OS X Let’s talk Metal 4 and M3

[Irishman]

Is it more likely to be coming during WWDC of 2023? Later? What kinds of new features are most likely to see? Less likely?

We know that Apple has caught the Real Time Ray Tracing bug, and if it weren’t for the tremendous power requirements to make it work on Apple silicon, we would have seen it already working to bring hardware-accelerated Ray Tracing to all Apple devices.

Do you think they’ve given up on it? Do you think they’re just pumping the brakes until they figure out how to get that 3nm process done and dusted?

Can there really be that much power draw savings that will allow hardware accelerated real time ray tracing without causing third degree burns or 10 second battery life?

 

[diamond.g]

I wonder if coming out with new versions of metal every year hurts the adoption rate of the API.

 

[galad]

It's not really a "new" version, it's just a bunch of new optional features. Sometimes some new additions, sometimes something that replace an existing API, but mostly new additions.

It's not like you have to trash your whole work and rewrite everything from scratch each year.

 

[Bodhitree]

Can there really be that much power draw savings that will allow hardware accelerated real time ray tracing without causing third degree burns or 10 second battery life?

They licensed the technology from Imagination, so perhaps they tried to use the reference implementation and found the power draw to be too much for iPhone. I do think we will see it in iPhone first, perhaps later this year with the A17, and then in the M4 on Macs. It depends on the time schedule, I think there is a slim chance it will make the M3, but the previous M-series chips have been a little behind the A-series in tech adoption.

 

[diamond.g]

They licensed the technology from Imagination, so perhaps they tried to use the reference implementation and found the power draw to be too much for iPhone. I do think we will see it in iPhone first, perhaps later this year with the A17, and then in the M4 on Macs. It depends on the time schedule, I think there is a slim chance it will make the M3, but the previous M-series chips have been a little behind the A-series in tech adoption.

seems like a waste putting it in the iphone before the mac.

 

[Spaceboi Scaphandre]

I wonder if coming out with new versions of metal every year hurts the adoption rate of the API.

What's hurting adoption rate of the API is the fact it's lacking compared to Vulkan and is exclusive to Apple devices. No game dev wants to learn a new API, especially one that is exclusive to one platform, a platform who's userbase is a farcry compared to Windows and who's platform holder actively hates game devs and makes their lives a living hell. Having to learn a new API costs time and money, and for Mac it's not worth it for the majority.

Divesting OpenGL and telling everyone to use Metal is why the API is not being used. Metal would probably be more used if they did a transition to it instead of just killing OpenGL overnight.

 

[camj]

I wonder if coming out with new versions of metal every year hurts the adoption rate of the API.

Personally, I don’t think this hurts it but rather helps it. The more functionality and love and attention the API gets, the better off we all are. Developers like having options. If they keep going at this rate, Metal is going to be on par with DirectX in 5 years.

Vulkan is a nightmare from what I understand and is much more verbose than comparable APIs. Anyway I just wanted to point out that releasing new major version updates is only going to help out and attract more attention. Devs will be grateful and more likely to use Metal 4+

 

[diamond.g]

Personally, I don’t think this hurts it but rather helps it. The more functionality and love and attention the API gets, the better off we all are. Developers like having options. If they keep going at this rate, Metal is going to be on par with DirectX in 5 years.

Vulkan is a nightmare from what I understand and is much more verbose than comparable APIs. Anyway I just wanted to point out that releasing new major version updates is only going to help out and attract more attention. Devs will be grateful and more likely to use Metal 4+

I'm pretty sure Metal is already on par with DX.

From an API feature usage standpoint we still haven't seen games use Sampler Feedback. Variable Rate Shading is quite rare, and we just got our first major game using Mesh Shaders (I actually think it is the second). All these things were introduced back in 2019.

 

[salamanderjuice]

I'm pretty sure Metal is already on par with DX.

From an API feature usage standpoint we still haven't seen games use Sampler Feedback. Variable Rate Shading is quite rare, and we just got our first major game using Mesh Shaders (I actually think it is the second). All these things were introduced back in 2019.

Does Metal have a DirectStorage equivalent?

 

[diamond.g]

Does Metal have a DirectStorage equivalent?

Yup. Called Fast Resource Loading. It was talked about in WWDC 2022.

 

[Spaceboi Scaphandre]

I wonder if coming out with new versions of metal every year hurts the adoption rate of the API.

No what hurts the adoption rate of the API is the fact it's the only option on Mac. Developers do not want to learn a new API usable on only one platform, a platform their games don't sell well on already, when Windows and Linux let them keep using the same APIs they've been using and have a much bigger audience.