Beyond the M1 hype and why the fairy dust needs to settle.

[Fomalhaut]

People keep bringing up the horrific to repair Surface.

Yes, true But the Surface is a direct competitor to the M1 Air/Pro, so it's a fair comparison.

I don't have a problem with this. I view these ultra-thin laptops as appliances, not a modular system that I'm every going to upgrade. In a Mac Mini....yes, I would like an additional M.2 connector. There is plenty of space for it. modular RAM is less likely now that it's moving to the SoC package.

 

[bobcomer]

I've been eyeing the Thinkpad X1 Nano too for an ultralight 2# but prefer one with AMD 5000U. Resale value of M1 is bad and have seen MBA M1 16GB/1TB for $1K so I either have to take a big loss or just hang on to it and hope an OS update fixes the major issues. Not hopeful it'll fix things like slow disk I/O though since that's likely a hardware limitation.

I like the X1 Carbon better. Not much heavier, but a bigger screen. My last 3 work laptops have been X1's of some kind or another.

 

[chouseworth]

What is SO great about the M1 macs?
They offer less than Windows counterparts. No real gaming support, no support for other OS natively, no touch and VERY VERY limited app compatibly. Sure its faster than i7 11th gen but AMD processors offer greater performance and around the same battery life as the M1.

The AMD Ryzen 7 4800U offers faster performance than an M1 Air/Pro and there are laptops that have that processor that are cheaper than the M1 Air with upgradable SSD and RAM.

Now with the SSD swap issue that Apple is quiet on is very serious IMO. I have an intel 16" MBP and I have written about 7TBW and I got this machine around January 2020 and I use this laptop very heavily everyday. The fact that I see people writing over 15TBW on their M1 macs that they got 5-6 months ago is very concerning.

All I am saying is look beyond the M1 hype and see that you are getting a computer with less features, no upgradeability and limited third party software. I say this because I see some people say the M1 Air is the best deal for an Ultrabook, I strongly disagree with that claim.
The reason the M1 macs seem so good is because the previous Macs were utter garbage in terms of specs and price to performance ratio.
Ever wonder why Rosseta 2 runs Intel software better on M1 macs than on intel macs is because those intel's that Apple replaced were not at all performant.
The M1 Air had a quad core i7 a weak one at that, the M1 Pro had a 8th gen i5/i7.

For $920 on the Windows side you can get a HP ENVY x360 with a FHD screen(1080p), Ryzen 7 4700U, 16GB RAM, a 256GB SSD(user upgradable) and a 1000 NITS display with touch. Click here to see HP Envy configure page. Yes it comes with Windows but Windows can do a LOT more than macOS can ever can.
The argument that macOS is better than Windows is no longer true as Windows vastly outperforms macOS in almost everyway. It's now even more obvious with the M1 macs.

I know I can't tell people what to buy or not, but people have been making extraordinary claims on YouTube, twitter and other social media
forums that M1 macs is the future and outperform most laptops and are the best value out there and I just wanted to clarify some points.

EDIT:
Ok I been researching the M1 more. It only consumes 15Watts max for the CPU alone. Thats very impressive.
The 4700U Ryzen costumes 40 watts max, not really as the spec sheet states which is 15watts. But after a while it comes to 15 watts.

Whereas the M1 goes up to 15Watts for the CPU only. NOW that is impressive. Can't wait for future Apple Sillicon now.

View attachment 1755852
source for watt info: https://www.anandtech.com/show/16084/intel-tiger-lake-review-deep-dive-core-11th-gen/7

I think the market has spoken. IDC just reported that Apple Mac shipments more than doubled (increased 110%) year over year from 1Q20 to 1Q21.

 

[BigPotatoLobbyist]

“I think the market has spoken. IDC just reported that Apple Mac shipments more than doubled (increased 110%) year over year from 1Q20 to 1Q21.”


I) Premium PC sales more generally were given a shot in the arm seeing as consumers were forced to spend an increasing sum of time with their computers (WFH), and it makes sense.
II) Butterfly keyboards plagued Macbooks for several years, this error was remedied in the Spring of 2020.
III) Definitely the M1 is a part of the growth and especially will be going forward (I think MacOS marketshare may even increase because of the ability to sell performant entry-level Airs now).

But I) and II) can’t be ignored in any honest narration of the data at hand.

 

[Joelist]

Actually the keyboards went to Magic starting in 2019. Interestingly, Apple's Mac revenue spiked year over year by 26% since the M1 Macs started selling (Statista).

 

[throAU]

The whole repairability thing is irrelevant as far as big business TCO goes.

An enterprise (e.g., like one where I work at) calls the vendor for repairs/replacement under warranty and Apple are easier to deal with. i.e., how difficult the thing is to repair is the vendor's problem, not ours. If it's out of warranty and fails it is likely due for upgrade anyhow and the machine is disposed of and replaced.

It's simply not worth spending hundreds of dollars in time to fix a machine that is fully depreciated/paid off (worth nothing on the books) and holding back your ability to deploy newer software. Chances are, you'll spend money to fix it and some other equally old component on it will fail shortly afterwards anyway.

But even then, even if we assume that repairs/replacement are the same - the big cost of ownership is end user support, software installs, software license tracking, SOE development/roll out tools, patch management, de-spywaring, dealing with application breakage, backups, etc. This is what contributed to IBMs experience that Macs were cheaper to deploy.

Every help desk ticket you log likely costs in the order of $60-80 even if it is a 15 minute fix - even if we assume a fairly nominal $20-40 hourly rate (plus benefits, cost to host the additional staff required or support contract volume, etc.) for all people involved..

• your lost time/productivity being unable to do your job while waiting for help with the machine
• your time to log the job and wait for a response (email IT, fill out web form or call and be on hold, etc.)
• the tech's time cost @ hourly rate
• the cost to log the actions/close the job out

Any ticket you can avoid logging is going to save the company 60-80 bucks in tangible expenses for an enterprise even if the solution is as simple as "yeah click the right mouse button instead of the left one" or helping the user out with "how do I sync my phone?".


2-3 tickets logged or not logged over the life cycle of the machine (lets say ~ 3-5 years) for anything more than trivial 15 minute fixes and you're looking at a 200-300 dollar cost between a PC and a Mac as being easily eaten. And most PC trouble tickets aren't 15 minutes to fix.


Now... if your mission critical software doesn't run on Macs then you've got to run Windows or whatever. But in 2021 you should be migrating that stuff to web UI anyway which makes the end user platform irrelevant...

Sure, most aren't there yet, but IBM is.

 

[Kung gu]

Currently using the G14 with a Ryzen 9 and RTX 3060 GPU. Makes no noise in silent mode and the battery lasts 9-11 hours with casual use in this mode. The first RAM slot is partially soldered, unfortunately, but I have a second RAM slot with 32 GB, so it currently has 40 GB of RAM. The SSD is user replaceable though, like the Nano.


View attachment 1757164


The Nano isn't a contender as it has garbage battery life. It has also an Intel chip, and nothing from Intel even remotely competes with the M1 for power and energy consumption. The laptop I have now is about as close to an M1 as I'm gonna get with a Windows laptop. Absolutely powerful in Turbo mode, but that requires being plugged in due to the extra power usage required.

There is a reason why your G14 gets similar battery life to an M1 Pro(although I think the M1 in casual use would get better battery life) in casual use, is because the battery capacity is much higher.

The G14 has a 76 Watt hour battery vs 58.2 Watt hour in the M1 Pro. Now if the upcoming 14"MBP has a higher battery capacity than the 13" M1 Pro and similar battery capacity as the Asus G14, the battery life of the 14" will be much higher than the G14 in casual use.

 

[Kung gu]

I've been eyeing the Thinkpad X1 Nano too for an ultralight 2# but prefer one with AMD 5000U. Resale value of M1 is bad and have seen MBA M1 16GB/1TB for $1K so I either have to take a big loss or just hang on to it and hope an OS update fixes the major issues. Not hopeful it'll fix things like slow disk I/O though since that's likely a hardware limitation.

The M1 is an iPad Pro chip with some Mac features like virtualisation and Rosseta IP. It's basically an A14X.

The true Mac chips will arrive later in the year as Bloomberg first said the Mac chips with be a 12 core processor, but instead we got a 8 core processor that was created for iPad Pro first. No wonder Apple used them in the M1 Air and low end Mac mini and 2 port MacBook Pro. The real Mac SoC will be in the 24" 27" iMac and 14" 16" MBP and high end Mac mini.

You can see the limits in the M1 SoC IO as it only supports 1 external display for MacBooks and can only have 2 USB4 ports. Clearly passed down from the iPad Pro style chip, as the iPad only supports 1 monitor.

My opinion is that we are yet to see the real Mac SoC.

 

[crazy dave]

“I think the market has spoken. IDC just reported that Apple Mac shipments more than doubled (increased 110%) year over year from 1Q20 to 1Q21.”


I) Premium PC sales more generally were given a shot in the arm seeing as consumers were forced to spend an increasing sum of time with their computers (WFH), and it makes sense.
II) Butterfly keyboards plagued Macbooks for several years, this error was remedied in the Spring of 2020.
III) Definitely the M1 is a part of the growth and especially will be going forward (I think MacOS marketshare may even increase because of the ability to sell performant entry-level Airs now).

But I) and II) can’t be ignored in any honest narration of the data at hand.

I think IDC pegged Apple’s Mac growth as twice the growth in the overall PC market. However you are right that this is not quite the same as comparing just “premium” category products. But previously in the pandemic a lot of the biggest growth was being seen in things like chromebooks rather than the premium segment ($800+).

As @Joelist said the magic keyboard is appeared on previous Mac models.

 

[crazy dave]

You can see the limits in the M1 SoC IO as it only supports 1 external display for MacBooks and can only have 2 USB4 ports. Clearly passed down from the iPad Pro style chip, as the iPad only supports 1 monitor.

My opinion is that we are yet to see the real Mac SoC.

*Some* of this may be software rather than hardware. For instance corellium reports that in their M1 Linux port TB3 works with eGPUs and I believe they said it connect to more displays. Can’t quite remember that latter one.

 

[thenewperson]

*Some* of this may be software rather than hardware. For instance corellium reports that in their M1 Linux port TB3 works with eGPUs and I believe they said it connect to more displays. Can’t quite remember that latter one.

If the latter one is true then I really wonder why Apple limited it to 1 display...

 

[crazy dave]

If the latter one is true then I really wonder why Apple limited it to 1 display...

I’m not sure they did, can’t find it anyway. I should also stress that getting eGPUs working just meant that it could see eGPUs, more than what can be done on the m1. Actually accelerating graphics on it is another story that exists beyond just thunderbolt.

 

[leman]

The whole repairability thing is irrelevant as far as big business TCO goes.

An enterprise (e.g., like one where I work at) calls the vendor for repairs/replacement under warranty and Apple are easier to deal with. i.e., how difficult the thing is to repair is the vendor's problem, not ours. If it's out of warranty and fails it is likely due for upgrade anyhow and the machine is disposed of and replaced.

It's simply not worth spending hundreds of dollars in time to fix a machine that is fully depreciated/paid off (worth nothing on the books) and holding back your ability to deploy newer software. Chances are, you'll spend money to fix it and some other equally old component on it will fail shortly afterwards anyway.

But even then, even if we assume that repairs/replacement are the same - the big cost of ownership is end user support, software installs, software license tracking, SOE development/roll out tools, patch management, de-spywaring, dealing with application breakage, backups, etc. This is what contributed to IBMs experience that Macs were cheaper to deploy.

Every help desk ticket you log likely costs in the order of $60-80 even if it is a 15 minute fix - even if we assume a fairly nominal $20-40 hourly rate (plus benefits, cost to host the additional staff required or support contract volume, etc.) for all people involved..

• your lost time/productivity being unable to do your job while waiting for help with the machine
• your time to log the job and wait for a response (email IT, fill out web form or call and be on hold, etc.)
• the tech's time cost @ hourly rate
• the cost to log the actions/close the job out

Any ticket you can avoid logging is going to save the company 60-80 bucks in tangible expenses for an enterprise even if the solution is as simple as "yeah click the right mouse button instead of the left one" or helping the user out with "how do I sync my phone?".


2-3 tickets logged or not logged over the life cycle of the machine (lets say ~ 3-5 years) for anything more than trivial 15 minute fixes and you're looking at a 200-300 dollar cost between a PC and a Mac as being easily eaten. And most PC trouble tickets aren't 15 minutes to fix.


Now... if your mission critical software doesn't run on Macs then you've got to run Windows or whatever. But in 2021 you should be migrating that stuff to web UI anyway which makes the end user platform irrelevant...

Sure, most aren't there yet, but IBM is.

Yep, that's pretty much it. Now, large organizations might be able to afford having IT support stuff that does minimal upgrades/repairs but as you say it makes much more sense financially and administratively to just buy premium support from a third-party service professional and let them deal with hardware. Computers are dirt cheap compared to labor (at least in post-industrial countries) and any downtime ends up costing $$$.

There is a reason why your G14 gets similar battery life to an M1 Pro(although I think the M1 in casual use would get better battery life) in casual use, is because the battery capacity is much higher.

There is no way that G14 gets similar battery life to M1 Pro, battery capacity notwithstanding. Maybe if you set brightness to minimum, turn off WiFi and just leave it there to do nothing. For actual use, M1 will last at least twice as long. The new G14 will probably do slightly worse than the 2020 version due to more power-hungry display panels.

*Some* of this may be software rather than hardware. For instance corellium reports that in their M1 Linux port TB3 works with eGPUs and I believe they said it connect to more displays. Can’t quite remember that latter one.

The eGPU part is definitely software, there is no technical reason why an eGPU would not work with M1's thunderbolt. ARM Big Sur simply does not include any third-party GPU drivers and I personally believe that it's in Apple's best interest for it to stay that way to make developers commit to their proprietary GPU programming model.

 

[throAU]

Yep, that's pretty much it. Now, large organizations might be able to afford having IT support stuff that does minimal upgrades/repairs but as you say it makes much more sense financially and administratively to just buy premium support from a third-party service professional and let them deal with hardware. Computers are dirt cheap compared to labor (at least in post-industrial countries) and any downtime ends up costing $$$.

Yeah to be clear when i say "out of warranty" i'm counting 3 year warranty. Like you'd get with AppleCare+ or Dell/HP extended warranty, etc.

Big business would normally buy that and run the machine until the warranty is out, any failure after that = bin, but you bank on retiring the machines at 4-5 years anyway to roll out new OS (again, pushing a new OS on your old fleet is pretty pointless waste of time for crap hardware at that point anyway).

OR - the other common practice is to lease/hire the machines. Dead machine = call vendor and they drop a new one out.

If you have proper data management (i.e., users are storing their stuff in server/cloud storage, or you at least do user workstation backups for those who don't), replacing a machine is relatively painless.

They log in their profile/data comes down, apps get auto-pushed, etc.

If trashing the machine and replacing it involves significant downtime for your users (and you've got more than a couple hundred employees), you're doing it wrong...

 

[leman]

Big business would normally buy that and run the machine until the warranty is out, any failure after that = bin, but you bank on retiring the machines at 4-5 years anyway to roll out new OS (again, pushing a new OS on your old fleet is pretty pointless waste of time for crap hardware at that point anyway).

Our service partner was offering us a 4 year warranty for a very reasonable price, so that's what we did.

If you have proper data management (i.e., users are storing their stuff in server/cloud storage, or you at least do user workstation backups for those who don't), replacing a machine is relatively painless.

Centralized solution was not an option for us, but every user was doing their own time machine backups, and we always kept some spare computers for the unlikely case that something breaks. In an event of failure we were able to provide a replacement machine with fully restored data in a time that it would take an employee to have lunch. All in all, it worked really well...

 

[Unregistered 4U]

The eGPU part is definitely software, there is no technical reason why an eGPU would not work with M1's thunderbolt.

There IS a technical reason why performance would suffer greatly, though.

 

[leman]

There IS a technical reason why performance would suffer greatly, though.

Could you expand on this a little bit? Not 100% sure what you mean. Are you referring to lack of UMA on eGPUs?

 

[Krevnik]

Intel still has compatibility and supportability where AMD still lacks. I have a few friends that have AMD systems and they ALL experienced the recent USB fiasco with AMD.

AMD Locates 'Root Cause' of USB Issues on B550, X570 Motherboards

A patch will arrive through the upcoming AGESA 1.2.0.2 update, which motherboard vendors will likely start rolling out in early April via beta BIOS updates.

www.pcmag.com

I would rather sacrifice a BIT of performance than deal with some random USB issues when I need it most.

Or the RAND bug with the 3000 series that made systemd issue a hotfix, and broke Destiny 2 for a couple months for folks that cared.

As I said at the 3000 series launch, AMD seems to have some issues making sure their stuff is rock solid at launch. Intel hasn’t been perfect here either, but they haven’t recently released a CPU that just randomly regresses something that worked in the previous generation to my knowledge. Let alone twice in two years. And this sort of thing does matter for B2B sales for the reasons you state. CPUs are not something you should take the “day one patch” approach with.

It doesn’t matter if the bug affects a minority of users, if those users are effectively blocked from doing what they want with the system. If I was looking to buy hundreds or thousands of machines for a business, I’d think twice about AMD in that setting when buying recently refreshed hardware, or risk getting bit by the next random regression.

And I say this as someone who’s gaming rig is an AMD system, and where I have no plans to go Intel with my next upgrade.

 

[GrumpyCoder]

And this sort of thing does matter for B2B sales for the reasons you state.

AMD is a bad choice when it comes to B2B or critical systems. When I bought desktop machines last year (Dell with Xeon Platinum and RTX8000) I seriously looked at Threadripper/EPYC systems. I probably could have lived with the 512GB memory limit in these particular machines, but the fact that AMD simply can not supply hardware in quantities is not acceptable. I spoke to a bunch of resellers and they all told me, that they usually have their AMD systems back up and running within a week. Sometimes it can take 4-6 weeks and in urgent cases they might replace the hardware with something else that is available. Sorry, that doesn't work for me for business machines. If a critical systems fails, I want to email/phone/ticket the supplier and expect to have it back up and running within 4 hours, which is done by one of their techs on site and without having to change drivers, etc.

Easy experiment, place an order for 10000 AMD CPUs and 10000 Intel CPUs and see which order arrives first.

I'm running some AMD systems in non-critical applications, so the above doesn't apply. Also for home use, where downtime is acceptable and playing some games is the primary use, none of this matters either.

 

[CMMChris]

*Some* of this may be software rather than hardware. For instance corellium reports that in their M1 Linux port TB3 works with eGPUs and I believe they said it connect to more displays. Can’t quite remember that latter one.

Nope, Corelliums Linux Port does not support more monitors. It doesn't even come with drivers for the GPU since there are none available at the moment.

The reason M1 Macs do not support eGPUs is that macOS simply doesn't offer any ARM drivers for any third party GPUs. If Apple would port their AMD drivers to ARM, eGPUs would work just fine.

 

[Unregistered 4U]

Could you expand on this a little bit? Not 100% sure what you mean. Are you referring to lack of UMA on eGPUs?

Not specifically UMA, as external GPU’s COULD have access to the memory, it would be more that Thunderbolt, regardless of how fast, won’t come anywhere near the access speed that the SoC provides. It actually has more to do with the rendering technology, TBDR vs. IMR. Apple has stated that rendering for Apple Silicon will be TBDR which comes with a lot of assumptions, including that there must be a very fast connection to a common pool of memory.

Apple COULD, at any time, change this stance, of course. BUT, as long as they stick with TBDR, then certain potential options would be far less likely due to the performance hit.

 

[crazy dave]

Nope, Corelliums Linux Port does not support more monitors. It doesn't even come with drivers for the GPU since there are none available at the moment.

The reason M1 Macs do not support eGPUs is that macOS simply doesn't offer any ARM drivers for any third party GPUs. If Apple would port their AMD drivers to ARM, eGPUs would work just fine.

The eGPU part is definitely software, there is no technical reason why an eGPU would not work with M1's thunderbolt. ARM Big Sur simply does not include any third-party GPU drivers and I personally believe that it's in Apple's best interest for it to stay that way to make developers commit to their proprietary GPU programming model.

I understand the reason why there is no eGPU support on the M1 on the Mac side. The multiple monitors was a brain fart. But, like with eGPUs, it still may be software enforced rather than an intrinsic limitation of the SOC hardware though which is what I was responding to.

Not specifically UMA, as external GPU’s COULD have access to the memory, it would be more that Thunderbolt, regardless of how fast, won’t come anywhere near the access speed that the SoC provides. It actually has more to do with the rendering technology, TBDR vs. IMR. Apple has stated that rendering for Apple Silicon will be TBDR which comes with a lot of assumptions, including that there must be a very fast connection to a common pool of memory.

Apple COULD, at any time, change this stance, of course. BUT, as long as they stick with TBDR, then certain potential options would be far less likely due to the performance hit.

I’m not quite sure what you’re trying to say. Metal works with both TBDR and IMR. While of course an eGPU suffers from having to connect over TB versus being on SOC, a powerful enough GPU will still be worth it just as it was on Intel Macs. As @leman says there may be other reasons why Apple would want to discourage eGPUs, but I don’t see why an eGPU would necessarily be too slow to be useful on an M1 Mac as compared to an Intel Mac. Could you explain further?

 

[GrumpyCoder]

While of course an eGPU suffers from having to connect over TB versus being on SOC, a powerful enough GPU will still be worth it just as it was on Intel Macs.

What makes you think the SoC isn't as powerful as an external GPU? Apple could very well target the 3090 and if they can do it, game over. There's still the market of data centers, but that's never been interesting for Apple.

Apple could bring back a lot of things, including OpenGL and Vulkan. I doubt they will.

 

[robco74]

I would guess that the lack of GPU support was more of a business rather than technical decision. There may be too few users using eGPUs to make supporting them in the first release worthwhile. Apple may not want to continue tasking engineers with writing the drivers and supporting them. Everyone tries to get their features in to each release, but it just isn't possible to include everything and maintain quality. Tough decisions get made and features get bumped.

In the case of Nvidia, they tend to want to write drivers themselves. For example, they won't release the info needed for open source drivers and will only release binary blobs for Linux. AMD and Intel OTOH are happy to release chip specs and let Apple write their own drivers. In that regard, it's more a case of Nvidia and Apple wanting more control over the end-user experience, not necessarily bad blood.

 

[cmaier]

I think the market has spoken. IDC just reported that Apple Mac shipments more than doubled (increased 110%) year over year from 1Q20 to 1Q21.

I agree that M1 has been wildly successful. I also think IDC's numbers aren't accurate.