Forum Discussion

🚨 This forum is archived and read-only. To submit a forum post, please visit our new Developer Forum. 🚨
TheAsimovInitia's avatar
TheAsimovInitia
Honored Guest
10 years ago

Better support for optimus (980M)

I'm not surprised that occulus doesn't think my 980M is good enough. I've had issues with both drivers and games. But once I have gotten them to work I get very good performance.

These power machines are very good gaming computers and developer platforms. I travel a lot so my Alienware 15 is perfect.

For example with Elite: Dangerous I had to rollback to a certain oculus runtime to make it detect it.

Is there any discussion in regards of supports for laptops now when you go CV1? I don't care if it doesn't match the official recommendation but will the runtime support it? As you seem to team up with Dell (Alienware) there might even come a laptop bundle?

6 Replies

  • "TheAsimovInitiative" wrote:
    I'm not surprised that occulus doesn't think my 980M is good enough. I've had issues with both drivers and games. But once I have gotten them to work I get very good performance.


    On the performance side, it makes sense for Oculus to set the expectations of developers and give them a target to optimize for. Given that 99% of laptops won't work because of technology limitations, they chose a target in the mid range of the current generation of consumer desktop GPUs. Your 980m, while powerful, is still only about 2/3rds as powerful as a desktop 970. So even if you could use it, it might not perform well for the launch games.

    However, the real issue is not, and has never been performance. Lots of good applications (even VR applications) can be written against lower power rendering hardware. The Gear VR proves that.

    The problem is support for direct mode. In the early days, the HMD showed up as a regular monitor on your desktop. This is a non-starter for a consumer product as it causes too much confusion for end users and lots of hoops you have to jump through. Additionally, Windows isn't very good at maintaining different refresh rates for different devices all connected to the same desktop. So the question isn't really "Why doesn't a 980M have enough performance?", but rather "Why doesn't the 980M support direct mode?"

    Virtually all gaming laptops that include a nVidia GPU also include an Intel GPU. The display panels are physically wired to the Intel GPU, and most applications run only on the Intel GPU. Applications that run on the nVidia GPU display content by copying data to the Intel GPU every frame. Additionally, the Intel GPU doesn't (currently) support the kinds of features that the SDK needs to drive the headset, like direct access to the display framebuffer.

    So, when Oculus says laptops are not supported, what they really mean is "dual GPU laptops with displays connected to the intel GPU are not supported". In fact you can get laptops with only nVidia displays, but they're not very popular because they tend to have terrible battery life. However, these laptops do in fact work with the Oculus runtime.
  • "jherico" wrote:
    "TheAsimovInitiative" wrote:
    I'm not surprised that occulus doesn't think my 980M is good enough. I've had issues with both drivers and games. But once I have gotten them to work I get very good performance.


    On the performance side, it makes sense for Oculus to set the expectations of developers and give them a target to optimize for. Given that 99% of laptops won't work because of technology limitations, they chose a target in the mid range of the current generation of consumer desktop GPUs. Your 980m, while powerful, is still only about 2/3rds as powerful as a desktop 970. So even if you could use it, it might not perform well for the launch games.

    However, the real issue is not, and has never been performance. Lots of good applications (even VR applications) can be written against lower power rendering hardware. The Gear VR proves that.

    The problem is support for direct mode. In the early days, the HMD showed up as a regular monitor on your desktop. This is a non-starter for a consumer product as it causes too much confusion for end users and lots of hoops you have to jump through. Additionally, Windows isn't very good at maintaining different refresh rates for different devices all connected to the same desktop. So the question isn't really "Why doesn't a 980M have enough performance?", but rather "Why doesn't the 980M support direct mode?"

    Virtually all gaming laptops that include a nVidia GPU also include an Intel GPU. The display panels are physically wired to the Intel GPU, and most applications run only on the Intel GPU. Applications that run on the nVidia GPU display content by copying data to the Intel GPU every frame. Additionally, the Intel GPU doesn't (currently) support the kinds of features that the SDK needs to drive the headset, like direct access to the display framebuffer.

    So, when Oculus says laptops are not supported, what they really mean is "dual GPU laptops with displays connected to the intel GPU are not supported". In fact you can get laptops with only nVidia displays, but they're not very popular because they tend to have terrible battery life. However, these laptops do in fact work with the Oculus runtime.


    Thanks for the answer it explains why it is not supported. Do you happen to know if the direct access to the framebuffer is something that in the future could be fixed with drivers or would it require a new generation of multi-gpu setups? I have a desktop with 980 in it but I wan't to explore fully mobile VR (backpack) and also work on it on my travels. I would very much prefer not having to have to buy a NVIDIA GPU-only laptop.

    I guess the graphics amplifier that allow a desktop GPU works the same way?
    http://www.dell.com/content/products/pr ... u=452-BBRG
  • Also, I saw now that it was posted in PC development and I guess General or Support is better if you want to move it.
  • Be careful of the new laptops that have full desktop 980 gpus too. I've seen one that claims to have optimus (the Sager NP9870-S / Clevo P870DM-G). Not sure if the store page is accurate on that one.

    (Clevo P770ZM is a 980m without optimus and runs oculus sdks fine. But it's probably too slow for CV1, sadly)
  • "TheAsimovInitiative" wrote:
    Do you happen to know if the direct access to the framebuffer is something that in the future could be fixed with drivers or would it require a new generation of multi-gpu setups?


    A driver can't change the fact that the actual hardware display port is physically connected to the Intel GPU, so I'm pretty sure that it can't be changed by a driver update. On the other hand, there's no reason a driver couldn't report that it supports direct framebuffer access, even though it doesn't, allowing the runtime to work. It would just introduce some latency while the frames are being copied over to the Intel GPU.


    "TheAsimovInitiative" wrote:
    I guess the graphics amplifier that allow a desktop GPU works the same way?
    http://www.dell.com/content/products/pr ... u=452-BBRG


    I don't think so. I haven't used one, but the impression I get is that you connect displays to the GPU in the amplifier. In essence the device is just an external PCI-X bus extension. It might work, although it might require that you disable the laptop nVidia GPU and the laptop primary screen as well. Hard to say without testing it out and I don't have the cash to try something like that.

    However it's worth noting that the X-51 Desktop machine on Alienware's site is one of the approved 'Oculus Ready' PCs and it supports the graphics amplifier, despite being a desktop machine. This is probably because it's a slim design computer and some video cards simply won't fit in the chassis.