EposVox got “Early Access” to closed beta builds of OBS and press closed beta drivers for Turing with NvFBC enabled. It’s one of his recent videos where he received a RTX 2080 FE from Nvidia after undoubtedly signing a few NDAs.
He said something along the lines of…
The old method involved copying GPU buffers to system memory, taking up more resources. The new Beta OBS allows the GPU buffers to remain on the card.
It’s NvFBC.
Interesting, i’d love to see how they are doing this. It’s almost certainly using the same method we already know about. Since OBS is GPL2, if there is a binary blob in there might be able to lift it into LG and leverage it.
Currently the beta releases with the blob are still limited to “influencers.” Jim is only likely to add the code to the release OBS once the next version is released, and it might be Windows only. Building from Git likely won’t get you the blob, since Jim collaborated with Nvidia after signing a few NDAs himself to make a special beta build.
Wasn’t the first time Jim has made non-git beta builds. AGDQ uses non-git OBS beta builds direct from Jim for their speedrunning convention and week long charity livestream.
We primarally want it for windows anyway, so no problems there. If he releases a build, even a limited beta of OBS without providing the source code on request he will be violating the GPL2 license.
That goes against the NDA he signed, so he might have modified the license for those off-git builds to respect the NDA.
The binary blob can be included and used, it wont be breaching the GPL2, but providing a build with the blob and not releasing the source to build it to use the blob will be in breach.
Huh, I truly wonder how the NDA will cope with this if early access to influencers is privileged access the public is not supposed to have.
Since Jim is the copyright holder he can change the license agreement of his code, however all contributors to the project would have to agree to the change.
NVIDIA directly approaching him might be a big enough influence on the project, especially if there was a deal with money behind it.
well lets hope that it’s released, there would be no issue with adding support to LG to enable NvFBC via the OBS blob, however you might have to obtain the blob from OBS manually, or install OBS along side LG.
Hmm, It looks like they’re using “texture streaming” instead of NvFBC. So while it is better performing, it’s not quite NvFBC, and it just makes calls to the driver.
It wouldn’t explain the NDA as we can do this easily without. We already use texture streaming in the client.
Yeah, something tells me there’s more if Influencers are getting the actual build that saves enough performance as stated in the CES presentation.
Hello, I am trying to make a VM to do some gaming on linux with looking glass, so in the guide it says to install ivshmem driver, so I download and install it. But when I run the “ivshmem-test.exe” it reponds with “Test: Find device…FAIL - Unable to enumerate the device, is it attached?”. What could be the problem?
Okay, first thing: we need your XML or qemu command line config. Can’t do anything without verifying you’ve actually installed the device.
Did you install the ivshmem driver manually via Device Manager?
Also the ivshmem-test application has not been updated/tested in a long time, it may not work correctly.
No, I right clicked the ivshmem.inf file and then install. I also tested through looking glass and it responded with same thing.
But in system devices in device manager there is no “PCI standard RAM Controller”, check the screenshot below. I have updated all nvidia drivers and reinstalled windows countless times.
As already asked, please share your domain XML/QEMU startup command.
