Jump to content
Search In
  • More options...
Find results that contain...
Find results in...


  • Content Count

  • Joined

  • Last visited


This user doesn't have any awards

About hishnash

  • Title

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

  1. One option i could see them going to is someting more like the `smart` connector on the ipad that is just a few contacts on the outer edge (supported by magets to help aligned the cable) of the case if you need a wired connection.
  2. The main difference metal has compared to other display technologies is you can issue commands from more than one CPU thread (and even issue draw calls from the GPU itself) what this means practically is that the single core performance is not the limiting factor of your frame rate. in OpenGL, DX and Vulkan you can only issue draw calls (etc) from a single CPU thread. So if you have a complex scene and you want to hit a given frame rate (say 120Hz, like the ipadPro) you need to do all of these calls from a single core on your cpu so you need a fast CPU core. However with metal you can split these calls between many cpu threads and even some GPU shaders and do things like update data on the gpu and issue a draw call without evolving the cpu. This means rather than having a single core running at a high speed you can have a few cors running at a lower speed. This saves a LOT of power since the power a cpu consumes is non-linear compared to its single core speed. This is a big boost when it comes to MOBILE battery performance.
  3. Im not sure you can find any laptops with more raw I/O. Do you use the Function keys (F1 through to F16..) as function keys or are you talking about media keys?
  4. yep its not a Vm, rather a shim. the biggest issue i have found with WSL is IO performance. basically windows is mapping its own apis to linux apis and doing so is slow in some cases were these things dont line up. (like disk and network). You can think of it a bit like WINE but in reverse.
  5. Given the OS uses Metal as the core display tec if you want to write a Display Driver you must support Metal fully, otherwise the user will not even see a boot screen, let alone a window. sure Nvidia could write a user-space driver today that only supports Cuda and they could ship it, there is nothing stoping them doing this. the funny thing is if they write a display driver for Nvidia cards to work on modern macOS then they implicitly also write metal compute driver for their cards as well. This might be harder than you think due to the features in metal that have been added over the years (a load of features that are not in normal display shader languages like being able to dereference a pointer and directly allocating/deallocating memory from within a shader at runtime).
  6. WSL is slow when it comes to filesystem ops. dont expect to do any data-sci work that what writen for a *nix filesytem in wsl if it uses lots of files. and this is not surprising since you need to map from the windows file-system to a vertual *nix file-system these are very different systems.
  7. What is interesting here is with the latest macOS 10.15 nvidia could release a UserSpace Cuda driver (without needing any extra kernel level reviews from apple) the driver would not be able to be used for regular display but for CUDA compute the user space driver api is there and Nvidia could release such a driver at any time if they feel there is a market for Cuda on the mac. Apple explicitly listed accelerator cards as being the main upsides of user-space drivers. they have a load of benefits over normal kernel space drivers: * if the driver has a bug (inf loop or crash) it does not lock up the Kernel * the driver still operations under the same memory limitations as normal programs so cant just read other applications memory (kernel space drivers can since they run within the kernel) * it runs as a user prosses so can be killed at any time, (without crashing the system) also you can have more than one instance running. * much much simpler to debug as a developer * does not require extra code review from apple, normal code signing by the devs account is enough to install (kernal space drivers do since they can bypass any security within the system) * can use mutliple cpu cores etc, and unlike kernal space drivers since it runs in its own prosses you dont risk locking up the system if you take a little longer doing a task.
  8. Native apps can give big benerfics: * download videos/music so you can watch/listen when you are flying (or otherwise not on good enough internet for 4k HDR high bitrate video) * lower battery consumption (important if you want people to watch for many hours) * better system integrations (eg being able to wake up in the background and pre-emptivly download the next episode of a tv show when it comes out so its ready for you)
  9. For BIONC runtime you can set up a linux VM and pass through the eGPU into that that will give you the best OpenCL support. Since macOS and Linux have very simple file system structures you can share a directory without the pain and performance hit you get on windows. This is quite a common workflow for devs of all types on macOS.
  10. If you what to make an Image of a disk then you expect it to be `byte` for `byte` the same as the original. However in macOS you can create Sparse Images, these will appear to be the full (16GB when inspecting from inside) but on disk (outside of the image) they will take up just the space that is needed. When creating an image in Disk Utility did you try selected the `Compressed` option? Both SuperDuper (the link above) and Carbon Copy Cloner https://bombich.com/ are considered the best tools in this space. And as the person above suggests you can also of course just do `dd` to a sparse volume (basically what both of the above tools do) If you don't want a Byte For Byte copy but rather a file for file copy (what you are used to in the windows space) then take a look at this article using sparse bundles and rsync: https://blog.fosketts.net/2015/07/22/how-to-use-mac-os-x-sparse-bundle-disk-images/
  11. It is an FPGA card apple said this back at WWDC.
  12. No this accelerator has nothing to do with the monitor. You will be able to use the screen on and device with good (duel display port over single connection) TB3 support (mostly just macs). The card is just for real-time decoding/encoding of raw 8k pro-res you will still need a gpu
  13. For the Companies that need this will still be cheaper than the equivalent 12 GPUs it replaces! (Just in power costs alone)
  14. I would say pros still want a laptop with good cooling and good battery life. what they could do is make multiple skews so if you don’t need it it does not damage your device.
  15. I suppose the trade off is cooling, you can either replace the Thunderbolt3 ports with others (and reduce your options) or add more ports but have less airflow (or loader airflow) or less battery. given these changes affect all users negitivly, unless you make lots and lots of diffrent case options with different port permutations.. For a user that never uses an ethernet port the space needed within the device for that port is wasted and makes the produce worce for that user. For a user that never uses the SD card reader (and yes there are lots of such users given MBP largest user group is devleopers and they dont normally use SD cards for work) the space used is a wast.