Phoronix just posted Xplane benchmarks for the upcoming release including OpenGL and Vulkan.
Basically Nvidia leads the pack as long as you are at low resolutions it can run at rediculous FPS but with AMD only about 20FPS behind… but once you get to 4k and high settings or multiple 1080p screeens or an ultrawide monitor… AMD pulls back toward the front especially when you consider value, also it seems the simulator has some sort of culling or ROP bottleneck as Navi 5600-5700xt all perform the same and 5500xt with 8 fewer ROPs than the 40 on the other cards predictably runs 25% slower. So… it seems they are leaving alot of performance on the table somewhere in the engine. It is likely if that bottleneck were removed AMD and Nvidia cards would probably all run faster.
It’s very telling that a $250 AMD card at 4k very high is only 8 FPS behind Nvidia’s $2500 card … the engine isn’t just putting the GPUs to work very well.
All this arguing over nVidia/AMD and for what? Has ANYTHING been advanced for Haiku? Hmm… NOPE! Constructive debate is one thing, but I see no value in any of this. How about… oooh… we try to further optimize the drivers we already HAVE?!? Work with what we HAVE, not argue over what we DON’T have and may never be able to get. Kids in a sandbox, throwing sand at each other… seriously!
Re:nVidia drivers
When running drivers that are open source, nVidia’s performance falls far behind the competition. This would suggest that most of nVidia’s performance comes from driver optimizations instead of actually having higher performance cards.
Re:3rd party OS
On minimally POSIX or non-POSIX operating systems you can’t use Linux drivers at all because of inefficiencies in X11. Even on Macs nVidia support is likely to be left behind because of drivers not being open to Apple’s engineers. It’s becoming time for nVidia to pay the piper for its closed nature because other companies can share tech to team up against nVidia by using open-source drivers.
From my background as an Amiga and AmigaOne backer, RadeonHD cards run accelerated software with shader support on AmigaOS 4.1 but nVidia wouldn’t issue register maps to the guy that wrote the current drivers. Therefore only older nVidia cards work on AROS because it is open source drivers only and they were ported across from Linux. The softpipe pipeline used by AROS limits the Gallium3d support such that newer drivers that require LLVMpipe aren’t even an option.
Beggars can’t be choosers. Stick to what you’ve got sources for. And by the way, AmigaOS is proprietary and still couldn’t get nVidia support. An open-source OS like Haiku doesn’t stand a chance of even being a blip on nVidia’s scope.
I agree, otherwise I wouldn’t be here. Getting nVidia to see the potential will be trickier than any of us imagine but porting an open-source driver set may lead in that direction.
NVIDIA provides full Vulkan 1.2 support and functionality on NVIDIA GeForce and Quadro graphics card with one of the following Turing, Volta, Pascal, Maxwell (first and second generation) and Kepler based GPUs.
That would be a partially incorrect assumption… If Nvidia’s cards had the same open source support as AMD, they would be faster most likely as they are graphics focused much like AMD’s new RDNA architecture, older AMD cards were more general since AMD did not have the money to invest it separate architectures for compute and for graphics. Both Nvidia and AMD do game specific optimizations, Nvidia just does more as they have a larger budget. (all the more reason to supprt AMD so they have the budget to implement things like this in the open drivers).
The main reason most Nvidia cards at least new ones are extremely slow with the open source drivers on Linux is that the drivers are reverse engineered with virtually no help from Nvidia and reclocking the GPU core and VRAM doesn’t work for years until someone figures it out… which means your GPU ends up running at base clocks all the time so no supprise that a 100Mhz 2080ti is pretty slow.
Also note that modern cards are not fast at some older rendering methods that were faster on older cards, which is why XPlane is slow on newer AMD and Nvidia cards. It is also why they are working on reengineering thier render backend to support vulkan… it is not there yet but does show some promise.
Yes. You helped in some key things yourself. Good constructive arguments are healthy. Even fact checking is needed since documents and certain information sources are either biased and/or outdated.
