I’m really happy with how this episode came out. I hope y’all enjoy it.
Next week: Our macOS Big Sur episode!
I really enjoyed the interview, so thanks for that.
I am very excited about my MacBook Air M1.
I loved my 2015 MacBook, with the lack of fan and ease of portability. Didn’t like the only-1-port and didn’t like having to get the keyboard replaced 3 times.
I kept it until 2019 when I upgraded to the MacBook Air, which I didn’t love, but the 2015 MacBook was no longer reliable.
In April of 2020 I upgraded to a 16" MacBook Pro because I needed more “oomph” for all the video/Zoom/WFH stuff that I was doing. Loved the horsepower. Got a great dock so I don’t actually use more than 1 port (when I’m at my home office desk anyway).
I don’t love the size. It’s fine on a desk, but on the couch it’s unwieldy.
I don’t love the Touch Bar. I tried. I really tried. I even tried Better Touch Tool. But I just don’t like it, and I only use it by accident.
The good news is that I can trade my MBPro in for a new MacBook Air with 16GB and 1TB SSD (which is what I have now) as basically an even exchange.
I’m more excited about this than the previous MacBook Air or the MacBook Pro because it reminds me of the MacBook (but with an extra port, which everyone always said was a big drawback) and not only is this one not underpowered, but it’s likely to equal to my MacBook Pro, but in a much smaller form-factor and without a fan.
Did I mention there’s no fan? I’m a big fan of no fan.
And no Touch Bar.
Plus, being able to run iOS apps, which seems like it might be cool… as long as the developers of some apps that I’d like to use don’t opt-out, I guess.
Everybody and their brother is talking about binning now. I think there is a lot of confusion on this manufacturing yield management process.
TL;DR binning is a way to salvage defective chips by re-using them as lower quality or lower performance chips. This ups the overall yield in semiconductor manufacturing helping to lower costs and/or increase profits.
Binning sorts can be based on cpu speed, power usage, heat, or other criteria. Semiconductors are manufactured on large wafers with hundreds or more of the same chip built at the same time. It is never possible to achieve 100% quality, so there are always some defective chips on every “batch”. Binning allows using the failed chips (by spec’ing them at lower speed, lower number for GPU or CPU units, or less efficient heat or power) instead of throwing them away.
New chip processes, like going to 5nm for Apple Silicon, will have relatively low yields. It takes time and experience to get the defect rate down. So over time the same chip manufactured the same way will cost less to make as the yields increase - even before more dramatic die shrinks or redesign.
LED lights, ram chips, and many things other than SoC or processors are regularly binned. It’s a part of any semiconductor quality control/manufacturing process and by itself is neither good or bad.
There is also a lot of confusion about SoC (system on a chip) and memory subsystems.
The Apple Silicon SoC does not have the 8GB/16GB of main memory (“unified memory”) on the SoC. That wouldn’t fit. 1 GByte of ram is approximately 8 billion transistors. So do the math.
Apple SoC is using state of the art MCM (multi chip modules). Multiple silicon dies are combined into a single module and that module is encased in epoxy or similar material so it looks like a single chip from the outside.
Apple SoC is most likely also using 3D stacking where one or more dies are physically stacked on top of each other to optimize interconnection and minimize physical space.
Dram memory are huge arrays of the same circuits repeated over and over to build up to the desired total memory. Processors and GPUs will have onboard buffers and transient memory for performance and working storage, but the primary “main memory” is still a large dram array which is not on the same chip.
Controversially, “unified memory” is a marketing term that takes a disadvantage - not having dedicated high speed memory for video, graphics, or other specialized processing and attempts to turn that into an advantage.
The original IBM PC, Apple, and other computers of the 1980’s all featured “unified memory”. That’s simply how processors and computers were designed. It wasn’t until more advanced video interfaces and graphics processors were created with dedicated video buffers and high-speed specialized memory arrays that unified memory became a major disadvantage/weakness in simpler systems.
This is much more difficult area to analyze, as having a single shared pool of memory that can carved up as needed into program, video, and gpu memory on the fly instead of having permanent separate memory arrays can be optimal, but will not be faster than dedicate memory for higher-end very high performance systems.
That’s why dedicated GPU/eGPU’s have large amounts of higher-speed video and graphics processing memory. It would be too expensive to build a consumer level desktop computer all the memory running at the speed and performance of dedicated memory found on GPU cards.
There is a latency/time penalty for moving data between regular memory and higher speed video/graphics memory, but that is why high speed buffers, highly parallel memory arrays, multi-ported direct memory access, and other advanced techniques (translation: much more expensive chip design) are used to eliminate or minimize those penalties but that results in more complex silicon design.
Apple Silicon is a tremendous step forward in chip and computer design, but as Scotty said “you still can’t break the laws of physics”
This was great! I had a bit of a “hair-blown-back” moment when @MacSparky said “and this is the slowest Apple Silicon that will ever be released!”
I had to pause playback and let that soak for a minute. He’s absolutely spot-on! These speeds and performance are the new “low” benchmarks for each generation that comes after it. It’s truly incredible to think about when put into context like that.
Thanks for the Touch Bar comment. I’ve tried it on other people’s machines and want to like it…but I don’t. Guess it’s the new MBA for me!
@SpivR - does the idea [memory swapping will be faster with the M1 + SSD and, therefore, compensate for less amount of memory, for “regular users”] stand , or is it incorrect?
Unfortunately, no way to tell without benchmarks which are notoriously biased and hard to compare easily between different architectures.
I have no doubt the Silicon Macs will perform very well. I just think Apple is pushing the marketing levers strongly here.
The real proof is probably 12 months away or longer when we see a next generation of iMac, iMac Pro and maybe higher end Mac mini and MacBook Pros and see whether Apple continues this design or has their own discrete GPU with separate memory arrays.
There is also a lot of pushback in PC circles that Apple’s “faster than 98 % of all PC’s sold last year” is vague too. A boatload of $150 to $300 PC’s sold are included in that stat, so without actual numbers and what kind of tests they ran (synthetic benchmarks, actual workload equivalent tasks, etc.?) it is somewhat meaningless.
Again, Apple has increased performance and battery life amazingly more than Intel Macs, but the bragging rights and marketing claims don’t deserve a blank sheet “pass” and should be grounded in more quantitative data.
The fact that Apple did not offer any direct comparisons of Silicon Macs versus current Intel Macs, iMacs, iMac Pro, and Mac Pro shows they are more interested in marketing hype than showing actual comparisons and real performance data. Can’t blame them, they don’t want to Osbornize currently selling Intel Macs prematurely.
It makes sense.
The M1 is going the opposite direction as far as user-replaceable parts goes. Take my 2019 Macbook Pro that I purchased in March… I got 32GB RAM and 2 TB drive, which I thought would be enough for the next 4-5 years. My local storage needs went up dramatically in 5 months later, and I’m already using half of my storage. It’s going to go up again soon , so I’ve been researching my options for upgrading the hard drive.
So far the memory is sufficient, but I really wish that I could get an internal 4TB drive for my Macbook Pro. I’ve tried external TB drives and they just aren’t practical.
It looks I’ll probably have to jump off the Apple computer train for my next computer unless Apple changes course and offers the ability upgrade hardware, whether self-service or by an Apple technician.
I wouldn’t count on future computers being any more user-upgradeable than these are. That’s just not the direction things are going. I don’t love it either (I just put a new SSD in my 2008 Black MacBook), but that’s where we are.
Not that I condone such actions, but a motivated person could go the Hackintosh route
For as long as new macOS releases support Intel, at least.