One of the more
interesting announcements that I believe flew under the radar was the joint
VMware-NVIDIA initiative to bring virtual graphics processing unit technology
(vGPU) to VMware’s vSphere and the VMware cloud on AWS so it’s really a good
time to understand GPU passthrough concept.
CPU is good at
handling multiple tasks but a GPU can handle a few specific tasks very fast.
CPU is good for
serial processing and has low latency while GPU is good at parallel processing
and hence high throughput. In marvel comic terms, CPU is like Antman, GPU is a
hulk – incredible hulk!
okay, what is this thing
"GPU passthrough" :
It is a
technology that allows you to directly present an internal PCI GPU to a virtual
machine. The device acts as if it were directly driven by the VM, and the VM
detects the PCI device as if it were physically connected. VMware calls this
operation as VMDirectPath I/O.
So why do we need such
stuff:
To execute
workloads with more computer power, simply for faster time to results.
Who uses all these crazy things:
Gamers - for
smooth graphics,
Data
scientists, Artificial Intelligence, Machine / Deep learning engineers to
compute more data in a faster manner.
Interesting
isn't it? and let’s see how
do we enable it:
To Enable the
Host for GPU Passthrough:
1. Select the
Configuration tab on the host, and click Advanced Setting. (If the host has
devices enabled for passthrough, these devices will be listed here.)
2. To configure
passthrough for the GPU, click Configure Passthrough.
3. In the Mark
Devices for Passthrough window, check the box that corresponds to the GPU
adapter installed in the host.
4. Click OK.
The GPU should now be listed now in the Window on the Advanced settings page.
Note: If the device has an orange arrow
displayed on the icon, the host needs to be rebooted before passthrough will
function. If the device icon is green, passthrough is enabled.
To Enable the
Virtual Machine for GPU Passthrough:
1.Right-click
the virtual machine and select Edit Settings
2. Add a new
device by selecting PCI Device from the list, and click Next.
3. Select the
GPU as the passthrough device to connect to the virtual machine from the
drop-down list, and click Next.
4. Click Finish
5. Download and
install the drivers according to the Virtual Machine's OS.
6. Reboot the
virtual machine.
Quick links:
Blog link to
explore more about the topic - https://blogs.vmware.com/vsphere/2019/03/exploring-the-gpu-architecture-and-why-we-need-it.html
Comments
Post a Comment