Virtualization GPU hardware for COP OpenCL workload ?
Greatings.
I have a special request: I need some tips about virtualizing GPU hardware for OpenCL workloads from Capture One Pro
#1 workload (for 70% of the time): multiple "mini"-clusters for provisioning and failure tests; software development / cross-compiling with GNU/Linux distro ; and all day regular usage for office/internet stuff
#2 workload (for 30% of the time): dedicated application usage with Capture One Pro under Windows 10 guest OS (digital postprocessing/darkroom)
I know virtualization technologies mainly from VMware workstation, VirtualBox onto which I'm rely on since 2011 but not for Capture One, which running until now on dedicated hardware and operating system (COP7 MacOS X 10.7/Lion, COP11 MacOS X 10.12/Sierra). Actually I'm using an Apple MBP8.2 (year 2011) which is capable of running macOS 10.12 (Sierra) and COP11 together but in "slug" mode.
It would be fine to let OpenCL API calls passed-through directly to the the GP/GPU hardware resources.
I bought a lic for COP11 which has - of course - higher requirements in comparison and better feature set.
I'm in planing to buy in mid 2018 a preconfigured single socket, x86_64-based system with >=10C/20T_VCPU, =>128GiB RAM, NVMe Boot mirror and SSD boundary as secondary storage, 10gbps network infra and >= 1 OpenCL enabled graphics board with >= 4GB on-board memory, 27" UHD enabled graphics monitor (color space)
This means I planning to migrate vom MacOS X (old hardware) to Windows 10 (new hardware) running in a virtual machine with GPU passthrough, if possible.
Anyone maybe knows which Hypervisor options are suiteable to fulfill this requirements?
Is Xen still in use today and/or state-of-the-art?
Is ESXi w/ free lic for GPU passthrough an valuable option, maybe with business lic?
More alternatives or experiences?
I don't expect much answers but I should try my luck.
You and your beloved ones an Happy Christmas time.
Thanks.
I have a special request: I need some tips about virtualizing GPU hardware for OpenCL workloads from Capture One Pro
#1 workload (for 70% of the time): multiple "mini"-clusters for provisioning and failure tests; software development / cross-compiling with GNU/Linux distro ; and all day regular usage for office/internet stuff
#2 workload (for 30% of the time): dedicated application usage with Capture One Pro under Windows 10 guest OS (digital postprocessing/darkroom)
I know virtualization technologies mainly from VMware workstation, VirtualBox onto which I'm rely on since 2011 but not for Capture One, which running until now on dedicated hardware and operating system (COP7 MacOS X 10.7/Lion, COP11 MacOS X 10.12/Sierra). Actually I'm using an Apple MBP8.2 (year 2011) which is capable of running macOS 10.12 (Sierra) and COP11 together but in "slug" mode.
It would be fine to let OpenCL API calls passed-through directly to the the GP/GPU hardware resources.
I bought a lic for COP11 which has - of course - higher requirements in comparison and better feature set.
I'm in planing to buy in mid 2018 a preconfigured single socket, x86_64-based system with >=10C/20T_VCPU, =>128GiB RAM, NVMe Boot mirror and SSD boundary as secondary storage, 10gbps network infra and >= 1 OpenCL enabled graphics board with >= 4GB on-board memory, 27" UHD enabled graphics monitor (color space)
This means I planning to migrate vom MacOS X (old hardware) to Windows 10 (new hardware) running in a virtual machine with GPU passthrough, if possible.
Anyone maybe knows which Hypervisor options are suiteable to fulfill this requirements?
Is Xen still in use today and/or state-of-the-art?
Is ESXi w/ free lic for GPU passthrough an valuable option, maybe with business lic?
More alternatives or experiences?
Recommended system requirements (COP11)
If working with high-resolution images and doing calculation-heavy adjustments we
recommend the following:
• Intel CPU with 4+ cores
• 16+ GB of RAM
• Solid State Disk (SSD)
• Dual (matching) Graphics card from NVIDIA or AMD with 4GB+ RAM per card
Note: Due to the signifi cant additional calculation overhead, systems using 4k/5k monitors
will require additional resources in addition to the above recommended spec (both in
GPU and CPU power). For a Professional experience we recommend at least doubling the
specifi cation above for these confi gurations.I don't expect much answers but I should try my luck.
You and your beloved ones an Happy Christmas time.
Thanks.
0
Post ist für Kommentare geschlossen.
Kommentare
0 Kommentare