Hello all,
I am currently looking into moving from Proxmox to Looking Glass for the sake of being able to use the host OS as a daily driver (I miss my Linux). Due to my complex setup, looking glass seems to be the only possible way to do this. What is the current state of it? I see some commits have been made somewhat recently on github, but a lot of it looks like it hasn’t changed since it was originally pushed. I see a thread about using some new features on the Quadro from early 2019.
Ultimately, I need to know:
Has Looking Glass overcome the bugs mentioned in the original 2018 video?
Are there currently any known major bugs? If so, is it under active development to resolve them?
Are any GPUs recommended/not recommended? I currently have a 1070 for the guest and a 710 for the host.
I don’t want to move into something that isn’t stable.
This reminds me, any plans on Mac guest support? Friend of mine want to get into iOS development but would rather avoid paying the 1200 bucks extra this would cost.
I read some months ago about AMD looking into VFIO, nothing came from those investigations then?
OSX support requires someone to write a IVSHMEM driver for the host application and the OSX specific capture code for the host application. I don’t know enough about OSX to know how feasable this is.
AMD and I were looking into them, but AMD have gone quiet.
Sorry for the obtuse question. For eg, if the host is an RTX4000 and the guest is an RTX8000, could one game at a resolution higher than 1920x1200 comfortably? (i.e with high updates per second?)
Ah, I see. I wonder how much work that would be - from what I understand Apple is very strict about the drivers they allow, then again I could confuse it with iOS. Since kernel is based on FreeBSD though, maybe a FreeBSD host could be a first good step? Then again, it forked like 20 years ago, so doubt much resemblance is left there.
I’ll see if I can dig up a Mac person willing to dig into this. Seems like an interesting problem for sure, if you want to learn about the guts of MacOS.
Then again, there are probably zombie dragons lurking in there. Who knows what’s in there after all this time…
From what I understand one could just compiled and install the driver themself. I am just not prepared at this time to support another platform nor do I have an indepth understanding of the OSX landscape.
As for FreeBSD, nope sorry:
Mac OS X is based upon the Mach kernel. Certain parts from FreeBSD’s and NetBSD’s implementation of Unix were incorporated in Nextstep, the core of Mac OS X. See http://en.wikipedia.org/wiki/History_of_Mac_OS_X
However, OS X Leopard is UNIX-compliant with some sort of certification. OS X does have FreeBSD’s virtual file system, network stack, components of its userspace and few other stuff. See apples on webpage about UNIX technology that is included with OS X - http://www.apple.com/macosx/technology/unix.html
So only a portion came from FreeBSD, it’s not based on it.
No, they have implemented a recovery routine into the AMDGPU driver, they have not fixed reset. It’s useless to us as we need to reset the GPU outside of any driver so the guest can use it.
Ah, I see. Then it might be time to poke them again post vacation times, I think AMD want to solve this but have not been able to find the time to allocate due to lockdowns and what not.
AMD are not interested in fixing this… not only have my contacts gone silent, I was contacted just today by a cloud compute company who have been poking AMD continually for a fix for this too, with zero movement desperate for a fix (or workaround) and willing to pay for it.
Even still though, I am having another look through the amdgpu driver for any changes that may help us…
Tell them they should do a post in the subreddit explaining what they need and why it’s so important that they’d throw money at it, like what gnif did a while back. Since begging and pleading aren’t working maby some negative attention might put it back on the radar. Even better if those companies did a post together.
I mean, if AMD can’t fix current cards that’s ok I get it, but let me know that next gen is going to be able to deal with it.
As it stands, the next video card I buy will likely be nvidia because at least you can make it work. Which is a shame because they are outright hostile to filthy peasants using vfio.
has anyone experimented with a powered riser
something like this but with a power killswitch added it
I think(?) hotplug is implemented data wise but not power wise
if successful it could become a workaround for AMD to use on Matx boards in ATX cases
kinda ghetto outside of the box solution but some people don’t mind that kinda thing
Huh. No idea why they want to leave money on the table like this, if it was a HW issue you would reckon they’d come out and say so. Strange…
Completely, 100% in agreement. We need to let them know this is costing them serious $$$$$ at the end of the day. Raise the stink in about a month or so when people are back from summer holidays.
