The World's First Hardware-Based GPU Virtualization Solution

AMD MxGPU is the world’s first hardware-based virtualized GPU solution, is built on industry standard SR-IOV (Single-Root I/O Virtualization) technology and allows up to 16 virtualized users per physical GPU to work remotely.

Unlike software virtualization offerings, AMD’s scalable VDI (Virtual Desktop Infrastructure) technology can better protect valuable data by using a dedicated share of local memory for increased security.

VMWARE DRIVER​ CITRIX DRIVER

PURE. VIRTUALIZED. GRAPHICS.

Predictable Performance

An AMD graphics card equipped with our MxGPU technology offers consistent, predictable performance. IT managers can easily configure these solutions to allow for up to 16 users on a single GPU. When the card is appropriately configured to meet the needs of the organization, users expect the same access to the GPU no matter their workload. There is no worry about any one user tying up the entire GPU while other users experience slow and unresponsive systems. Each user now has an equal share of the GPU to allow them to design, create, and execute their workflows.

AMD MxGPU virtualized graphics delivers the same performance across multiple virtual desktops with the SPECapc® SOLIDWORKS benchmark, while the performance difference between virtual machines using a competing solution is up to 68%. 1

Data Secure

Users will have peace of mind knowing that their valuable work and data cannot be viewed by other prying eyes. Unlike software virtualization, AMD’s hardware-based virtualization solution makes it extremely difficult for a hacker to break in at the hardware level, whereas with software virtualization, any breaches or holes in the underlying driver can be exploited to access a guest virtual machine in an unauthorized fashion. For organizations that highly value security, AMD’s MxGPU solutions are the clear option for their virtualized workstation needs.

Full Certifications

There are virtually no limits what can or can’t be done in VDI, with full OpenGL®, DirectX®, and OpenCL™ acceleration and full ISV certification for Dassault Systèmes SOLIDWORKS®, PTC Creo®, Siemens® NX™, and more.

AMD MxGPU virtualized graphics delivers superior performance over the competition in these applications, with up to 35% faster performance in SOLIDWORKS 2, up to 38% faster performance in PTC Creo 3, and up to 156% faster performance in Siemens NX 4.

RADEON PRO SOFTWARE AND VDI ECOSYSTEM

One Driver For All

AMD Multiuser GPU (MxGPU) virtualized graphics delivers workstation-class graphics with local desktop-like performance from the data center to virtually any device, virtually anywhere.

In bringing the power of a physical GPU to virtual environments, AMD MxGPU virtualized graphics uses the unified Radeon Pro Software Enterprise driver which lets users effortlessly run professional applications on both physical and virtual machines.

VMware Support

With support for VMware vSphere® 6.0 and higher, system admins will love how easy AMD MxGPU virtualized graphics is to set up. The new Radeon Pro Settings for VMware seamlessly integrated into VMware vCenter® allows for management of physical GPUs at the host and datacenter level with a simple and intuitive interface.

​Admins can quickly set up a VDI environment, rapidly deploy virtual GPUs to existing virtual machines, and enable hundreds of professionals with just a few mouse clicks depending on the number of GPUs in the system
configuration.

DOWNLOAD NOW GET THE USER GUIDE

Citrix Support

AMD FirePro™ S-Series cards support Citrix HDX 3D Pro technology, enabling IT departments to create virtual workstations for greater economies of scale, better security, and higher performance of graphics apps. Multi-user GPU (MxGPU) technology is now fully integrated with the Citrix’s HDX 3D Pro protocol. HDX 3D Pro enables users to deliver desktops and applications that leverage the GPU for hardware acceleration, including applications that require OpenGL®, OpenCL™, and DirectX® support.

DOWNLOAD NOW

AMD MxGPU Products

Recommended AMD FirePro™ S-series cards

AMD FirePro™ S7100X

Supports up to 16 concurrent users per card

Available on: Dell PowerEdge M630 (via Amulet Hotkey DXG-S7100X), HPE ProLiant WS460c Gen9

AMD FirePro™ S7150

Supports up to 16 concurrent users per card

Available on: Dell PowerEdge R730, Dell PowerEdge T630 (Actively Cooled solution), Supermicro SuperServer 1028GQ-TR(T), Supermicro SuperServer 1028GR-TR

AMD FirePro™ S7150 x2

Supports up to 32 concurrent users per card

Available on: Dell PowerEdge R730, Dell PowerEdge C4130, Dell PowerEdge T630, HPE ProLiant XL190r Gen9, HPE ProLiant XL250a Gen9, HPE ProLiant DL380 Gen9, Supermicro SuperServer 1028GQ-TR(T), Supermicro SuperServer 1028GR-TR

ADDITIONAL RESOURCES

Learn more about how MxGPU is the right solution for your business in:

Predictable, reliable infrastructure is paramount in an enterprise environment. AMD MxGPU graphics virtualization technology delivers leading Quality of Service (QoS) with performance that is up to nine times more consistent than the competition. 5 Learn more about our QoS advantage by downloading our whitepaper here.

To download your copy of the whitepaper, please complete your details.


GPU-accelerated virtual desktops are becoming more commonplace as the software we use becomes more demanding for graphics hardware and the ROI improves. Without proper selection and management of GPU resources for virtual desktop environments, the user experience when running these applications can be suboptimal, resulting in not only a bad experience but poor adoption or worse: a failed project.

Today, it’s more important than ever to factor in GPU resources to your desktop virtualization projects, but if you’re not GPU-savvy, you may need some help. In this white paper, we’ll take a look at the approach AMD uses so you can better understand what’s going on behind the scenes. The goal is to arm you with the information you need for success as you head into a GPU virtualization project.

To download your copy of the whitepaper, please complete your details.


AMD Multiuser GPU: Hardware-Enabled GPU Virtualization for a True Workstation Experience
To download your copy of the whitepaper, please complete this form.

  1. With driver 16.50, FirePro™ S7150 x2 MxGPU delivers same performance across VMs on SPECapc SOLIDWORKS benchmark, while the performance difference between VMs on Nvidia GRID for Tesla M60 is up to 68%. Testing conducted by AMD Performance Labs as of January 10th, 2017, under following configuration: Host: VMware ESXi 6.5.0. Virtual Machine: Intel Xeon E5-2650 v3 @2.30 GHz, 4 Logical Processors, 8 GB Physical Memory, 120 GB Storage, Windows 7 Ultimate 64 Bits SP1 Build 7601, 16.50 graphics driver for AMD FirePro™ S7150 x2 and 369.17 for Nvidia Tesla M60. For both AMD FirePro™ S7150 x2 and Tesla M60, only one physical GPU was involved for the benchmark. Tesla M60 was configured to use the 2Q profile. S7150 x2 MxGPU virtual GPU number was set to be four. In both configuration, one of the physical GPU is split into four virtual GPUs and assigned to four independent VMs. Another physical GPU is remained unused. The configurations for Tesla M60 and AMD FirePro™ S7150 x2 are equivalent. Benchmark Application: SPECapc SOLIDWORKS benchmarks under M60-2Q profile/AMD FirePro™ 7150 x2 MxGPU – 4 virtual GPUs configurations. AMD FirePro™ S7150 x2: Highest graphics composite in four VMs: 2.43. Lowest graphics composite in four VMs: 2.43. Performance difference: 2.43/2.43 = ~0%. Tesla M60: Highest graphics composite in four VMs: 2.67. Lowest graphics composite in four VMs: 1.59. Performance difference: 2.67/1.59 = ~67.92% difference. PC and server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers. FP-211
  2. AMD FirePro™ S7150 x2 delivers up to 35% faster performance on SPECViewperf 12.1 SOLIDWORKS benchmark than Nvidia Tesla M60. Testing conducted by AMD Performance Labs as of January 10th, 2017, under following configuration: Host: VMware ESXi 6.5.0. Virtual Machine: Intel Xeon E5-2650 v3 @2.30 GHz, 4 Logical Processors, 8 GB Physical Memory, 120 GB Storage, Windows 7 Ultimate 64 Bits SP1 Build 7601, 16.50 graphics driver for AMD FirePro™ S7150 x2 and 369.17 for Nvidia Tesla M60. Benchmark Application: SPECViewperf 12.1 SOLIDWORKS test under single GPU pass-through. AMD FirePro™ S7150 x2 graphics composite: 68.05. Tesla M60 graphics composite: 50.4. Performance Differential: 68.05/50.4 = ~35.02% higher score on AMD FirePro™ S7150 x2. PC and server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers. FP-208
  3. AMD FirePro™ S7150 x2 delivers up to 38% faster performance on SPECViewperf 12.1 PTC Creo benchmark than Nvidia Tesla M60. Testing conducted by AMD Performance Labs as of January 10th, 2017, under following configuration: Host: VMware ESXi 6.5.0. Virtual Machine: Intel Xeon E5-2650 v3 @2.30 GHz, 4 Logical Processors, 8 GB Physical Memory, 120 GB Storage, Windows 7 Ultimate 64 Bits SP1 Build 7601, 16.50 graphics driver for AMD FirePro™ S7150 x2 and 369.17 for Nvidia Tesla M60. Benchmark Application: SPECViewperf 12.1 Creo test under single GPU pass-through. AMD FirePro™ S7150 x2 graphics composite: 42.64. Tesla M60 graphics composite: 30.8. Performance Differential: 42.64/30.8 = ~38.44% higher score on AMD FirePro™ S7150 x2. PC and server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers. FP-210
  4. AMD FirePro™ S7150 x2 delivers up to 156% faster performance on SPECViewperf 12.1 Siemens NX benchmark than Nvidia Tesla M60. Testing conducted by AMD Performance Labs as of January 10th, 2017, under following configuration: Host: VMware ESXi 6.5.0. Virtual Machine: Intel Xeon E5-2650 v3 @2.30 GHz, 4 Logical Processors, 8 GB Physical Memory, 120 GB Storage, Windows 7 Ultimate 64 Bits SP1 Build 7601, 16.50 graphics driver for AMD FirePro™ S7150 x2 and 369.17 for Nvidia Tesla M60. Benchmark Application: SPECViewperf 12.1 Siemens NX test under single GPU pass-through. AMD FirePro™ S7150 x2 graphics composite: 40.48. Tesla M60 graphics composite: 15.83. Performance Differential: 40.48/15.83 = ~155.72% higher score on AMD FirePro™ S7150 x2. PC and server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers. FP-209
  5. 1. Testing conducted by AMD Performance Labs as of August 1st under following host and virtual machine configuration: Host: VMware ESXi 6.5.0 with host driver Radeon Pro VMware vSphere Driver– Revision 1.02 / Nvidia VMware ESXi 6.5 Host Driver 367.106. Virtual Machines: Intel(R) Xeon(R) E5-2698 v4 @2.20 GHz, 4 Cores, 4 Logical Processors, 8GB Physical Memory, 120 GB Storage, Windows 7 Enterprise 64-bit SP1 Build 7601 with Radeon Pro Software 17.30 driver/Nvidia GRID 370.12 guest graphics driver. Benchmark: SPECapc Solidworks simultaneously on 16 virtual machines. Nvidia GRID vGPU configuration: Each VM has 1 vGPU assigned from Nvidia GRID M60-0Q profile (16 vGPU on a physical GPU). AMD MxGPU configuration: Each VM has 1 vGPU assigned from 16 VFs configuration (16 vGPU on a physical GPU). Raw, normalized, AMD scores across 16 VMs: 0.994174757 0.994174757 0.994174757 1.025242718 0.994174757 0.994174757 0.994174757 1.025242718 0.994174757 0.994174757 0.994174757 0.994174757 1.025242718 0.994174757 0.994174757 0.994174757. Raw, normalized, Nvidia scores across 16 VMs: 1.181981982 1.066666667 0.893693694 0.922522523 1.066666667 0.922522523 0.922522523 0.893693694 1.066666667 0.951351351 1.181981982 0.836036036 0.951351351 1.124324324 0.951351351 1.066666667. FirePro™ S7150x2 graphics standard deviation: 0.00390. Nvidia Tesla M60 graphics standard deviation: 0.03618 Performance Differential: 0.03618/0.00390 = ~9.27x higher standard deviation on Nvidia Tesla M60. FirePro™ S7150x2 graphics performance predictability (inverse of standard deviation): 1/0.00390 = 256.41. Nvidia Tesla M60 graphics performance predictability (inverse of standard deviation): 1/0.03618 = 27.64. PC and server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers. RPW-17

© 2017 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, Radeon, LiquidVR, and combinations thereof are trademarks of Advanced Micro Devices, Inc. DirectX is a registered trademark of Microsoft Corporation in the US and other jurisdictions. OpenCL is a trademark of Apple Inc. used by permission by Khronos. OpenGL is a registered trademark of Silicon Graphics, Inc. used by permission by Khronos. Vulkan and the Vulkan logo are registered trademarks of Khronos Group, Inc. Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.

HDMI, the HDMI logo and High-Definition Multimedia Interface are trademarks or registered trademarks of HDMI Licensing, LLC in the United States and other countries.