It is not Nvidia's job to prop up the desktop linux ecosystem. It's wise not to invest time in that ecosystem, as far as I can tell. It's got an ongoing history of disappointing its customers and vendors over and over. And we get good enough computational support ever business or individual out there, Nvidia GPUs are still preferred for clustering.
> I will avoid Nvida like the plague till the day I die.
The way you talk makes it sound like it's a religious or political debate. You're mad that they don't support your demagogue or your shared values. Isn't Hacker News a place where we're supposed to keep politics at bay and focus on the technology?
NVIDIA released Linux drivers hand in hand with their Windows based ones, you get at least one update every month.
So while it's true that you can't have a completely open source driver for NVIDIA you do have excellent Linux drivers.
On the compute side NVIDIA's support for Linux is also unmatched, in fact for many things CUDA actually works better on Linux than on Windows.
There is still problem with switchable GPUs: namely those, where one GPU is using Mesa and the other is not (i.e. Nvidia). Those computers didn't disappear for the face of the Earth, people are still using it and expect it to work, even if there are never laptops available on the market. It is being worked on (https://blogs.gnome.org/uraeus/2016/11/01/discrete-graphics-...).
Intel doesn't supports GPU switching on its MESA drivers it requires the proprietary one.
Get the prop nvidia and intel drivers and get nvidia prime and you are golden.
Same issue with AMD GPUs with intel CPUs especially the PowerXpress ones.
They do have a binary firmware, for Broxton and newer, however that's a thing that runs on the GPU. It does not have anything with how Mesa or other OpenGL stack shares buffers and video output devices.
Do you have a source for this? In any case, just because some hardware has been recently phased out doesn't mean the manufacturer shouldn't be expected to support it. Not everyone has enough money to buy a new computer every year; I'm a student with an Optimus laptop and I know I won't be able to afford a replacement in the next year or two.
If NVIDIA can't be asked to fully support their hardware on Linux, the least they can do is release some documentation so the Nouveau devs don't need to grope in the dark.
I don't disagree in principle, but they've killed it, releasing information to make it work might be a problem for them, and since this thing is no longer supported adding compatibility for older devices isn't a high priority for them.
It worked poorly on many windows machine, it's effectively still doesn't work on Windows 10 [0] and it's been effectively killed since 2014.
[0]https://forums.geforce.com/default/topic/860554/geforce-mobi...
It's also worth noting that NVIDIA couldn't get the darn thing to work on Linux, it's also worth noting that AMD also had/has similar issues with PowerXpress on non-Iris based Intel CPUs, it does somewhat work well on AMD's own CPU's with integrated graphics and it's more modern "APUs".
>Do you have a source for this?
Source for what? under Iris Intel the CPU handles the switching between GPUs not that it works that well, but it doesn't work well on most OSs other than windows mostly due to how the display driver infrastructure is setup and how the composition manager works (even on OSX you still likely to want to use gpu-switch than rely on the OS).
If you want to blame some one blame the OS tbh, if OSX and Linux had proper user mode display drivers like WDDM, had modern display drivers like WDDM and actually had a composition manager capable of using multiple GPUs then there might be better solutions for this.
But overall it works, it does require you to use the proprietary drivers from both NVIDIA and Intel but it works, at least on Ubuntu I never had issues with following a similar guide to this http://ubuntuhandbook.org/index.php/2016/04/switch-intel-nvi... (reading the comments on that guide it seems that it missing an important part which is apt-get install nvidia-prime) So this just works...
To switch between the 2 on demand you can simply use these commands:
-prime-select intel
-prime-select nvidia
On other distros (or if you are not using nvidia-prime) you might need to use /proc/acpi/bbswitch And modprobe/rmmod to do the switching, there are bash scripts online ready to use which are available as long as you have the proper drivers installed.
P.S. Sorta OT but relevant: http://i.imgur.com/Ql1dsZC.jpg
Also it looks like PRIME supports synchronization now if you have a distro with DRM support (no not the bad kind ;) https://devtalk.nvidia.com/default/topic/957814/prime-and-pr... so if you are using their currently suggested setup you don't need to switch manually anymore.
This is much more expensive than hardware implementation of Nvidia Optimus.
It's more expensive on Linux because the DRM on Linux is broken, there is no longer "Optimus" since NVIDIA no longer uses a dedicated chip, you share a frame buffer with the IGP.
PRIME also supports the older Optimus chip laptops, but again this doesn't work that well, but then again these don't work well with Windows 10 either due to the changes to WDDM.
But that won't stop me from trying to fairly apply the policy to the free software people who are utterly convinced that serving their agenda is a net good rather than a lifestyle choice.
Shouldn't be a problem anymore under Linux as most distros today install Nouveau drivers by default. https://nouveau.freedesktop.org/wiki/
CUDA only exists because Nvidia is attempting to pretend OpenCL, Vulkan, and DX12 don't exist [1]. These require hardware scheduling on the GPU to switch shaders. Rather then dedicated X amount of chip hardware to Y shader for Z ms.
It should be noted for GPGPU compute Nvidia is not the correct choice. AMD RX 480 has 5.8TFLOPS @$200 ($37/TFLOP) vs Nvidia GTX1080 8.9TFLOPS @$600 ($67/TFLOP). In reality you should be doing GPU programming in OpenCL so you are GPU agnostic. You can switch vendors or platforms seamlessly (in most cases if you avoid proprietary extensions) even target AMD64, ARM, and POWER8/9 hardware.
That being said I own a boat load of Nvidia stock because their marketing is excellent. Really marketing is all 80% of people pay attention too. CUDA has some great marketing around it. In reality CUDA is slower then OpenCL (on Nvidia's platforms even) and no easier to work in.
Let's hope it gets picked up by machine learning frameworks etc., because this market badly needs the competition, as your comparison of per-dollar raw performance numbers shows.
But I'm curious in how the FLOPS on these cards were measured. For example one concern I have is that presumably these two cards have slightly different levels of parallelism. So it may be more or less difficult to extract the full performance from a particular card due to parallelism overhead. Then there's driver overhead, ease of programming, etc.
F * (1/Hz) * 2 = FLOPS
Where Hz is the well.. the clock rate, inverse to get cycles per second. This is wrong because stalls happen, memory transfers, cache misses etc. It is also wrong because the clock rate is throttled and you are not always at Maximum boost clock.
Then multiply by 2 for FMA (fused multiply add). This is wrong because well not every operation is a one cycle FMA. Division can be many (>100). Also scalar pipelines don't have FMA.
Ultimately all vendors use the same crappy calculation so we are comparing apples to apples. Just rotten apples to rotten apples. It gives you a good ideal circumstance you can optimize towards but never actually attain.
Sample applies to integer math, long double math and so on.
AMD RX 480 has 5.8TFLOPS @$200 ($37/TFLOP) @231Watt Peak (2.5GFLOPS/Watt)
Nvidia GTX1080 8.9TFLOPS @$600 ($67/TFLOP) @318Watt Peak (2.8GFLOPS/Watt)
See Furmark benchmark for wattage values [1]
Typical KW-H in the US $0.12KW-H [2]. So the delta-cost of GTX1080 vs RX480 will be mitigated by GFLOPS/Watt efficiency savings in 4 years, 4 months. Which on a typical 2, 3, or even 4 year hardware replacement cycle the extra cost will NEVER be re-couped.
[1] http://www.anandtech.com/show/10325/the-nvidia-geforce-gtx-1... for load wattage numbers
[2] http://www.npr.org/sections/money/2011/10/27/141766341/the-p...
Meanwhile, AMD are active contributors to Mesa and DRM, which benefits everyone.