This just came in…
“Vulkan Videos From This Week’s GDC 2016 Conference”
http://www.phoronix.com/scan.php?page=news_item&px=Vulkan-API-GDC-2016-Videos
Have not watched them, figured I’d link it as it is probably relevant to the conversation.
This just came in…
“Vulkan Videos From This Week’s GDC 2016 Conference”
http://www.phoronix.com/scan.php?page=news_item&px=Vulkan-API-GDC-2016-Videos
Have not watched them, figured I’d link it as it is probably relevant to the conversation.
Well, for what it’s worth, I’m for DX12.
Here’s hoping, but I have a hunch that your resources will grow to allow you guys to do the stuff that you feel are also important. 
I wish I could be as optimistic (sure, Sony and Microsoft I’ll buy), but I’m still not convinced that Nintendo has realized that they’re the one that needs to be courting third parties yet. The NX rumours aren’t making me feel any better about my inevitable purchase (because, dammit, I’m an addict and I need my fix).
Specifically, the rumors that even long-standing partner studios haven’t heard anything solid about the NX yet.
But at least it sounds like they’re looking into standards for once.
Why else would they join the committee? I know this is Nintendo we’re talking about, but at this stage, I’d be shocked if Vulkan was not the primary graphics API on NX.
Nintendo has a direct line to your wallet too? #Smash4Life
Asking why Nintendo does things is the first step toward madness. I mean, why did they spend all of that money and time on the game pad only to support it as poorly as 3rd parties? You’d think at least they would have had some game ideas for it!
I’m trying so very, very, very hard not to drop $80 on Twilight Princess. Again. I mean, it’s sitting RIGHT THERE on my shelf already, but… But…
Too many good points in this thread and too little time to address all that would deserve to be adressed, so instead i am just going to drop my opinion on the point i care the most about:
I do absolutely NOT see backwards compatibility as something unimportant, or as something that should be given up lightly.
Agreed.
Also, the promise of having the game work under Win7 is not something to be broken lightly either.
So far, the best option seems to stick with D3D11, in my opinion. Unless there’s a major shift from the 50%+ users from Win7 to Win10, or a performance improvement in Vulkan.
Not being a tech geek, I was wondering if there will be a lot more work switching over once the game has neared completion as opposed to now?
Yes there will be more work however the tradeoff is that we’d be more likely to be able to afford that work.
I personally wait a few months, if Nintendo, Valve and Google are serious about Vulkan and about supporting it then they will surely throw resources and personnel at fixing the most obvious problems with Vulkan, if the solution don’t look like appearing after a few months, choose DX12 and leave support for Vulkan for the future when it big commercial supporters pull their fingers out and start sorting out issues with Vulkan.
Apparently Vulkan was going to be use for Marshmallow but it wasn’t ready in time. I be pretty surprise if Android N doesn’t come with Vulkan.
Nintendo, haven’t got a clue, their strategy makes no sense to me at the best of times, sometimes I think it a company that just give up competing with everyone else in the industry and just does it own thing. Which is perfectly fine but not exactly profitable strategy.
Valve their strategy make perfect sense to me, it just a crap strategy that isn’t going to work but it does demonstrate the company is run and control by geeks who thinks the general public is exactly like them.
My wife drives by Nintendo everyday. I’ll get her to stop in and ask WTF are they doing - but in a very polite, passive sort of way.
It is not a very good idea to base your whole judgement on only one game that is not even optimized.
Let’s see the results on the Source 2 Engine: https://youtu.be/xXyZ4YaktyU?t=1h33m25s
Vulkan is absolutely crushing DX9/OpenGL/DX11. Valve is not even looking to port their engine to DX12 because Vulkan is all they need.
Just wait until Unreal Engine 4 supports Vulkan (a matter of weeks) and the Ashes of Singularity port is available. DirectX 12 is not relevant anymore.
The benchmark you’re quoting is rendered at 640x480 which is absolutely not a coincidence and completely undermines your entire post. Even if Khronos wasn’t so good at defeating itself D3D12 would still be relevant because it’s supported (or will be) on the Xbox One.
It might not show off the shader performance issues you are mentioning. But it shows that the API is very efficient. Once this bug is squashed there is no reason to not consider Vulkan. In my opinion Windows 7/8/8.1, Linux, Android and possibly iOS/OS X are much more important than the Xbox One. I am at GDC right now and I can say Vulkan is rperceived very positively, it’s nothing like OpenGL.
I encourage you to watch the two talks from Valve Software and Oxide developers, it is really interesting.
There isn’t any issue with the API, at no point have I mentioned there being any problems with the API. The issue is with its shader compiler ecosystem.
The problem is that your post is implying that it is a major issue with the Vulkan API when it is an early bug that will be fixed quick. By the time your game is published all these issues will be gone.
It’s a major problem with “Vulkan” and I don’t see how you could possibly say that it will be fixed by the time we publish our game as there has been absolutely nothing publicly stated so far, at least nothing that I am aware of, that mentions a timeline for resolving this problem. I raised the issue with the LunarG guys here https://github.com/LunarG/LunarGLASS/issues/5 and as you can see they didn’t have much to say about fixing it. I recommend you re-read what I wrote and the nature of the argument I was making if we’re to continue debating this.
I saw those Talos benchmarks and I was surprised.
I originally heard of it and how it’d be based on Mantle. Like, literally AMD just gave them the code to run with as a basis for the Kronos Group’s next gen codebase.
Well, Mantle had huge performance improvements over DirectX11 on AMD hardware.
So, what’s going on here?
I also saw that beautiful UE4 demo running on the new Samsung Galaxy at 1080p. If it can run that on a phone well, why would it be slow on a PC?
Perhaps it’s Talos itself and not Vulkan?
In the end though, I would think it’d be best to support DX12 and Vulkan both. That seems to be what most games are doing. But if you had to chose one, well it’d seem like Vulkan would be better than DX11 moving forward. Vulkan is likely to be updated and supported more, while DX11 is being ditched for DX12 that’s limited to Windows 10.
As someone else pointed out, Valve did a conference sort of thing, updating from DX11 to Vulkan in Dota2. They said the performance differences were extremely minor, except when there are lots of draw calls. That’s counter to the great decrease in Talos Principle.
See: https://www.youtube.com/watch?v=xXyZ4YaktyU&t=1h27m25s
Dota 2 Vulkan Performance Results:
- Reduced latency
- Improved framerate in expensive scenes"
I guess you already said it’s only because the resolution is low, but eh. Aren’t there enough draw calls, like all those asteroids in ring systems, that’d benefit heavily?
I guess Valve isn’t above lying to push Vulkan, and it being BS how they’re saying the resolution is low because they’re seeking to measure CPU performance and latency, but it’s getting really wide adoption.
Unity, UE4, Source2(obviously), in addition to all sorts of platforms, virtually everything seems to go toward supporting Vulkan.
I’d imagine that Talos Principle is just an outlier. I’d imagine it has a lot to do with it being a slow, mostly static FPS.
Anyway, like Flavien said, the min reqs were stated to be Win7. So if you have to choose one, Vulkan seems to be the obvious choice.
Worst case scenario, if the performance isn’t great at high resolution, it just isn’t great at high resolution. It still works and is still a good game. No one is demanding the best optimization that could be possible as long as there aren’t a bunch of hiccups that plague some games.
To be honest, whether it is DX12 or Vulcan, I’m good with either. But I am wondering in terms of ergonomics, will the development come to a standstill in other departments while DX12 or Vulcan is implemented? I ask this because I distinctly remember Star Citizen developers having to wait until certain departments were finished before continuing; this really created a bottleneck and slowed down production. If indeed this is the case then adapting DX12 or Vulcan could be a very time consuming process once the game is near completion so I keep wondering if sooner is better in the long run. But I know it will work out either way in the end - now or later. Just curious I guess.
@innociv as I was explaining to @libcg the problem isn’t Vulkan the API, which is what the Valve DOTA presentation was benchmarking. The problem is in Vulkan’s shader compiler ecosystem. The performance differential you see in the Talos benchmark is due to unoptimized shaders because Vulkan only has a reference compiler. This has nothing to do with the API and everything to do with the shader compiler that was used.