Would these VR headsets have worked well with Crossfire/SLI?
Would these VR headsets have worked well with Crossfire/SLI?
A potato card isn't suitable for them, but a solid GPU is better. The power required, the cabling, the heat—two cards would be unnecessary. One good card is more practical. Given the low resolution and refresh rate compared to a monitor, running two cards for VR doesn't seem worthwhile.
They are just most users limited to Meta/Oculus with a maximum of 90hz, which I find quite low. It wasn't just the frame rate that bothered me but also the resolution and racing experience made me feel unwell. Not dizzy, just like staring at an arcade machine. I didn't enjoy it and am back to using a 65"120hz OLED, which is a huge improvement.
There are, of course, I use this term carefully—'higher-end' headsets such as Vives and Primarks—but they come with inconsistent quality control and are four to ten times pricier. If you can afford them and are open to risking it, you're investing in a top-tier GPU.
Quest 2 and 3 both work at 120Hz. On the subject, the common VRAM used in SLI configurations isn't the best choice for VR systems.
Ah the quest 2 was 90 but received an update that allowed 120. Quest 3 always had it. I was corrected.
On-Topif: I concur. Perhaps I'm in the wrong stadium, but your words have me reconsidering. Latency and frame pacing matter. If two cards drifted slightly out of sync, it would feel like experiencing insanity—where one eye sees something different from the other.
Even in a shared setup, a few frames miscalculated by one GPU would make you go wide-eyed.
I was never familiar with VR SLI, but there was a fascinating remark in that article:
'With this capability, if an engine already supports sequential stereo rendering, it’s quite straightforward to activate dual-GPU support. Just add a few lines of code to configure the mask for the first GPU before rendering the left eye, and then set it for the second GPU before rendering the right eye. For scenarios involving shadow maps or GPU physics simulations where data is shared between both GPUs, you can configure the mask to encompass both, allowing draw calls to be broadcasted. It’s remarkably simple and easy to incorporate into an engine.'
Additionally, this applies to any number of GPUs on your system, not limited to just two. You can also use affinity masking to precisely manage how tasks are distributed across multiple GPUs, such as 4 or 8.