Announcement

Collapse
No announcement yet.

Apple M1 Ultra With 20 CPU Cores, 64 Core GPU, 32 Core Neural Engine, Up To 128GB Memory

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Originally posted by drakonas777 View Post
    I was hoping AMD would launch 5-9W APU, but it did not,
    I'm pretty sure most of the 15W processors can be restricted to <= 9 W, if an OEM should want to. I think you don't need a specialty SKU, just for that. It's more for the ~5W segment where you're essentially talking about a tablet-spec processor.

    Originally posted by drakonas777 View Post
    or I'm getting a Macbook,
    Last time I priced out a Macbook was in 2005, but they seemed to cost about 2x what a PC with equivalent specs was going for. Is that still true?

    Comment


    • Originally posted by Developer12 View Post
      Apple made this linux support _possible._ They had to purposely engineer the boot mechanisms unique to the M1 to enable all of this. That takes real developer effort.
      Enjoy it while it lasts. You know, they could change their minds at any time and slam the door on Other OS for all new units shipping. Or, maybe they'll rethink the matter when the platform is updated for the M2 generation. Since we simply don't know if/when they could change their minds, I wouldn't be spending my time on reverse engineering support for a platform that could render all of my efforts worthless, whenever they feel like it.

      Originally posted by Developer12 View Post
      Their engineers have also left more than a few backdoor gifts to help along the process, modifying things here and there.
      That's great, but still no guarantee. If management cracks down, then those gifts can go away in a hurry.

      Comment


      • Originally posted by mdedetrich View Post
        now there is no reason for MacOS to not implement a Vulkan driven. Heck it would greatly introduce adoption onto their platform
        I believe their official line is that people should use MoltenVK. So, that probably slams the door on getting native Vulkan support. As long as they maintain their technical lead and their phones remain so popular, they don't have much incentive to support "industry standard APIs".

        The other point people make about iPhones is that their userbase is more likely to pay for apps and will pay more money for them. So, it's a much more lucrative market for app developers. Given that, people will do what it takes to port to Apple devices, in which case the main impact of Apple using industry-standard APIs is making it easy for app developers to support other platforms, as well.

        See, when you have a walled garden, it's not only your users you want to keep locked inside -- it's also the app developers!

        Comment


        • Originally posted by Ladis View Post
          1. The industry standard is OpenGL, not Vulkan. At least we agree that there's no point for Apple to implement Vulkan (if ever was).
          OpenGL is effectively dead. It's not getting any new features. Certainly no ray tracing, though I'm not 100% sure about VRS or mesh shaders.

          Moreover, Apple dropped it from iOS and at least deprecated it from MacOS. They had it, and then they dropped it. Nice, eh?

          Want to play with Apple? Just take care not to get smacked by their big dick they like so much to swing around.

          Comment


          • Originally posted by coder View Post
            I'm pretty sure most of the 15W processors can be restricted to <= 9 W, if an OEM should want to. I think you don't need a specialty SKU, just for that. It's more for the ~5W segment where you're essentially talking about a tablet-spec processor.
            Yeah, but the problem is OEMs do not want that, they want good performance during reviews. Furthermore most of 15W SKUs are actually boost'ing beyond that, at least temporarily, but I believe AL ones can sustain higher power limits as long as cooling is sufficient. I do not want this. I want low wattage and big battery + great build quality. It's rare combination now. I mean, OEMs can do whatever they want for out of the box, but at least they could leave and option in UEFI settings to change power settings for advanced users. AFAIK that is very rare case also. I would gladly configure what I want myself.

            Originally posted by coder View Post
            Last time I priced out a Macbook was in 2005, but they seemed to cost about 2x what a PC with equivalent specs was going for. Is that still true?
            Unfortunately this is almost opposite now, at least in EU there I live and talking specifically regarding Macbook Air. Base Air costs around 1000 euros in my country. At this price you can get comparable x86 form factor notebook with more RAM (16GB instead 8) and more storage (512/1TB instead 256). However, almost in all cases you also would get:
            - Much worse build quality (seriously, some of the 900-1200EUR range x86 ultrabooks build quality is beyond despicable considering the price);
            - Worse screen;
            - Much worse battery life;
            - Much worse CPU performance;
            - Worse keyboard;

            Even looking at say 1700-2200EUR range ultrabooks situation regarding build quality, screen and kb gets better but you are still getting worse battery life ant CPU.

            I'm really not an Apple fanboy (was using Linux and cost effective x86 hardware almost all my life), but the price/value of 2020 Macbook Air is really amazing. The only catch is MacOS ecosystem of course. I want something similar, at similar price for Linux. That's all. I really hope that at least few good quality Ryzen 7 6800U ultrabooks with configurable power settings will emerge this year...
            Last edited by drakonas777; 12 March 2022, 04:58 AM.

            Comment


            • Originally posted by Developer12 View Post

              Clearly you haven't read their actual progress reports. They already have full control of the chip as far as power management and reclocking. The mainline patches for that are being finished up.

              There's already an open source userspace driver that Asahi have written, which drives the macos kernel interfaces and contains the all-important shader compiler and other bits. It already passes many opengl tests and all that remains after that is to write a thin linux kernel shim to interface with the PCIe device itself. I wouldn't be surprised if they have a prototype (if unusably buggy) linux driver within 2-3 months.
              For OpenGL yes, but Asahi also mentioned several issues, as pretty much entire Metal can be translated to Vulkan, but not entire Vulkan can be translated to Metal. There is some cases where MoltenVK doesn't work that well (because it can't be done) and if MoltenVK has those problems this will be even worse for Vulkan driver on M1.

              Think about it from this persective, DX12 and Vulkan are much more similar to each other (comparing to Metal), Vulkan has bleesing of Intel/Nvidia/AMD, GPUs are designed to work with both so there aren't cases "it cannot be done" and in case DXVK had issues with some stuff, GPU makers were making additional extensions just for their sake and this continues even for DX12 over Vulkan. Metal doesn't have those priviliges.
              Last edited by piotrj3; 12 March 2022, 06:04 PM.

              Comment


              • Originally posted by drakonas777 View Post
                Unfortunately this is almost opposite now, at least in EU
                People always cite build quality, and I don't disagree but it's hard to quantify. Also, I gather Apple has made some missteps on that front, in recent years. So, that's why I shy away from it, in direct comparisons.

                Specs-wise, I mostly mean things like RAM, SSD, and screen (resolution, contrast, gamut, refresh).

                Obviously, (perf * hours / weight) is going to suffer from the whole CPU situation. I'll sacrifice CPU performance before I'd sacrifice battery life or size/weight. However, I also don't tend to run heavy-duty tasks on laptops. For me, they're more about editing stuff (code, docs, emails), network connectivity (chat, video conferencing, remote login), and media playback. Since I didn't travel much, battery life only had to last me through a meeting (at the office) or while I'm away from my desk, at home.

                Even when it comes to my desktops, I'm still getting by with old quad-core machines. At work, the best Linux machine I use has only 6 cores. It's okay, as long as there's enough RAM and a decent SSD.

                My personal laptop is a Skylake i3. I made that decision based on cost, but I've really appreciated the low fan noise. Back then, the only thing separating Intel's 15W processors was clock speed and L3 cache.

                Originally posted by drakonas777 View Post
                I really hope that at least few good quality Ryzen 7 6800U ultrabooks with configurable power settings will emerge this year...
                If you're concerned about noise, I definitely feel that.
                Last edited by coder; 12 March 2022, 07:16 PM.

                Comment


                • Originally posted by piotrj3 View Post
                  pretty much entire Metal can be translated to Vulkan, but not entire Vulkan can be translated to Metal. There is some cases where MoltenVK doesn't work that well (because it can't be done) and if MoltenVK has those problems this will be even worse for Vulkan driver on M1.
                  Really? What if the problem is just a Metal API limitation, rather than an underlying hardware limitation?

                  Originally posted by piotrj3 View Post
                  Think about it from this persective, DX12 and Vulkan are much more similar to each other (comparing to Metal), Vulkan has bleesing of Intel/Nvidia/AMD, GPUs are designed to work with both so there aren't cases "it cannot be done"
                  Apple's GPUs are derived from Imagination Tech's designs, starting in the post-Vulkan era. And Imagination Tech had numerous contributors to the Vulkan 1.0 specification.



                  Also, considering Vulkan's portability goals, I find it hard to believe the core feature set wouldn't work well on M1 hardware.

                  Comment


                  • Originally posted by piotrj3 View Post

                    For OpenGL yes, but Asahi also mentioned several issues, as pretty much entire Metal can be translated to Vulkan, but not entire Vulkan can be translated to Metal. There is some cases where MoltenVK doesn't work that well (because it can't be done) and if MoltenVK has those problems this will be even worse for Vulkan driver on M1.

                    Think about it from this persective, DX12 and Vulkan are much more similar to each other (comparing to Metal), Vulkan has bleesing of Intel/Nvidia/AMD, GPUs are designed to work with both so there aren't cases "it cannot be done" and in case DXVK had issues with some stuff, GPU makers were making additional extensions just for their sake and this continues even for DX12 over Vulkan. Metal doesn't have those priviliges.
                    Who the hell said anything about vulkan on the M1? Yeah it'd be nice but nobody is working on it or talking about it *at all.*

                    If we do ever come to that bridge? You can patch over a LOT with compute shaders. It just won't be pretty.

                    Comment


                    • Originally posted by Developer12 View Post
                      Who the hell said anything about vulkan on the M1? Yeah it'd be nice but nobody is working on it or talking about it *at all.*
                      Well, I had said Imagination's newly open-sourced Vulkan driver could potentially prove useful, in this endeavor.

                      Comment

                      Working...
                      X