Slashdot Mirror

On what type of hardware are you trying to install Linux? It is true that Linux does not support all hardware, but it supports more then Windows out of the box (without driver CD's). Second video card. Sound like Nvidia? => http://bumblebee-project.org/ Blame Nvidia for this that it took them ages to add support for it in their drivers! I personally never use hibernate (takes longer then cold boot), but standby works without any problems on my pc's and laptop. Still it is dependent on the hardware. If you try to install Linux on a Windows-machine most of the time it will work, but not always.

I can't get it working with the 3.8 kernel in the new ubuntu beta... wonder if this will make that project unnecessary..?

Bumblebee. I use it, and it works.

Incorrect; Bumblebee allows you to boot into Intel and run Optimus stuff on the Nvidia with a simple command of "/usr/bin/optirun (application)". I use it all the time and it works great on a couple of different Optimus-equipped laptops I've tested it on. On my own Alienware M11xR2 I can get 4 and a half hours out the battery if I don't run 3D stuff... about 2 and a half when I do.

I have an nvidia optimus/k1000m on a shiny new thinkpad w530. The kernel-included nouveau and intel drivers work fine. Switching between these without restarting is even theoretically possible with vga_switcheroo (though it apparently doesn't entirely work on the w530 specifically). Using the "optimus" bit is also perfectly possible with bumblebee.

However, even with bumblebee, the drivers are included with the kernel, allowing you to fully use KMS, bootup logos, etc. For full 3D, you can even still rely on the builtin intel drivers and use the proprietary nvidia drivers with bumblebee (or not, if your system allows you to switch fully to discrete mode).

*NVIDIA makes good binary drivers, but those have problems when a new kernel version comes out with changed interfaces: Only NVIDIA can adapt them, and until they get around to it, NVIDIA may not work with the latest kernel version.

In addition to that, Nvidia tends to do things in their own way in order to share code between the various OSes.
And some feature can't be implemented in the Windows-y way on Linux.
On the other hand, the linux kernel provides plenty of software facillities, thus increasing code share between drivers. Nvidia tends to ignore these. (For example, Linux has standard interface for handling multiple monitors, changing the resolution, and so on: XRandr. Until now (version 302) all that the nvidia drivers could do was only their own TwinView, which is proprietary, has nothing to do with XRandr, doesn't play nicely with other Linux software. But is the same technology as what they do on Windows - Meanwhile Xrandr has been available on ATI for ages, both on the proprietary and opensource drivers).

The Optimus situation boils down to that.
- A new generation of optimus laptop has an embed Intel gpu which is always on, and which is permanently connected to the display. It has also a discrete Nvidia GPU. Unlike older dual GPU laptops, it lacks any electronics to switch the output from one GPU to the other. The Nvidia GPU will never see a monitor connected to it, it stays always headless. Instead the Nvidia GPU has to render its 3D ouput on the frame buffer of the Intel GPU , which will display it.
- The Linux operating systems has several software facilities which could make such an "shared framebuffer" architecture workable (both as in kernel code to route ouput, and a modular Gallium3D driver stack) - in fact, current experimental implementation of such software works to do crazy stuff. Like hot plugging an USB attached display adapter and have the computer's high performance GPU off loads the 3D rendering and then stream the graphics to the external 2D LCD.
- But the Nvidia people don't use these because they differ massively with how they deal with these things in their common OS-shared code. End result. no Optimus for you, which can in some circumstance even mean no graphic output for you at all.

Nvidia's answer? Go see Bumblebee.
(Bumblebee is a hack by a small 4-man team [kudos to them!]. Basically it will start on-demand a separate X server on the Nvidia hardware and use VirtualGL to copy back the visuals on the main X Server running on the intel. Users can select which software to run on the real Intel X server, and which software to remote to the Nvidia X server. It can solve the optimus problem in a few situations. But that is still a hack. And that's because Nvidia has a closed driver that can't be fixed, so volunteer have to make such hacks aroudn the unfixable problems)

And as far as I've heard, the Tegra situation isn't any better. Nvidia keeps doing things their way.

*Proprietary drivers - 1) monitor upcoming kernel builds and proactively update drivers before the next kernel release or 2) have a dedicated nVidia contact to work on updating drivers ASAP when notified that an upcoming kernel build breaks them

Well that would be the strict minimum. It would be even better if Nvidia could at least try to play nicely along with what's done by everybody else.

We have a real clash of mentalities.

On the linux side, we have developpers which like consolidation, code sharing and so one. If some function has to be performed by more than 2 modules, the developers dream is to have that function move out into a common module with a nice API to export such functionnality to anyone who can need it. (for exemple, in storage, the long term is to consolidate all the various data redundacy functions (MD linux software raid, vendors sfake raid, LVM stripping, DRBD network redundancy) into the same facility - currently LVM and Fakeraid alrea

Isn't that problem solved by the bumblebee project?