The main thing here is not the move to ARM, but the move from MacOS 10.x (aka OS X, or to be more exact, a very good NeXTSTEP successor) to MacOS 11.x which brings much of the iOS design, turning the Mac into a jail of services. A very, very, very sad day today. We can consider the Mac dead as a UNIX machine, because it's not going to be anymore like that.
Announcement
Collapse
No announcement yet.
Apple Confirms Their Future Desktops + Laptops Will Use In-House CPUs
Collapse
X
-
Originally posted by vladpetric View PostModern Intel/AMD cpus maintain much higher instructions per cycle throughputs (like 3x) versus standard ARMs.
RPI 4 is Cortex A72 cores there are new cores than that in the arm line up.
The Arm Cortex A76 cores have 3x the instructions per cycle throughput as a A72. A76 and new arm standard cores are in the ball park of Intel x86 but AMD x86 not so much when it comes to instructions per cycle.
Items like the RPI4 are built for cost. RPI4 is a 28nm silicon production part this is quite a cheap nm production at this stage. Now to make a A76 arm standard core you have to use at least 12nm production process that is a lot more expensive at this time. A76 arm cores are in production for different phones at this stage.
Originally posted by vladpetric View PostApple's ARM implementation is the only ARM worth using in a high performance context (like, a desktop).
There's good news for Android performance enthusiasts: The Arm Cortex-X1 is a big CPU to compete with Apple's powerhouse processors.
The reality is apple has to work now to stay ahead of reference now that the x series has started.
- Likes 5
Comment
-
Originally posted by oiaohm View Post
This is a true and false statement. True if you are talking about arm cpu you can get on protoboards like RPI4. Not in true for all standard ARM core designs in the market.
RPI 4 is Cortex A72 cores there are new cores than that in the arm line up.
The Arm Cortex A76 cores have 3x the instructions per cycle throughput as a A72. A76 and new arm standard cores are in the ball park of Intel x86 but AMD x86 not so much when it comes to instructions per cycle.
Do you actually have numbers that A76 has 3x IPC as A72? Keep in mind that having more issue slots is necessary for high IPC, but hardly sufficient.
As for my 3x claim - it actually comes from my own workload. It's not exactly 3x (more like 2.96) but whatever.
I'm also comparing RPi4 with a 5-year old Haswell, which is built on 22nm FWIW.
Finally, what a vendor claims when it comes to performance on their front page is highly unreliable. Third party benchmarks ...
So, do you have a third party, single threaded benchmark that shows roughly 3x IPC for either A76 or ARM X series, versus the processor in A72 in RPi4?
- Likes 3
Comment
-
Originally posted by oiaohm View Post
This is not true. The top500 supercomputer list the top one at the moment is using arm standard cores just the latest generation of them and 48 cores per chip. There are quite decent high performing arm standard cores these days issue is you have to use 14 nm or better to use the standard design that is not cheap but it will get cheaper as production moves to 5nm. Apple is most likely to spend the money to by access to 7nm and 5nm silicon even if they are using arm current standard designs that is going to perform fairly well. Apple own design is not that far head of the best Arm standard reference design at this point.
TTBOMK they don't even have a proper standardized benchmark suite to measure multithreaded performance for the top 500 (e.g., the phoronix suite).
Another thing is that oftentimes the supercomputers operate with a high-throughput, but also high latency model (throughput-oriented computation). You can put a lot more wimpy cores and get great performance this way, but your latency is not going to be great. An example is the Xeon Phi line - you get as much as 256 threads per chip, but you don't get great latency from one thread.
For the most part desktops and servers use really beefy cores that deliver high IPC and low latency (latency-oriented). They tend to perform well on a broad variety of workloads. There are situations in which the high-throughput/high-latency systems do better, sure.
Ok, so let me clarify - when I said high performance I really mean low latency.
- Likes 3
Comment
-
Originally posted by oiaohm View Post
- Likes 4
Comment
-
I'm going to take the plunge and guess Apple's transition to their own ARM chips is going to help app compatibility for Windows 10 on ARM and thus, in the long run, ARM on Linux.
And is it just me or are Apple's keynotes always so well choreographed? As a former Mac user I always want to go out and switch all my devices to Apple before realizing that would really not be a good idea. Even as a gamer macOS has much less game support then Linux (especially since they don't have DXVK and 32 bit apps can't run anymore).
- Likes 2
Comment
-
Originally posted by torsionbar28 View PostThere is nothing "eco" about the Apple company or its products. Let us count the ways... planned product obsolescence, non replaceable batteries, poor repairability, lack of support for 3rd party or DIY repair, frivolous fancy packaging, economic support of oppressive communist regimes with abysmal human rights records (China), sweatshop slave labor conditions in their offshore factories. Did I miss any? I'd wager that Apple is responsible for an outsize portion of global e-waste and human suffering.
- Likes 7
Comment
-
Originally posted by WolfpackN64 View PostI'm going to take the plunge and guess Apple's transition to their own ARM chips is going to help app compatibility for Windows 10 on ARM and thus, in the long run, ARM on Linux.
And is it just me or are Apple's keynotes always so well choreographed? As a former Mac user I always want to go out and switch all my devices to Apple before realizing that would really not be a good idea. Even as a gamer macOS has much less game support then Linux (especially since they don't have DXVK and 32 bit apps can't run anymore).
- Likes 1
Comment
Comment