Hi Rune,
Red Matter 2 dev here. For the record, I’m fairly certain you tested RM2 back in the day when it still hadn’t received the latest graphical update with the most recent bells and whistles. That probably account for the difference.
Thanks, and of course you could be right - but usually the 5090 is not 2.5 times (=150%) faster than the RTX 4090, and that difference seems very odd to me.
I was more wondering if the RTX 4090 could hit some ram limit with Quest 3 res 400% - where the 5090 could benefit from the 32GB. Running low on ram could explain the sudden massive slowdown of the RTX 4090. For the RM2 testing, the following config was used:
GPUS: RTX 5090/4090
CPU: 7950X3D
RAM: 64GB 6000Mhz/CL30 DDR5 RAM
Quest 3 with Virtual Desktop, Godlike: 6144x6432(400%) Per Eye
That res translates into 79 million pixels per frame combining both eyes - so quite an insane res, more than 2 x 8k res. Also the Quest 3 was using 120 Hz.
Also results from Metro could indicate that the 4090 hits a wall even with 24GB vram:
Here the 5090 is more than 3 times faster than the 4090 - extremely unusual results. But a lot of other variable could interfere. I'd love to see same test made with Index and other hmds not using streaming or VD. And I'd love to see other Quest 3 users verify these results made by one YouTuber, who does not seem to have much experience testing hardware.
Of course if the 32GB should make the above possible that would be a valid reason to get the 5090 - for those wanting to use such extreme resolutions, and that might be a very small minority. 2c.
(Btw, if some thinks I'm joking about any game needing 32GB of vram, see how easily Alyx used nearly 20GB vram here with a Reverb G2 and RTX 3090 with 24GB vram as res increased - jump to 9:36 to see the 18.5GB vram use:
Thus it's very likely that RTX 4090 may have hit a vram wall when extreme Quest 3 resolutions were used. )
