AMD RTG Used Looking Glass for a Split Screen GoldenEye Demo

AMD RTG put on a cool demo using Looking Glass to replicate split screen GoldenEye.

Reddit: https://www.reddit.com/r/goldeneyesource/comments/9lqmt3/local_4_player_splitscreen_on_4k_55_oled_using/

19 Likes

OK. that’s pretty cool.

awesome!
where was that photo made?

Not sure, reddit link there might shed some light on it

1 Like

Yeah, not a whole lot in the reddit thread, but this is definitely cool.

That’s so cool. How is this working? One Linux host with 4 windows VMs all running on one GPU and on TR?

Here is what I was told by the original author:

4 looking glass instances running on one Linux host, 4 windows VMs, each with their own Radeon Fury GPU (although I will use MxGPU SR-IOV in the future and share a Vega V340) Looking Glass performed the most reliably out of all the moving parts in my demo.

5 Likes

congratulations!
that’s really cool :sunglasses:

1 Like

That is actually really cool to see it being adopted (somewhat)

It’s a damn shame but understandable that they are locking the SR-IOV to the Radeon Pros. Well one more excuse to build crazy mixed multi GPU systems. :smiley:

1 Like

@alphazero1990

i don’t think radeon pros have it. Atleast i have heard so. Not that i have one, but i’d like one ; )

AMD is setting up MxGPU Cards as the dedicated SR-IOV enabled Solution, and the Instinct cards might have it too. Though i currently doubt that!

Interestingly, i assume that only really the vBios is differentiating between the cards. hence the hardware verifying it.

1 Like

Unfortunately, this wouldn’t work for CS:GO because of VAC, but it looks like this was on a local LAN server without VAC, which makes this possible.

What would be pretty cool is to see if you could run 4 instances of LAN Screencheat, and see if any players know which viewport is theirs… This game also supports Native Linux, so 4 Linux VMs would sweeten the deal.

Best part, the Looking Glass clients would randomize which shared memory file to which VM is read after each match, making it even more confusing.

Yes and no. If they were to give 1-2 instances for end user cards, that would be plenty without eating into their datacentre market share.

Scale just happened. I’d assume there.