Why Apple don't use NVME m2 slot for mac mini/mac studio

[dmccloud]

You confuse me no end, define uneducated guess? This research is quantitive and empirical, testing common SSD across most manufacturing.

Do you think a "brand" makes the difference when it comes to actual scientific research / testing?
Or should it be what they put in to the product?

What makes apple internal SSDs different besides the SSD controller being part of the SoC?
Are the memory bank chips specific to apple?
Do they fabricate them in some special way?
How does this make them 100% resistant to SSD wear?

Your conclusions are based upon an assumption that the SSDs Apple is using are virtually identical to the third party SSDs from other manufacturers. It's already been well established that some of the controllers are on the SSD itself with Apple products rather than in a chipset or CPU, so there's already one major difference your assumptions do not account for. Since there's no published documentation specific to Apple's SSDs, all of your assumptions are just uneducated guesses based upon stereotypes of SSD technology as a whole.

In the consumer eyes, if the company is loyal and trustworthy (which apple tends to be), then they can go by this and provide loyalty. But in reality, apple is a business and will do what is best for business. The question really is what is best for business? And that is why i continue to believe that a company investing in consumer level AI research and not just simply offer a high end enterprise solution, would increase their sales, and change the market fully. Brining on a consumer level AI boom, and not just a industrial one.



Normal is relativistic, and dependent on project that a person is doing at that time, so i think when you say normal, you mean normal for you?



I have said many times why i am using this, and why replaceable ssds are needed. Consider this. £40 a time for a 1tb NVME ssd Samsung pro 970 vs 1.3k replacing a mac mini m2 pro.... hmmm

I think you are a) severely underestimating the durability of modern SSDs and b) even more severely overestimating how much data you will be reading/writing to any SSD. Consequently, a lot of your arguments are based on speculative inferences rather than measurable facts or observable behaviors.

Regarding the steel toe cap boots, i fear you have no army experience and perhaps not been in cadets? as well, army boots and steel toe cap boots are essentially the same thing, so i do not understand your remark regarding " as useful here as a pair of steel cap working boots when running over a minefield?

You can buy steel toe boots virtually anywhere, even at WalMart - they're often used in construction or related fields where heavy objetcs, tools, etc. are often carried on the job. They are also not even remotely similar to military issue boots used by the Army, Navy, Air Force, or Marines, which is why the analogy is valid. Whether or not someone has been in the military has absolutely no relevance to the analogy or discussion at hand.

 

[Kr0n05K!ngR]

Hello mr_roboto, thank you for the reply.

What do you mean by "kernel mode access", and why would you need it for this application? All you need is storage. Machine learning should be a regular userspace process.

Kernel mode access, from a programming perspective should be enough to determine what i am talking, but it can depend on the experience of the user. Milage may very, as they say. Sorry your having problems understanding it.

Background:

Kernel mode access: "drivers that run in the same memory space as the operating system kernel, which is the core component of the system that manages resources, processes, and security. This means that kernel mode drivers have direct access to the hardware and can perform faster and more efficiently than user mode drivers."

Ref: https://support.apple.com/en-gb/guide/security/sec8e454101b/web

Anything that is not run within the kernel, is sandboxed in a different memory space, and apple use programs such as "gatekeeper" to manage this. You can add apps using spctl command, and whats supposed to happen is gatekeeper to prompt you when an app is ran, if have set security level within the macOS recovery environment > startup security utility > reduced security > allow user management of kernel extensions from identified developed.

The SIP commands can disable protection and the SPCTL command to disable protection, both do not work when running the os off of an external NVME, hence “kernel mode access is being refused” regardless of the settings the user makes. Again this is M2 mini pro Mac.

And i am not just speaking about machine learning, i am also talking about FOSS running on the mac, such as veracrypt having access to mount and unmount encrypted partitions.

Also your generalising machine learning, as some need drivers ran in the memory space of the kernel. This can also be achieved installed on a virtual machine that has kernel access within its own elastic environment, which is the case for your VPS i believe. But this is software or would need a redirect to hardware resource, which some GPU manufactures provide API access for, which can require an additional licence. With bare metal, in no VM i need local kernel mode access. This can give better access the computer hardware, such as the neural engine.

Also, what would be different about booting from an external? You gave a fairly weird answer before that's hard to make any sense out of.

Explain whats wired about wanting to be in an environment that allowed for replacement of an SSD when it reaches TBW? Once it is dead in the mac mini (soldered on) its dead.

How is this strange to you?

What Is not understood? Help me understand you.

Also, who cares where you boot from, why can't you just tell your software to use whatever storage you want it to use? It's almost inconceivable that it's so badly written that it forces all writes to go only to the boot drive.

Well the point of communicating in this forum was to ask for help and let me know what the experiences are of others. So heck, if you think its inconceivable, why not try out my setup and let me know how you would program this.

Things to note, its not made to be a "consumer level" application, its made to "engineer" with. If your looking at it from a user perspective, then that may be why its “inconceivable", try opening your mind more, not everything is set in stone or only what you see. Again, milage of understanding on what I am attempting to achieve may very, but again, I am happy to hear any ones opinion and feed back on the matter. ^^

A lot of the things you've been saying in this thread lead me to believe that you're in way over your head, have no idea what you're doing, and have been freaking out because you googled things, misunderstood them, and leapt to the wrong conclusions.

Way over my head? I had no idea you thought you where my supervisor rotfl, XD.

And Even if it was, i welcome it so i can find a solution, is that not what a good engineer does, find and develop solutions?

This is not over any ones head, whats the problem, kernel mode is not providing access to known developers, applications on macOS when running off of an external drive, even when setting it up to. How is that over my head?

And who's freaking out here? I am not the one who seems to be triggered by asking a question. I am not to do this?

Also I would echo @leman in saying that if you're training ML models on such a tremendously large data set that you need 2500 TB/month worth of writes, why on earth are you doing that kind of work on a Mac Mini?

To answer why i am doing this on a mac, TOPS performance, the macs "neural engine" that any one can access and use.

How did you ever plan to do all this on its internal drive?

I didn't, hence the problem i have stated many times i am having, how much clearer do i need to be?

Kernel access to when running macos on an external NVME. <-- still not clear?

the normal machine to "acquire" for this kind of work is (if you're only doing it once or twice) not to buy, but rent a big AWS instance for a month or two. One stuffed with a bunch of GPUs to accelerate model training. (This is one of the many reasons why I suspect you and/or your boss have no idea what you're doing.)

Normal for who? Perhaps you? Should we all follow only what you do?

You realise there are massive caveats to doing things only in the cloud? For one I am not a company looking to build a solution at scale, I am a student with limited funds. Rotfl!!

There are many many solutions to a problem, and renting tenants on a VPS instances can have hidden additional charges, navigating pricing on providers like AWS is often a nightmare, and if your doing this within a company, they usually have an accounts manager they work with. I am not a company.. Also more and more companies are seeing the inflationary costs with cloud, and moving back to premises equipment and private cloud as a result. Are you perhaps doing this with a company that has unlimited funds?

The point is to be off line, and maintain Intellectual property, which doing so on a VPS will not cater to.

On data privacy and security, regardless of what the provider tells you. have you read your contracts lately? There is an indemnity clause regarding the safeguarding of data, which often states (depending on provider) that they take no responsibility for your data in transit, and will try their best to protect it, often prompting for you to pay another service fee for backups..

Depending on your location, the recourse will be with a governing body (The office of the information commissioner in the UK), and they will not action unless a large number of people (often more then 1000 in the uk) are affected. So, the individual has a problem if they agree to this, and continue to use these 3rd party systems, rather then come up with a solution.

Also, why would i want to train the VPS providers AI (AWS in your case) with my data? Your response makes no sense!

Again apple have stated the m2 Mac mini pro supports known developer kernel access on macOS off of an external drive. So this should be working. I am still working with them to solve this, but at the same time was asking here if any one has experienced this issues.

I am starting to regret asking the public on a forum, as I seem to be triggering a few people who will not actually reflect on what I have to say with an open mind, refuse to answer my question, attack me with ad-hominems “ freaking out” and then state that their way is the only way….

 

[Kr0n05K!ngR]

The issue is hypothetical because the computer itself is more likely to die before modern SSDs. TBW is so high with modern SSDs that you're unlikely to ever hit that limit, even if doing a lot of heavy 4K video editing. For example, the Samsung 980 Pro series of SSDs have TBWs of 600x the listed capacity (e.g., 250GB - 150TBW, 500GB - 300TBW, 1TB - 600 TBW, 2TB - 1200 TBW), meaning you could fully rewrite the entire drive 600 times. Crucial's 2TB MX500 SSD has a TBW of 700TB, which is good for 350 complete rewrites of the drive. Enterprise SSDs, such as those used in datacenters and services such as AWS, Apple's own iCloud datacenters, etc. have even higher TBW ratings, and MTBF ratings in the millions of hours. The Microsoft article you cited is from 2019, which is an eternity in the tech sector.

The attached screenshot is from Crucial's specsheet for the MX500 series. Note that the MTTF is equivalent to 750,000 days of continuous usage.

View attachment 2269373

On a side note, there is no way an individual is writing 2500TB a month to any SSD.

Thanks for the input, but again how can you make a broad statement like "there is no way an individual is writing 25000tb a month on any ssd" without actually looking at the journal articles (peer reviewed scientific research) i provided.

Are you referring to your own experience like some? Or do you have access to everyones systems in the world to check? How can you make this claim with out actually understanding what is out there beyond what you and others in your group do with the hardware?

You provide a screen shot, how about an actual link to something beyond a picture you have control over?

Also, referring to the pick, it even sates 5 years, but you insist "the computer itself is more likely to die before modern SSDs."?

May be build your own computer, as it sounds like your getting ripped off. Every computer I have ever built has lasted way beyond 10 - 15 years. I am guessing but are you using laptops and portable devices that look cute, but are not very robust? As I know they do tend to produce allot of heat.

You can buy steel toe boots virtually anywhere, even at WalMart - they're often used in construction or related fields where heavy objetcs, tools, etc. are often carried on the job. They are also not even remotely similar to military issue boots used by the Army, Navy, Air Force, or Marines, which is why the analogy is valid. Whether or not someone has been in the military has absolutely no relevance to the analogy or discussion at hand.

Provide a brake down of how they are different, actually providing diagrams and research on the topic, other wise this is a baseless assertion and red herring.

Feel free to use the link I provided to keep you keep points and notes on the matter.

I think you are a) severely underestimating the durability of modern SSDs and b) even more severely overestimating how much data you will be reading/writing to any SSD. Consequently, a lot of your arguments are based on speculative inferences rather than measurable facts or observable behaviors.

Well i can only go by the research i have done, and i do not have the funds to test this on the m2 to confirm, feel free to do so on your dime! I will be happy to hear your findings.

 

[Kr0n05K!ngR]

You are making guesses about endurance of an SSD without having tested the said SSD or looking at the empirical data pertaining to the said SSD. That's your own admission. This makes what you say about endurance of Apple SSDs an uneducated guess.

You continue to avoid answering my questions, and just go round in a circle to the same conclusion which is a red herring at best, please answer my questions and stop avoiding them..


Because you have not done all things related to tech, or heard of all things related to this, you assume assume when some one is doing something perhaps out of your remit, that its fool hardy, unreal, or just uneducated?

I have already said many many times, the aim, run off of an external NVME SSD, there done.
Does apple allow this? Yes!

There for what i want is NORMAL according to apple.

As you have rightly stated, yes it will burn out in 3 - 6 months (under halt a year), so i would like to run it off of an external SSD, so that is why i need that good old "kernel access" which i do not have.

You know if your group of friends feel this is over my head, but yet provide no advice besides judgment, i do wonder what your intent is? Perhaps malicious?

As stated before, i noted my research and you keep going back to what seems "not with apple!, test it or do not ask for help!"

I love apple too but ALL SSDs have a TBW, and i want to keep my investment as long as possible. I AM A STUDENT and a minority, please understand that fact before talking from such a high place of privilege, not everyone can afford 7k systems or perhaps as your friend has stated the unlimited usage of an expensive VPS solution..

At any rate, I agree with you that these posts strike me as confused and incoherent. Especially the claims that this is typical use for machine learning. I'm not a machine learning expert, but I've been working in empirical research as data analysis specialist for two decades and I've never seen anything like that.

Even i am able to admit when i do not know something and ask for advice, but in your case, you say "I've been working in empirical research as data analysis specialist for two decades and I've never seen anything like that."

Which means you have worked with simple data bases, perhaps no ML and AI or any kind or re-enforcement learning. While AIs do use data bases, the data is pulls and written on the SSD at a much higher rate then a human can, and then some algorithms can. Again i am still learning this, but the SSD TBW was one thing my director told me to look for, as they have been known to provide too many writes, which is why often data centres will work on clusters of SSDs that will not write all to one drive at a time before move to another, but instead write to them each in a pattern (little bit for each drive) so to maximise the life of the SSD. This is comparable to how microSD cards can preserve data banks, writing little are a time to each bank. Now I cannot afford an array of NVME SSDs or a data centre, so i am working on a Proof of concept, meaning caveats i have mentioned and limited funds. I am still learning ML and AI, and happy to get advice. Open your mind beyond the silos you may be comfortable with.

Now, if you don't like that, its your problem not mine, i need help with he kernel issue, please focus on that if you wish to reply, rather then acting triggered with the things you have admitted "I'm not a machine learning expert", you have no experience of, relating it only to data analytics rather then ML or AI specifically, again they use Data analytics, but not at the legacy level you may be. Please stop trying to poison the well and my knowledge on the matter. I report to my university, not you or any one of your friends on this form.

Thank you!

 

[dmccloud]

Thanks for the input, but again how can you make a broad statement like "there is no way an individual is writing 25000tb a month on any ssd" without actually looking at the journal articles (peer reviewed scientific research) i provided.

No need to look at journal articles when simple math dictates that in order to write 2500TB in a month, one would need to write between 3.36TB (31 day month) and 3.72TB (28 day month) EVERY hour of the day without any breaks. Given that most consumer-grade SSDs have a TBW rating of up to 600x their storage capacity, that means you would need at a minimum a 5TB SSD just to have a margin of error. Even with a 5TB drive, you'd likely be looking at replacing it for too often to even be considered cost-effective. What you would really need for what you are claiming to write every month are enterprise-grade SSDs, such as those used for AWS.

Are you referring to your own experience like some? Or do you have access to everyones systems in the world to check? How can you make this claim with out actually understanding what is out there beyond what you and others in your group do with the hardware?

You provide a screen shot, how about an actual link to something beyond a picture you have control over?

You also lack "access to everyones systems in the world to check", so that entire argument is a red herring and deflection at its worst. At least I provided actual currently VERIFIABLE evidence whereas you passed off opinion, conjecture, and articles that are out of date as current. Sorry that conventional wisdom proves you wrong, but the math and facts don't add up in your claims.

Also, referring to the pick, it even sates 5 years, but you insist "the computer itself is more likely to die before modern SSDs."?

May be build your own computer, as it sounds like your getting ripped off. Every computer I have ever built has lasted way beyond 10 - 15 years. I am guessing but are you using laptops and portable devices that look cute, but are not very robust? As I know they do tend to produce allot of heat.

You continually make unfounded assumptions and cherry-pick things that fit into your predetermined beliefs while rejecting everything that disagrees with your preconceived notions. My Mac is a laptop, but my gaming PC is a custom-built machine. How many of those computers you have "built" were able to keep up with software advances after 10-15 years? Unless every single one of them have a Trusted Platform Module and meet every other arbitrary requirement from Microsoft, they can't run Windows 11 without some back-end hackery. Likewise, as software such as the Adobe Suite and games have evolved, so have the minimum hard requirements needed to run them with any semblance of performance. If you are upgrading ANY components in these 10-15 year old machines, then they really aren't lasting 10-15 years as originally configured.

Provide a brake down of how they are different, actually providing diagrams and research on the topic, other wise this is a baseless assertion and red herring.

Feel free to use the link I provided to keep you keep points and notes on the matter.



Well i can only go by the research i have done, and i do not have the funds to test this on the m2 to confirm, feel free to do so on your dime! I will be happy to hear your findings.

I'm not going to use an out of date link you provided for ANYTHING when advances in SSD technology since 2019 have led to noticeable improvements in reliability, stability, and overall performance. Just like your 10-15 year old PC, a 5 year old SSD will not be as reliable as current models.

I find it interesting that you admitted in your last sentence if it doesn't exist outside of your narrow realm of "research" (which honestly comes across as confirmation bias of things that support your personal misconceptions), then you discount it on its face without even a half-hearted attempt at critical analysis.

 

[dmccloud]

Now, if you don't like that, its your problem not mine, i need help with he kernel issue, please focus on that if you wish to reply, rather then acting triggered with the things you have admitted "I'm not a machine learning expert", you have no experience of, relating it only to data analytics rather then ML or AI specifically, again they use Data analytics, but not at the legacy level you may be. Please stop trying to poison the well and my knowledge on the matter. I report to my university, not you or any one of your friends on this form.

Thank you!

I would strongly advise you to get another advisor at your university if they are putting these misconceptions into your mind. By your own admission you are also not a machine learning expert but a student, and I would wager that both Leman and myself have done far more in the tech space and conducted far more research than you have even been exposed to, which makes your repeated arguments that anyone who doesn't know "everything" shouldn't be commenting even more inane.

 

[jinnyman]

There's no technical reason behind it. They just want to earn more money aka maximize their margin, and Apple, at the current position, can get away with it.

 

[Kr0n05K!ngR]

No need to look at journal articles when simple math dictates that in order to write 2500TB in a month, one would need to write between 3.36TB (31 day month) and 3.72TB (28 day month) EVERY hour of the day without any breaks. Given that most consumer-grade SSDs have a TBW rating of up to 600x their storage capacity, that means you would need at a minimum a 5TB SSD just to have a margin of error. Even with a 5TB drive, you'd likely be looking at replacing it for too often to even be considered cost-effective. What you would really need for what you are claiming to write every month are enterprise-grade SSDs, such as those used for AWS.

So confused by you answer, which makes me wonder if you like some are reacting rather then accepting you do not know everything, i am not claiming to, but your board statement's are a strange one.

Understand each use cases are different, and the point is to get speed and flow of data for the ML, thats not "normal" consumer level use by no means, but the fact remains, and the point is to extend the life of the mac i am using. So please stop trying to dissuade me (like others), from doing so..

Also, there are not always 31 days in a month, this is why you multiply by 52 weeks then divide by 12..

You also lack "access to everyones systems in the world to check", so that entire argument is a red herring and deflection at its worst. At least I provided actual currently VERIFIABLE evidence whereas you passed off opinion, conjecture, and articles that are out of date as current. Sorry that conventional wisdom proves you wrong, but the math and facts don't add up in your claims.

I am not the one making over arching statements like you have, so do not need access to "everyones systems" just a project aim, and what i wish to achieve and how to go about it. Thats what your arguing against why i am doing a think, rather then actually helping with the main issue, kernel access on macOS running on a NVME drive, which does not work for the reasons i have said many many times. Actually read history of my posts, rather then jump at the most recent..

You continually make unfounded assumptions and cherry-pick things that fit into your predetermined beliefs while rejecting everything that disagrees with your preconceived notions. My Mac is a laptop, but my gaming PC is a custom-built machine. How many of those computers you have "built" were able to keep up with software advances after 10-15 years? Unless every single one of them have a Trusted Platform Module and meet every other arbitrary requirement from Microsoft, they can't run Windows 11 without some back-end hackery. Likewise, as software such as the Adobe Suite and games have evolved, so have the minimum hard requirements needed to run them with any semblance of performance. If you are upgrading ANY components in these 10-15 year old machines, then they really aren't lasting 10-15 years as originally configured.

Cherry picking? What?

Ok here is your argument:

The issue is hypothetical because the computer itself is more likely to die before modern SSDs.

My response:

May be build your own computer, as it sounds like your getting ripped off. Every computer I have ever built has lasted way beyond 10 - 15 years. I am guessing but are you using laptops and portable devices that look cute, but are not very robust? As I know they do tend to produce allot of heat.

The point was death of a PC vs ssd, not weather or not it would run the latest and greatest of things..
Also Windows 11? Thats another privacy can of worms i will not touch, with or without a trusted platform module, not interested in serialisation of my data, so that i can be tracked and profiled thanks! GDPR is a thing here, and i intend to keep those rights regardless of the product or service on offer, the shinny does not tempt me.



You know steam can run on linux and there are fully library of things you can play on linux thanks to the steam deck. While its not practical as a pick up and use solution, does not mean your limited to what ever Microsoft scraps we get thrown. And last i check Microsoft put out a statement that we should all move in to the cloud, and stop buying hardware..

I'm not going to use an out of date link you provided for ANYTHING when advances in SSD technology since 2019 have led to noticeable improvements in reliability, stability, and overall performance. Just like your 10-15 year old PC, a 5 year old SSD will not be as reliable as current models.

I find it interesting that you admitted in your last sentence if it doesn't exist outside of your narrow realm of "research" (which honestly comes across as confirmation bias of things that support your personal misconceptions), then you discount it on its face without even a half-hearted attempt at critical analysis.

Hmm, i think your reacting, as this has nothing to do with my statement to you, it was regarding boots..

Your initial post:

You can buy steel toe boots virtually anywhere, even at WalMart - they're often used in construction or related fields where heavy objetcs, tools, etc. are often carried on the job. They are also not even remotely similar to military issue boots used by the Army, Navy, Air Force, or Marines, which is why the analogy is valid. Whether or not someone has been in the military has absolutely no relevance to the analogy or discussion at hand.

Hence:

Provide a brake down of how they are different, actually providing diagrams and research on the topic, other wise this is a baseless assertion and red herring.

Feel free to use the link I provided to keep you keep points and notes on the matter.

Try and read more of the post, and not react with assumptions that end up having nothing to do with the statement being made, or the reply being given..

 

[Kr0n05K!ngR]

I would strongly advise you to get another advisor at your university if they are putting these misconceptions into your mind. By your own admission you are also not a machine learning expert but a student, and I would wager that both Leman and myself have done far more in the tech space and conducted far more research than you have even been exposed to, which makes your repeated arguments that anyone who doesn't know "everything" shouldn't be commenting even more inane.

Yeah no, sorry but they have the power to grade me, your a talking head on the internet who seems to get triggered by words.

The point, need to achive is kernal access in macos when running on an external NVME on a mac mini m2 pro, is this achivable? Yes, it was on other mac minis up to the m2, so why not the m2? Apple have confirmed that kernal access not being avialable on macos running on external drive is unexpected behavior.

Regardless of how many writes you think i will or will not be performing, i am not willing to test my hardware to prove anything, as i do not have the time or money to do so, hence student.

Also, i never said i was a machine learning expert, but i did say that according to my research, which you have by your own admision:

I'm not going to use an out of date link you provided for ANYTHING when advances in SSD technology since 2019 have led to noticeable improvements in reliability, stability, and overall performance. Just like your 10-15 year old PC, a 5 year old SSD will not be as reliable as current models.

Your personal experiences are not all things to all people, no ones are. Nor are mine, i am simply working towards what i am told vs research done, i cannot simply quote: "some dude on internet said this", no director, teacher or whom ever marks my work will grade me well with it. I need more evidence beyond what you say. So again, provide actual evidence with links, not screen shots, and i will refer to that if the argument can go beyond the baseless assertion. It must be related to ML and AI, and it must solve the the issue of using a single SSD rather then an array of SSDs.

There is a process when writing a report: point, evidence explain. May be i am not doing a good job here, which is why i seem to have some up in arms..

If you wish to provide actual input that is helpful, rather then react and simply try to villainies me, please help with the kernel mode issue, do you have an M2 mac you can test, can you boot to an external drive and get kernel access (regarding to the gate keeper issue i mentioned).

Again thanks for the input, but i am getting tiered of trying to defend myself in order to get help..

Must the few villagers light their torches?

 

[Kr0n05K!ngR]

There's no technical reason behind it. They just want to earn more money aka maximize their margin, and Apple, at the current position, can get away with it.

Perhaps, business do need to make money after all, but i still feel apple are the best in terms of performance, style and robustness. As stated before, I LOVE APPLE!

While the SSDs may be allot stronger and robust compared to others on the market, i cannot prove this with out grate cost in time and money. (measuring the TBW myself).

Instead i can only quote the research that is out there for commodity SSD NVME, and then correlate this related to the mac mini.

 

[leman]

You continue to avoid answering my questions, and just go round in a circle to the same conclusion which is a red herring at best, please answer my questions and stop avoiding them..

I am not interested in your question. I am commenting on claims you are making. These forums are visited by many people, most of them looking for information. I happen to take an issue with some of the statements that you make, which is why I am rebutting them, so that people reading it get a more clear picture of the issues.

You know if your group of friends feel this is over my head, but yet provide no advice besides judgment, i do wonder what your intent is? Perhaps malicious?

My intent is to point out the contradictions and factual inaccuracies in your claims and statements in order to promote a healthier discussion on the forums.

Even i am able to admit when i do not know something and ask for advice, but in your case, you say "I've been working in empirical research as data analysis specialist for two decades and I've never seen anything like that."

Which means you have worked with simple data bases, perhaps no ML and AI or any kind or re-enforcement learning. While AIs do use data bases, the data is pulls and written on the SSD at a much higher rate then a human can, and then some algorithms can.

What you write here doesn't make any sense. You don't have to hammer your SSD to do reinforcement learning. Why not just keep the data in memory? I understand that you are learning, and there is a lot of confusion. But your approach has been criticised with people that have more credentials than you (myself included), maybe you can admit that there is at least some chance that your approach might be flawed?

Again i am still learning this, but the SSD TBW was one thing my director told me to look for

I'd recommend you to change your institution or your advisor. If they insist that hammering an SSD while doing low-level ML training routine is normal (because you can't be doing anything demanding on an M2 Pro, it's severely underpowered for training), I don't have much hope for them as an ML expert.

Please stop trying to poison the well and my knowledge on the matter. I report to my university, not you or any one of your friends on this form.

My goal is not to poison you, my goal is to help you do things better. If you university insists that this is how things are done, they are most likely very wrong. Of course, I could be misunderstanding the nature of your work, but the more you write the more I am convinced that the problem is low quality of training at your institution.

 

[Kr0n05K!ngR]

I am not interested in your question. I am commenting on claims you are making. These forums are visited by many people, most of them looking for information. I happen to take an issue with some of the statements that you make, which is why I am rebutting them, so that people reading it get a more clear picture of the issues.

So you can not prove how or why apple SSDs are more robust then commodity ones? You cannot prove what makes them different beyond the SSD controller being part of the SoC?

Right...

My intent is to point out the contradictions and factual inaccuracies in your claims and statements in order to promote a healthier discussion on the forums.

But you provide nothing beyond baseless assertions, and anecdotal evidence, not providing any proof for your argument, besides exaggerated some sense of misplaced authority.

How do i quote you?

What authoritative research have you done in the field, that qualifies you and your friends as “Experts”?

Remember, i am a student, not an expert, but i cannot simply quote " people on a forum told me". This is why on the scale of reliable research, forum posts are below mainstream new articles in terms of reliability.

If you can provide actually research links beyond you trying to tell me off, banging a drum, then i will use it.

Again, need actual help, hence my initial posts. If you cannot offer that, then what is the point of you replying to my post?

What you write here doesn't make any sense. You don't have to hammer your SSD to do reinforcement learning. Why not just keep the data in memory? I understand that you are learning, and there is a lot of confusion. But your approach has been criticised with people that have more credentials than you (myself included), maybe you can admit that there is at least some chance that your approach might be flawed?

Critiqued, and what credentials? Again how do i quote you? What authoritative research have you done in the field?

Perhaps offer some of your own research on the matter, guidance, help me learn the right way providing evidence beyond the anecdotal? Do not simply try and villainies me because you do not agree. I would say your bullying rather then offering Critique, as Critique would offer alternatives and guidance, your small group seem to only offer vitriol.

I'd recommend you to change your institution or your advisor. If they insist that hammering an SSD while doing low-level ML training routine is normal (because you can't be doing anything demanding on an M2 Pro, it's severely underpowered for training), I don't have much hope for them as an ML expert.

Again silly statements like this which make me not want to take you seriously..

How does one change institutions based on a post on a forum, from a person who has done no professional authoritative research in the field?...

If your so convinced of this, will you fund my change? Perhaps support this economically, can you put your money where your mouth is? I doubt it… your not even willing to answer my question or test with M2..

Help me understand this reactionary statements your making..

Your an advanced user at best, who may have a job in tech, but has done nothing beyond working for which ever company you do, thats does not make you a professional, that makes you a skilled worker. This is why skilled works rhetoric is not used by anyone beyond their bubble (supervisor, anyone working within their department)..

Your authority may look good on your resume, but that does not mean institutions (RFC, IETF, IET, NIST, ETC) will take your seriously..

Again there are levels to research and forum posts are below mainstream news articles in reliability. ( you get deducted marks for using them..) Provide actual evidence, and links, looking at my history of posts, and I will review and respect your opinion more.

My goal is not to poison you, my goal is to help you do things better. If you university insists that this is how things are done, they are most likely very wrong. Of course, I could be misunderstanding the nature of your work, but the more you write the more I am convinced that the problem is low quality of training at your institution.

Well this is a contradiction, you say you want to help, but then just pass judgment rather then provide links to research, or other ways of doing things, perhaps tutorials? Somthing you have tried and has worked for you (meaning actualy links to knowledge bases, or raeding materials, not just statments..)?

Nope judgment only. Sorry its not helpful, its spiteful!

 

[NT1440]

So you can not prove how or why apple SSDs are more robust then commodity ones? You cannot prove what makes them different beyond the SSD controller being part of the SoC?

Right...


But you provide nothing beyond baseless assertions, and anecdotal evidence, not providing any proof for your argument, besides exaggerated some sense of misplaced authority.

How do i quote you?

What authoritative research have you done in the field, that qualifies you and your friends as “Experts”?

Remember, i am a student, not an expert, but i cannot simply quote " people on a forum told me". This is why on the scale of reliable research, forum posts are below mainstream new articles in terms of reliability.

If you can provide actually research links beyond you trying to tell me off, banging a drum, then i will use it.

Again, need actual help, hence my initial posts. If you cannot offer that, then what is the point of you replying to my post?



Critiqued, and what credentials? Again how do i quote you? What authoritative research have you done in the field?

Perhaps offer some of your own research on the matter, guidance, help me learn the right way providing evidence beyond the anecdotal? Do not simply try and villainies me because you do not agree. I would say your bullying rather then offering Critique, as Critique would offer alternatives and guidance, your small group seem to only offer virtual.



Again silly statements like this which make me not want to take you seriously..

How does one change institutions based on a post on a forum, from a person who has done no professional authoritative research in the field?...

If your so convinced of this, will you fund my change? Perhaps support this economically, can you put your money where your mouth is? I doubt it… your not even willing to answer my question or test with M2..

Help me understand this reactionary statements your making..

Your an advanced user at best, who may have a job in tech, but has done nothing beyond working for which ever company you do, thats does not make you a professional, that makes you a skilled worker. This is why skilled works rhetoric is not used by anyone beyond their bubble (supervisor, anyone working within their department)..

Your authority may look good on your resume, but that does not mean institutions (RFC, IETF, IET, NIST, ETC) will take your seriously..

Again there are levels to research and forum posts are below mainstream news articles in reliability. ( you get deducted marks for using them..) Provide actual evidence, and links, looking at my history of posts, and I will review and respect your opinion more.



Well this is a contradiction, you say you want to help, but then just pass judgment rather then provide links to research, or other ways of doing things, perhaps tutorials?

Nope judgment only. Sorry its not helpful, its spiteful!

What was the point of this post?

It’s clear your (unfounded, as demonstrated previously with actual data you’ve ignored) using the wrong tool for the job (in the context of your concerns), why not use the right configuration of product you actually need?

You’re going to extreme lengths to justify what in 2023 is basically the modern equivalent of an old wive’s tale.

 

[Kr0n05K!ngR]

What was the point of this post?

It’s clear your (unfounded, as demonstrated previously with actual data you’ve ignored) using the wrong tool for the job (in the context of your concerns), why not use the right configuration of product you actually need?

You’re going to extreme lengths to justify what in 2023 is basically the modern equivalent of an old wive’s tale.

Did you read the whole post? Should be apparent what the point was..

I have only seen baseless assertions, provide the links you mention in relation to the user i am replying to.. What research articles have they used? What evidence have they provided beyond the anecdotal?

Or can you only provide baseless assertions and anecdotal evidence also?

Remember i cannot quote "some guy on a forum told me".

Want to help, help, not throw insults or judgment.

Thanks..

 

[leman]

So you can not prove how or why apple SSDs are more robust then commodity ones? You cannot prove what makes them different beyond the SSD controller being part of the SoC?

Of course I cannot. Nor do I have to. I am not the one claiming that Apple SSDs are unsuitable for some basic ML training use.

How do i quote you?

What authoritative research have you done in the field, that qualifies you and your friends as “Experts”?

What do you want to quote? You want a paper that shows that you don't have to destroy SSDs to do reinforcement learning? Seriously? What else? Do you also want a paper citation as proof that a screwdriver is not good for hammering in nails? Or that a chilli pepper will burn if you put it in your eye?

Well this is a contradiction, you say you want to help, but then just pass judgment rather then provide links to research, or other ways of doing things, perhaps tutorials? Somthing you have tried and has worked for you?

I did give you a suggestion: stop writing your intermediate results to the disk. If your pipeline is set up in a way that trainings steps are written to disk, and you are unwilling to rewrite your code, use a RAM disk. If your model is too large to fit in memory, well, then you are using a wrong tool. There are plenty of commercial cloud solutions for ML learning that will likely be cheaper than buying new SSDs.

 

[Kr0n05K!ngR]

Of course I cannot. Nor do I have to. I am not the one claiming that Apple SSDs are unsuitable for some basic ML training use.

Provide evidance they are, as stated my reserch is comparable to commodity hardware related to SSDs, there is nothing to prove apple SSDs are beyond that.

What do you want to quote? You want a paper that shows that you don't have to destroy SSDs to do reinforcement learning? Seriously? What else? Do you also want a paper citation as proof that a screwdriver is not good for hammering in nails? Or that a chilli pepper will burn if you put it in your eye?

Provide evidence based on factual research sure, not specifically on that they don't destroy SSDs but rather a journal that corroborating your assertions, this is called an evidence based argument after all.

Screwdrivers and chilli peppers have nothing to do with this, stop trying to divert from the topic, stick to proving your point, i have provided evidence that ML most definitely does push SSD to reach TBW sooner, its the implementation of the SSDs (in an array and how they are written) is what increases its life. So, provide research beyond your anecdotal evidence that refutes this.

I did give you a suggestion: stop writing your intermediate results to the disk. If your pipeline is set up in a way that trainings steps are written to disk, and you are unwilling to rewrite your code, use a RAM disk. If your model is too large to fit in memory, well, then you are using a wrong tool. There are plenty of commercial cloud solutions for ML learning that will likely be cheaper than buying new SSDs.

Ok this is some what helpful, but the problem i have ultimately is getting the writes elsewhere when the ML is happening. I can use RAM, but with 16gb, it will fill very fast. So the goal again is to get kernel access to an external SSD, so that i can boot to it, and do everything from there. That way i avoid reaching TBW sooner on the internal drive.

Cannot use cloud, as i would be training the cloud providers AI with data that is technically some one else Intellectual Property, i must use private cloud on bare metal hardware.

The mac m2 pro mini seemed like a good choice related to power vs performance. I was considering Jetton nano, but they have other problems and cost more for the performance they offer. The Mac offers allot more TOPS for the money.

The M2 can provide 15.8 trillion operations per second, or 15.8 TOPS.

Ref:-https://www.apple.com/uk/newsroom/2023/01/apple-unveils-m2-pro-and-m2-max-next-generation-chips-for-next-level-workflows/

Ref:- https://www.notebookcheck.net/Apple-M2-Processor-Benchmarks-and-Specs.632312.0.html

 

[dmccloud]

So confused by you answer, which makes me wonder if you like some are reacting rather then accepting you do not know everything, i am not claiming to, but your board statement's are a strange one.

Understand each use cases are different, and the point is to get speed and flow of data for the ML, thats not "normal" consumer level use by no means, but the fact remains, and the point is to extend the life of the mac i am using. So please stop trying to dissuade me (like others), from doing so..

Also, there are not always 31 days in a month, this is why you multiply by 52 weeks then divide by 12..

I clearly said there were NOT always 31 days in a month:

No need to look at journal articles when simple math dictates that in order to write 2500TB in a month, one would need to write between 3.36TB (31 day month) and 3.72TB (28 day month) EVERY hour of the day without any breaks.

How can one take you seriously when you deliberately misconstrue the arguments of others? You said 2500TB in a month, so I figured out how many TB/hour that would be in a 28day month (February) and a 31 day month, which is where that 3.36TB to 3.72TB number comes into play. This "multiply by 52 weeks then divide by 12" doesn't even make sense in this scenario.

I am not the one making over arching statements like you have, so do not need access to "everyones systems" just a project aim, and what i wish to achieve and how to go about it. Thats what your arguing against why i am doing a think, rather then actually helping with the main issue, kernel access on macOS running on a NVME drive, which does not work for the reasons i have said many many times. Actually read history of my posts, rather then jump at the most recent..

You are the one making sweeping generalizations about SSD lifespans then attempting to change the discussion to a tangential issue. Nobody has to have access to "everyones systems" for purposes of this discussion, and since you clearly do not have such access, the last thing you should be doing is to place undue burdens on others you yourself are unable and/or unwilling to apply to your own claims.

Cherry picking? What?

Ok here is your argument:



My response:



The point was death of a PC vs ssd, not weather or not it would run the latest and greatest of things..
Also Windows 11? Thats another privacy can of worms i will not touch, with or without a trusted platform module, not interested in serialisation of my data, so that i can be tracked and profiled thanks! GDPR is a thing here, and i intend to keep those rights regardless of the product or service on offer, the shinny does not tempt me.



You know steam can run on linux and there are fully library of things you can play on linux thanks to the steam deck. While its not practical as a pick up and use solution, does not mean your limited to what ever Microsoft scraps we get thrown. And last i check Microsoft put out a statement that we should all move in to the cloud, and stop buying hardware..

Other than again ignoring the arguments of others which contridict your preconceived notions, you keep dropping nonsequitur after nonsequitur here and going off on tangents that are more about your personal biases than anything relevant to the initial discussion.

Hmm, i think your reacting, as this has nothing to do with my statement to you, it was regarding boots..

Your initial post:



Hence:



Try and read more of the post, and not react with assumptions that end up having nothing to do with the statement being made, or the reply being given..

I have read your posts, and your posts say things that you either can not back up or will not back up and attempt to dodge the questions by gaslighting everyone.

 

[dmccloud]

Provide evidance they are, as stated my reserch is comparable to commodity hardware related to SSDs, there is nothing to prove apple SSDs are beyond that.



Provide evidence based on factual research sure, not specifically on that they don't destroy SSDs but rather a journal that corroborating your assertions, this is called an evidence based argument after all.

Screwdrivers and chilli peppers have nothing to do with this, stop trying to divert from the topic, stick to proving your point, i have provided evidence that ML most definitely does push SSD to reach TBW sooner, its the implementation of the SSDs (in an array and how they are written) is what increases its life. So, provide research beyond your anecdotal evidence that refutes this.



Ok this is some what helpful, but the problem i have ultimately is getting the writes elsewhere when the ML is happening. I can use RAM, but with 16gb, it will fill very fast. So the goal again is to get kernel access to an external SSD, so that i can boot to it, and do everything from there. That way i avoid reaching TBW sooner on the internal drive.

Cannot use cloud, as i would be training the cloud providers AI with data that is technically some one else Intellectual Property, i must use private cloud on bare metal hardware.

The mac m2 pro mini seemed like a good choice related to power vs performance. I was considering Jetton nano, but they have other problems and cost more for the performance they offer. The Mac offers allot more TOPS for the money.

The M2 can provide 15.8 trillion operations per second, or 15.8 TOPS.

Ref:-https://www.apple.com/uk/newsroom/2023/01/apple-unveils-m2-pro-and-m2-max-next-generation-chips-for-next-level-workflows/

Ref:- https://www.notebookcheck.net/Apple-M2-Processor-Benchmarks-and-Specs.632312.0.html

You can't demand others prove evidence for every potential situation when you yourself have provided NONE. Even when people have provided links to other websites, you have dismissed them on their face without even the slightest attempt at critical analysis. The irony of your calling for "evidence based argument" when your own posts are full of speculation and reliance on out of date research which has been supplanted by newer technologies in the SSD space is laughable at best.

 

[Kr0n05K!ngR]

You can't demand others prove evidence for every potential situation when you yourself have provided NONE. Even when people have provided links to other websites, you have dismissed them on their face without even the slightest attempt at critical analysis. The irony of your calling for "evidence based argument" when your own posts are full of speculation and reliance on out of date research which has been supplanted by newer technologies in the SSD space is laughable at best.

Evidence for every anecdotal statement, your baseless assertions mean nothing if you cannot back it up with evidence. As stated before, I cannot quote in an report “some taking head on a forum told me”. Again, you opinion is respected, but also rejected, i will not take advice that cannot be backed up just on "trust me bro" logic.

I have provided links, and rational based on what i am doing why i am doing it. I need help with kernel access when running macOS on external NVME, regardless of what you "think", thats the problem i need to overcome. And the main reason i posted..

 

[Kr0n05K!ngR]

I clearly said there were NOT always 31 days in a month:



How can one take you seriously when you deliberately misconstrue the arguments of others? You said 2500TB in a month, so I figured out how many TB/hour that would be in a 28day month (February) and a 31 day month, which is where that 3.36TB to 3.72TB number comes into play. This "multiply by 52 weeks then divide by 12" doesn't even make sense in this scenario.



You are the one making sweeping generalizations about SSD lifespans then attempting to change the discussion to a tangential issue. Nobody has to have access to "everyones systems" for purposes of this discussion, and since you clearly do not have such access, the last thing you should be doing is to place undue burdens on others you yourself are unable and/or unwilling to apply to your own claims.



Other than again ignoring the arguments of others which contridict your preconceived notions, you keep dropping nonsequitur after nonsequitur here and going off on tangents that are more about your personal biases than anything relevant to the initial discussion.



I have read your posts, and your posts say things that you either can not back up or will not back up and attempt to dodge the questions by gaslighting everyone.

Not even going to reply to each, as you are deleting your statements when i put in yours vs mine, so to provide rational, your purposefully attempting to hide what you said, which is why it is not their, and your continue to give red herrings to divert the topic away from you being unable to prove your point. Example regarding the boot comparison..

If you cannot provide assistance or evidence, please stop replying. Your not going to change my mind, and my focus is to find the solution i need.

Please help providing advice or alternatives, no ad-hominems, no tu quoque fallaciess, no red herrings, focus on the issue please. Or simply get on with your life, as i should not be the main focus, i am sure you have better things to do..

 

[NT1440]

The solution here is to use the right tool for the job.

Though as others with real world experience have pointed out, your workflow is nonsensical as somehow you’re looking to write data center amounts of data. That alone should warrant a rethink if *how* your workflow is set up because something definitely isn’t right.

 

[Kr0n05K!ngR]

The solution here is to use the right tool for the job.

Though as others with real world experience have pointed out, your workflow is nonsensical as somehow you’re looking to write data center amounts of data. That alone should warrant a rethink if *how* your workflow is set up because something definitely isn’t right.

This obviously seems to triggers you in some way, i say again, focus on if you can help regarding kernel access on an external SSD running macOS. The other element to it is irrelevant. I cannot quote you guys, and i cannot restart, i can only work with what i have and make the best of it.

If you think its "nonsensical" then i am sorry you do not understand, or wish to try, you have no evidence that i can study, no links to journals, and refuse to believe that data flow for Ai and ML on a single SSD drive will produce that much data, just because you have not seen it..

Also "real world experience" what does that even mean?

Are you and the others professionals in the field? Or are you advanced users who work for the company, and should simply be followed just because you just say it?

Where is your research related to the topic?

What papers have been published? Are you chartered engineers? Which professional body do you guys belong to?

Your opinions while respected, are not authoritative and I cannot trust them with out real research on the topic (related to SSDs).

I cannot quote in my research " a forum said".. What do you not understand?

I cannot use "trust me bro" logic.

Until you and others in your small group can provide actual links, methods, and give real advice beyond vitriol, your responses are useless to me.

 

[thebart]

And that's why you simply get an external ssd...

Which is going to be slower. And another wire and another box hanging off your computer, ruining that nice compact aesthetic form. And they don't give you enough ports so add a hub, yet another box

One more instance of Apple making things worse for their customers and ruining their own design, through sheer stinginess and greed

 

[ovbacon]

Which is going to be slower. And another wire and another box hanging off your computer, ruining that nice compact aesthetic form. And they don't give you enough ports so add a hub, yet another box

One more instance of Apple making things worse for their customers and ruining their own design, through sheer stinginess and greed

My external ssd tests in Blackmagic on my mini M2: steady at W:2800+ MB/s, R:2700+/- MB/s which seems pretty fast to me and is definitely faster than 99% of people need.

 

[mr_roboto]

Background:

Kernel mode access: "drivers that run in the same memory space as the operating system kernel, which is the core component of the system that manages resources, processes, and security. This means that kernel mode drivers have direct access to the hardware and can perform faster and more efficiently than user mode drivers."

LOL. You do not need to lecture me on the basics, I'm an engineer who's been in the industry for decades.

The reason I asked you to clarify is because you started by claiming that Apple is doing a terrible job of supporting people who want to experiment with ML, then began babbling about unrelated topics like "kernel access" in a manner which suggested you were introducing them in support of the original complaint.

Anything that is not run within the kernel, is sandboxed in a different memory space, and apple use programs such as
"gatekeeper" to manage this.

Gatekeeper does not manage memory spaces or sandboxing, its function is to check code signatures to determine whether a binary was supplied by a trusted source without alteration.

Also your generalising machine learning, as some need drivers ran in the memory space of the kernel.

Then name these things! What ML training software, when running on macOS, particularly Apple Silicon macOS, needs installation of a special driver? Apple provides APIs for using all the relevant hardware, there's no need for a third party kext.

This can also be achieved installed on a virtual machine that has kernel access within its own elastic environment, which is the case for your VPS i believe. But this is software or would need a redirect to hardware resource, which some GPU manufactures provide API access for, which can require an additional licence. With bare metal, in no VM i need local kernel mode access. This can give better access the computer hardware, such as the neural engine.

Nonsense. You don't need VMs to train a neural network, and outside of VMs, you don't need "bare metal" access to the Apple Neural Engine (ANE). You don't even want that, unless your actual project is to reverse engineer the ANE - Apple does not document its instruction set, or any of the other low level details you'd need to use it without letting macOS run the show.

Explain whats wired about wanting to be in an environment that allowed for replacement of an SSD when it reaches TBW? Once it is dead in the mac mini (soldered on) its dead.
How is this strange to you?

That is not strange. Your statements before were. You are not communicating clearly, nor is it clear whether you're trying to.

Well the point of communicating in this forum was to ask for help and let me know what the experiences are of others. So heck, if you think its inconceivable, why not try out my setup and let me know how you would program this.

How can anyone even know what your "setup" is? You've never been specific about it, instead you like to go off on irrelevant digressions about VMs, bare metal, kernel mode access, veracrypt, and so forth.

Things to note, its not made to be a "consumer level" application, its made to "engineer" with. If your looking at it from a user perspective, then that may be why its “inconceivable", try opening your mind more, not everything is set in stone or only what you see. Again, milage of understanding on what I am attempting to achieve may very, but again, I am happy to hear any ones opinion and feed back on the matter. ^^

My opinion is that if you're sincere, you really need to pay more attention in class.

To answer why i am doing this on a mac, TOPS performance, the macs "neural engine" that any one can access and use.

People usually don't want to use the Apple Neural Engine for training, as it's too specialized for inference. Apple's own documentation suggests training on the GPU.

That said, if you really want to try to target the ANE for training, it's right there, you can use it. Go wild! But please don't pretend it's impossible just because of countless made-up issues which don't really matter.

I didn't, hence the problem i have stated many times i am having, how much clearer do i need to be?

Kernel access to when running macos on an external NVME. <-- still not clear?

No, because you're ignoring one of the most important questions, one which makes it completely unclear why you're complaining about this:

Why do you need macOS to boot from the external NVME???

Many non-technical users boot Macs from an internal drive, then work with data stored on external drives. You're claiming to be an expert, yet you don't know how to do this very basic thing? Are you sure you're in the right field of study? Is this really the position you want to take? Please clarify.

You realise there are massive caveats to doing things only in the cloud? For one I am not a company looking to build a solution at scale, I am a student with limited funds. Rotfl!!

I was assuming you were being honest about needing to write 2500 TB per month. That's why I thought you were talking about your manager, not your student adviser.

Toy undergrad projects don't need 2500 TBW/month; they should be designed to teach you how to use the tools rather than require expensive hardware. I can believe there's graduate level projects which might legitimately need that, but they'd come with enough grant money to rent or buy appropriate hardware.

Also, why would i want to train the VPS providers AI (AWS in your case) with my data? Your response makes no sense!

You want to come at me about not making sense when you say things like this?!

AWS rents you a computer (or a slice of a computer), not an "AI". If you use your rented compute power to train a neural network, you are not training the "VPS provider's AI". You're simply generating data stored in disk space that you're also renting from AWS. AWS doesn't know or care what that data is (they're not supposed to look at it), nor do they own any IP rights to it. If Amazon tried to write their contracts such that they owned all data generated on or uploaded to AWS, nobody would use their services.

Again apple have stated the m2 Mac mini pro supports known developer kernel access on macOS off of an external drive. So this should be working. I am still working with them to solve this, but at the same time was asking here if any one has experienced this issues.

I'm guessing that even if you're not actually a troll, whatever support person has the misfortune of talking to you is as confused as we are.

I am starting to regret asking the public on a forum, as I seem to be triggering a few people who will not actually reflect on what I have to say with an open mind, refuse to answer my question, attack me with ad-hominems “ freaking out” and then state that their way is the only way….

It's not that my way is the only way. It's that if you're not trolling, you simply don't need most of the things you think you need.