give you some they sure have alot.
By the way I have born with this name. No need to hide in the bushes like you.
This will be my last comment in this article so be free to write what ever you want.

 

[Frick]

Good grief this is turning into Ars Technica.

 

[Lionheart]

after yesterdays events, nbadia had to recruit some more people to shill on the public forums so they can damage control for them.

how is that working out for you @Musaab? won't they ask you why havent you used an older account to shill so you don't stand out so much like that (instead of going at it with a fresh account)? will they still pay you after this rookie mistake? btw how much are they paying you to "promote" their stuff?

It's pity to hear some thing like that from kids. By any way if you are a man worked in the middle east in construction and manufacturing there is a chance you worked for me once. No need for your money but if it's of any use go buy yourself some manners if there is still some in the market. If you don't have money I'll talk to NVIDIA to give you some they sure have alot.
By the way I have born with this name. No need to hide in the bushes like you.
This will be my last comment in this article so be free to write what ever you want.

 

[theJesus]

I can't wait for the day when this sort of stuff can be found off-lease on eBay for cheap. I'd so love to put something like this in my vSphere server and then I could replace almost all of my machines with thin-clients. The only thing I think I might still want a dedicated machine for is a digital audio workstation (not that I have time to make music these days).