|
Suspicious Dish posted:so here's the thing: you've been suckered by nvidia, and you don't know it. as much as i agree about everything else i can't agree with this. amd has been p. bad at drivers over the years. nvidia has always been more stable for me. on all platforms, too, this has held constant for me across many generations of macs dating back to when macs used powerpc cpus and amd's graphics division was still independent company ati i know gl/d3d drivers are shambolic horrors inside, and who knows maybe nvidia has been hurting all of us by forcing that model to stay alive longer than it should have, but if so the reason it worked for nv is that they were and are legit better at stabilizing those giant mountains of awful code
|
# ? Mar 11, 2015 05:47 |
|
|
# ? Mar 28, 2024 19:39 |
|
they're better since they're in bed with the gamedevs. each aaa game is mostly tested on nvidia, and each aaa game has a liason and a product team of 3-4 part-time engineers, working to make their driver not suck for your game. nvidia's driver works by rewriting aaa engines. and not only at the gl/d3d level. it has custom code for every single engine in existence, and for certain games, even patches higher-level parts of the engines. they basically rewrote skyrim's renderer and redid the scene graph code so it was more efficient. in the past, nvidia's driver patched flash to make calls to gettimeofday return a fake date under certain circumstances since it made the renderer run faster. when chrome changed how the flash sandbox worked, this broke all websites: https://code.google.com/p/chromium/issues/detail?id=16800 nvidia has no shame doing this, and they've built up this large codebase over the years. amd is not starting from the same space, and aren't in bed with aaa gamedevs, and can't afford to do the custom hacks that nvidia has. amd competes by trying to play some of the heuristics, but they're not playing all the same tricks so as a result, you get system instability, and nvidia plays it off as "amd is a bunch of idiots" and you eat it up. it is true: amd is not as good as nvidia at building a complex mountain of code that has no shame patching skyrim. all of this will be completely irrelevant when vulkan comes around. nvidia has tried *super* hard to keep the apis the way they are since they have such a significant competitive and marketing advantage built up over time. they squashed longs peak. they squashed opencl. i also want to point out that while i was at red hat, i had access to a super secret mailing list set up by nvidia to talk with aaa gamedevs, as a result of some of my linux contracting work for nvidia. very regularly, there would be an email from some big aaa company (@ea.com, @blizzard.com, @epicgames.com were all super common) complaining about some strange behavior with nvidia's driver, and the response would first be "that's needed for compatibility with ~*~ other application ~*~", then pushback along the lines of "please make a compliant driver", then a response along the lines of "a fix has been put in the next release", which patched the specific game to make things work. i could go on...
|
# ? Mar 11, 2015 07:08 |
|
Suspicious Dish posted:right, i think we're going to see vendors implement opengl on top of vulkan that aren't nvidia / amd. who pays for the middleware and why
|
# ? Mar 11, 2015 07:44 |
|
google will probably make something like ANGLE but for Vulkan for WebGL. there are going to be lots of pre-existing games using OpenGL that should be able to be ran. I'd imagine the community will do something like the plan I described. CAD / Adobe / other sort of workstation companies have a vested interest in keeping their OpenGL-based rendering stacks working, since they've long been an opposition to cleaning up OpenGL as well
|
# ? Mar 11, 2015 07:47 |
|
Suspicious Dish posted:amd competes by trying to play some of the heuristics, but they're not playing all the same tricks so as a result, you get system instability, and nvidia plays it off as "amd is a bunch of idiots" and you eat it up. it is true: amd is not as good as nvidia at building a complex mountain of code that has no shame patching skyrim. all of this will be completely irrelevant when vulkan comes around. sure but i have seen stability/bug diffs on mac and i doubt nvidia pulls any of this cray cray poo poo on the mac platform (*). please don't dump another wall of text on me just because i think it's legit that nvidia does 3d drivers better. also nvidia hardware superiority, i've seen plenty of amd visual artifacts where they were cutting corners for performance (e.g. across several generations of gpus they had a lossy z buffer compression scheme which would cause z fighting on some polygon seams, nvidia hardware would do the same scene clean). * - i would expect the ghost of steve jobs to rise from the grave and poo poo all over any vendor which tried to patch out gettimeofday() to eke out a few extra frames, because (a) stebe don't care 'bout bideo james even when he's showing them off in a spooky undead keynote and (b) apple institutionally loving hates outsiders patching their syscalls (it's scar tissue from the classic macos era). also thanks to (a) busting rear end on game performance doesn't actually help nvidia (or amd) win apple's business. hence the eternal mac owner disappointment when super throat slasher 666 goty edition actually got a mac port (holy poo poo) but it turns out to perform like rear end compared to windows dual booted on the same hardware
|
# ? Mar 11, 2015 09:23 |
|
Suspicious Dish posted:in the past, nvidia's driver patched flash to make calls to gettimeofday return a fake date under certain circumstances since it made the renderer run faster. when chrome changed how the flash sandbox worked, this broke all websites: https://code.google.com/p/chromium/issues/detail?id=16800 i ran strace on some hung phpunit code today and it was 99% gettimeofday for some reason. god i loving hate php.
|
# ? Mar 11, 2015 12:09 |
|
my linux experience has been too good lately i am thinking about installing fedora 22 alpha
|
# ? Mar 11, 2015 12:10 |
|
install gentoo then rant all day about how systemd is a military-industrial complex conspiracy
|
# ? Mar 11, 2015 15:03 |
Mr Dog posted:install gentoo then rant all day about how systemd is a military-industrial complex conspiracy I think you mean social justice warrior conspiracy
|
|
# ? Mar 11, 2015 15:20 |
|
ahmeni posted:who pays for the middleware and why developers pay for it, because targeting vulkan directly is probably going to suck and involve shitloads of boilerplate and hardware detail developers pay for or download all kinds of other middleware today:
adding a 3d graphics abstraction layer to the long list of things you buy/download before starting a game isn't gonna wreck the industry i expect it will make lots of things better rather than worse
|
# ? Mar 11, 2015 15:25 |
|
Notorious b.s.d. posted:adding a 3d graphics abstraction layer to the long list of things you buy/download before starting a game isn't gonna wreck the industry yup cheap turnkey engines have been the best thing to happen to gaming epic and unity are gonna work with vulkan or w/e and developers will continue to just use unreal and unity the same poo poo is happening on the most popular os for gaming, ios, and their metal graphics apis
|
# ? Mar 11, 2015 15:33 |
|
Cocoa Crispies posted:the same poo poo is happening on the most popular os for gaming, ios, and their metal graphics apis apple and powervr are both contributors to the vulkan working group, and vulkan has specific features to benefit powervr's unusual design i'd be shocked if ios didn't support vulkan early and loudly
|
# ? Mar 11, 2015 15:38 |
|
So how does Vulkan relate to OpenCL? Are the APIs going to be similar? OpenCL seems okay (although it's very cumbersome to use manually as you have to specify everything), although it is at times a bit silly how it bends over backwards to pretend you're not just programming a GPU, but rather some general "accelerator". Or are we supposed to use Vulkan for doing computations on GPUs now?
|
# ? Mar 11, 2015 16:20 |
|
yes. the marketing glossy says that integrating rendering and compute into one api was a major goal
|
# ? Mar 11, 2015 16:25 |
|
Also, major s at Nvidia's OpenGL implementation overriding dlsym/dlopen for the entire process.
|
# ? Mar 11, 2015 16:28 |
|
Suspicious Dish posted:nvidia has no shame doing this, and they've built up this large codebase over the years. amd is not starting from the same space, and aren't in bed with aaa gamedevs, and can't afford to do the custom hacks that nvidia has. in a bit of a twist, amd's opengl drivers actually do follow the spec more closely than nvidia. however, since most developers are running nvidia hardware and drivers they end up writing things that seem to work until they run it on amd and get a black screen or shader compiler errors. of course, since it works on their nvidia hardware without any warnings or errors they just assume that the amd drivers are broken when it's their code that's wrong. i hit this at some point when i had a program that worked on my nvidia machine but drew nothing on amd because i wasn't initializing some vertex attribute correctly but never knew there was a problem because nvidia's drivers were default initializing it and didn't throw a warning or error.
|
# ? Mar 11, 2015 16:55 |
|
Notorious b.s.d. posted:apple and powervr are both contributors to the vulkan working group, and vulkan has specific features to benefit powervr's unusual design Are you sure about that? Apple isn't even listed as a Khronos member. https://www.khronos.org/members/contributors
|
# ? Mar 11, 2015 17:05 |
|
Suspicious Dish posted:Are you sure about that? Apple isn't even listed as a Khronos member. an Apple guy led the WebGL stuff, so I think so e: yeah, since 2006: https://www.khronos.org/news/press/apple_dell_google_and_others_join_the_khronos_group
|
# ? Mar 11, 2015 17:07 |
|
this is the graphics thread now, right? probably How Video Game Breasts Are Made (And Why They Can Go Wrong)
|
# ? Mar 11, 2015 17:10 |
|
apple logo is right in the middle between epic and unity. edit: they are listed as a promoter member The_Franz fucked around with this message at 17:27 on Mar 11, 2015 |
# ? Mar 11, 2015 17:15 |
|
prefect posted:this is the graphics thread now, right? not appropriate for a Linux thread, Linux users have no experience with female breasts
|
# ? Mar 11, 2015 17:35 |
|
kde abandoned nepomuk, the big complicated rdf-based "semantic desktop" that was supposed to change human/computer interface forever kde was the only thing that ever adopted nepomuk in the first place so i believe this leaves the project with 0 users http://vhanda.in/blog/2015/03/the-semantic-desktop-is-dead/
|
# ? Mar 15, 2015 00:56 |
|
Notorious b.s.d. posted:kde abandoned nepomuk, the big complicated rdf-based "semantic desktop" that was supposed to change human/computer interface forever Oh thank god. One less thing to disable if I go with kde5.
|
# ? Mar 15, 2015 01:40 |
|
something that annoyed me of kde was how you could easily accidentally drag things off the kde menus and task bars/rearrange without actually wanting to, and to lock them i had to manually lock all of them individually lol.
|
# ? Mar 15, 2015 10:06 |
|
Celexi posted:something that annoyed me of kde was how you could easily accidentally drag things off the kde menus and task bars/rearrange without actually wanting to, and to lock them i had to manually lock all of them individually lol. I'm glad I'm not the only 1 in yospos bad at the computer.
|
# ? Mar 15, 2015 11:51 |
|
my first task at my linix job is dealing with some realtek driver oh god
|
# ? Mar 16, 2015 17:53 |
|
the drivers for a wifi card i had used a deprecated kernel macro so i had to edit the source and recompile every time i wanted to install it welp that's my realtek story
|
# ? Mar 16, 2015 17:55 |
|
if you can't get your wireless driver accepted by upstream, that's a really, really bad sign
|
# ? Mar 16, 2015 17:57 |
|
we have a no realtek policy at work intel > broadcom > marvell > nvidia > tin cans and string > realtek
|
# ? Mar 16, 2015 18:06 |
|
carry on then posted:the drivers for a wifi card i had used a deprecated kernel macro so i had to edit the source and recompile every time i wanted to install it lol
|
# ? Mar 16, 2015 18:09 |
|
but really lol @ wifi on linux because it implies you're running it on something other than a server
|
# ? Mar 16, 2015 18:10 |
|
intel wifi is the only good wifi
|
# ? Mar 16, 2015 18:49 |
|
sometimes i gotta turn wifi off and then on again ... on my uBUNTU lappy
|
# ? Mar 16, 2015 18:52 |
|
Notorious b.s.d. posted:intel wifi is the only good wifi i wish apple would realise this
|
# ? Mar 16, 2015 19:22 |
|
Broadcom hates freedom
|
# ? Mar 16, 2015 19:24 |
|
broadcom is bad even under osx
|
# ? Mar 16, 2015 19:25 |
|
Broadcom is for broads
|
# ? Mar 16, 2015 20:18 |
|
Don't you guys have your servers on wireless? Really simplifies cable management tbqh.
|
# ? Mar 16, 2015 20:21 |
|
looks like kdbus will finally be merged into kernel 4.1 so applications that need fast ipc won't have to roll their own anymore.
|
# ? Mar 16, 2015 21:04 |
|
|
# ? Mar 28, 2024 19:39 |
|
oh good i've been watching the latest review thread on lkml and it's been a week, still basically no review comments though.
|
# ? Mar 16, 2015 21:15 |