“Starting at” is the Biggest Lie in Tech

[SorryBella]

8 hours ago, JordB said:

it’s a lie

Lies are a form of love, you know?

 

[emcue]

Wow, this was an astounding 10/10 video. Great dive into a great (ewaste, device longevity, right to repair) topic that's always on the surface in LTT.

[jfred]

1 hour ago, hishnash said:

Form how LTT have talked about right to repair they in general mostly seem to view it as cheap upgrades not repairing a product to bring a dead product back to life. 

Modularity, like repair documentation and part availability, also extends the useful life of a product. And ultimately the reason many of us advocate for right-to-repair is to cut down on e-waste and keep useful devices out of landfills. Being able to pull the motherboard out of your Framework and make it your next media center PC as opposed to buying something completely new is a perfect example of this.

 

It also makes it exceedingly more likely that end users will actually repair something rather than throwing it away. Yes, SMD rework is doable if you're experienced, but realistically most people would need to find a repair shop, and shops that can actually do work that involved are less common than they should be. Meanwhile, anyone handy with a screwdriver can replace a bad NVMe SSD or wifi card in their laptop. (And diagnosing and fixing that kind of thing is exactly the sort of stuff I remember doing as a teenager.)

[papajo]

7 hours ago, dogwitch said:

isp do this now.

charging for wifi.

that already built into the cost of the router/modem combo.

In my country(Greece)  if an ISP did that they would burn their headquarters to the ground.. 

 

On the other side they exploit us by not having competition because its a small country so its basically 1 "A tier" (cant remember the right term) ISP and the rest are a couple of satellite companies leeching of the "A tier" 's  infrastructure so speeds are low because of a very slow upgrade rate and prices are relatively high (compared to western rich countries) and batshit crazy (compared to countries of the same size which are not rich e.g Bulgaria) 

E.g a 50mbps connection (which is most likely the max you can get in your area only like less than 5% has access for 100-200 mbps and even less 1gbps <- oh and the upload speeds are horrendous like 10% of the download speed)  would cost 35 euro monthly (but with a "free" wifi router which is cheap huawei but it does its basic job to "transfer" the internet via lan or wifi without releasing magic smoke at least for a year or so lol) 

But I am pretty sure "my people" would put the line on charging extra for wifi lol. 


 

9 hours ago, JordB said:

It’s easy to look at the ‘starting at’ price for a new laptop and think it’s a great deal. But when you look closer, it’s a lie. That entry level config is barely good enough for what you need to do today, let alone tomorrow, and you can’t even upgrade the thing after you buy it!

Nice to see linus doing the exception and speak straight about how horrible the PC market has gotten (although by having an ulterior motive since he has interest on modular laptops which compete directly the SoC chromebooks and entry/mid grade laptops so its not a coincidence he "remembered" how shitty the market is but I am an unbiased reasonable man, I have nothing against linus' bussiness flourishing  so  if I see a good message no matter why it got out I like and share the video ! ) 

The even enraging thing is that by having everything unupgradable  by having a SoC with soldered ram and storage etc in a "monolithic" tiny pcb the production cost is actually CHEAPER but their "base" models dont defer from the price of their base models of the old which had separate pcbs slots etc.. so they not only making more money e.g by "selling future theoretical needs for storage today with today's prices" as linus well said but they make more money by having a much simpler and faster production line where everything is on a tiny pcb and the customer see none of these cost saving but rather pays even more!!!! 

 

[hishnash]

16 minutes ago, jfred said:

Yes, SMD rework is doable if you're experienced, but realistically most people would need to find a repair shop, and shops that can actually do work that involved are less common than they should be

Well they should be more common, if you look back over history not everyone was expected to do the blacksmiths job but every town had a blacksmith.   I do not see the issue with paying an expert to repair a product rather than throw it away. 

 

21 minutes ago, jfred said:

Meanwhile, anyone handy with a screwdriver can replace a bad NVMe SSD or wifi card in their laptop.

And in both of those cases they are throwing away working HW, yes the NAND might be dead on the NVM drive but the Controler is still perfectly working, or the a capacitor might be blown on your wifi card but the radio works fine.     It can be repaired, and HW vendors should provide schematics and firmware flashing tools to enabled this. 

[papajo]

2 minutes ago, hishnash said:

Well they should be more common, if you look back over history not everyone was expected to do the blacksmiths job but every town had a blacksmith.   I do not see the issue with paying an expert to repair a product rather than throw it away. 

 

Theoretically its possible to upgrade your non upgradable GPU by buying the cheapest GPU laptop combo and then solder a GPU (from the same SKU usually) you harvest from e.g broken more expensive laptop...  but how many people did this? maybe 1 in the entire world lol 

So yea if something is theoretically possible doesnt mean it will or realistically can happen. 

[hishnash]

1 minute ago, papajo said:

so they not only making more money

Not quite that simple.   The cost of having modular memory (in $, power and other factors) is much higher as you increase mem bandwidth and generation over generation that is what is happening.  So given a fixed budted of $ and power and space if you want to improve bandwidth (needed to support better compute) you need to go this direction or increase device cost, power draw and size.   

With the M2 for example you have 100GB/s of bandwidth to do this with LPDDR5 sodiums you would need 2 separate dims. That is a good amount of space within the device and it would draw massively more power.  Then if you consider the M2 Max that would require 8 dim slots!  

[hishnash]

3 minutes ago, papajo said:

Theoretically its possible to upgrade your non upgradable GPU by buying the cheapest GPU laptop combo and then solder a GPU (from the same SKU usually) you harvest from e.g broken more expensive laptop...  but how many people did this? maybe 1 in the entire world lol 

So yea if something is theoretically possible doesnt mean it will or realistically can happen. 

Again you are talking upgrade, not repair here, upgrade is a feature, repair is a right.    If right to repair were supported by HW vendors that does not alter the upgrade options but would mean if your GPU dies it would be easy for any vendor out there to buy the GPU chip package and replace your dead chip.    

People doing upgrades would not be intreated in doing this as the cost of such a part will not be that low, most of you GPUs cost is in the GPU so replacing a working GPU chip (that has most of your cards $$$ in it) with another one would not be economic unless you can sell your older one... this would not be worth the time and effort but it would be worth to repair a broken one. 

[Gokul_P]

We all know how texas instruments and nvidia has a better profit margin than apple. Here most of the times there is no whitelisted calculator list so students dont have to buy the expensive one. and most exams dont allow to use calculator here and if it is a online test (computerized) you will get a onscreen calculator. 

[hishnash]

2 minutes ago, Gokul_P said:

Here most of the times there is no whitelisted calculator list so students dont have to buy the expensive one.

On the consumer facing side we all think of TI as the calculator making but they do a LOT more than that, there is a very high chance that if you pick up any bit of tec there will be some chip within it that TI made or own patents for and license it out.  They are one of the most prolific vendors making a massive select of chips this is were they make the money not calculators. 

[Blademaster91]

2 hours ago, hishnash said:

So the thing is right to repair is all about the "repair" part not the upgrade part.   using solder is not at all anti repair infact is very pro repair, its not glue, its also much easier and cheaper to fix than if a socket gets damaged (water damage), solder you can desolder it and then clean up the pads - re-ball and your good to go, with a LGA or PGA socket its going to cost you a LOT of time and effort to repair.

That isn't the definition of right to repair at all, and is what is pretty much what companies like Apple do now, you have to have a certified repair shop with specialized tools fix the laptop or phone. It isn't realistic for the consumer to spend thousands on a soldering station and SMD rework tools. And it doesn't make any sense to solder everything down when RAM or an SSD could be easily swapped out by the user or any repair shop if the owner doesn't want to do it themselves. IMO it should be the same as vehicles in terms of right to repair, standardized parts the consumer can buy that aren't locked down by the manufacturer.

29 minutes ago, hishnash said:

And in both of those cases they are throwing away working HW, yes the NAND might be dead on the NVM drive but the Controler is still perfectly working, or the a capacitor might be blown on your wifi card but the radio works fine.     It can be repaired, and HW vendors should provide schematics and firmware flashing tools to enabled this. 

I think hardware manufacturers should provide schematics, and replacing a capacitor on a wifi card is easy enough to do, but

If the SSD or RAM fails then you recycle it, it isn't economically viable for a company like Samsung or Micron to replace a dead NAND chip on an RMA.  Having to throw away a stick of RAM or an SSD is less wasteful than someone throwing a whole laptop away because the SSD is soldered in making the cost of repair and labor more than the value of the laptop.

[papajo]

50 minutes ago, hishnash said:

Again you are talking upgrade, not repair here, upgrade is a feature, repair is a right.    If right to repair were supported by HW vendors that does not alter the upgrade options but would mean if your GPU dies it would be easy for any vendor out there to buy the GPU chip package and replace your dead chip.    

People doing upgrades would not be intreated in doing this as the cost of such a part will not be that low, most of you GPUs cost is in the GPU so replacing a working GPU chip (that has most of your cards $$$ in it) with another one would not be economic unless you can sell your older one... this would not be worth the time and effort but it would be worth to repair a broken one. 

Yea sorry I didnt read all you messages before so I thought you were saying that repairing/messing with your hardware by doing SMD stuff on it is easy or at least should become common because its possible to do so. 

Now I see that you say we shouldnt "mix" the right to repair with the right to upgrade... well those go hand to hand repair/upgrade its essentially the right of being able to do WTF you want with the  you paid with your hard earned money for. 

Its only that Rossman and co (who are likeable folk have nothing against them) are into the repair business and make money from that and thus focus on that. 

And I am not against them making more money because their profit coincides with my (the consumer) profit, but dismissing my other right (as the one to upgrade ) so that I increase/enhance their debate about wide spreading the right to repair (= also their right to make money from it) is like shooting my right foot so that the dog that wants to bite my left foot gets scared away.. ok my left foot isnt gonna get bitten by the dog (e.g apple etc who dont want right to repair) but my right foot has now a hole in it... lol 

So taking only one issue (right to repair) and dismissing the other (right to upgrade) will only make people that fix stuff like Rossman richer but hurt me the consumer 

Governments should realize that consumers should be able BOTH to repair AND upgrade (AND mod in general actually ) our hardware that we bought with our hard earned money. 

[papajo]

54 minutes ago, hishnash said:

Not quite that simple.   The cost of having modular memory (in $, power and other factors) is much higher as you increase mem bandwidth and generation over generation that is what is happening.  So given a fixed budted of $ and power and space if you want to improve bandwidth (needed to support better compute) you need to go this direction or increase device cost, power draw and size.   

With the M2 for example you have 100GB/s of bandwidth to do this with LPDDR5 sodiums you would need 2 separate dims. That is a good amount of space within the device and it would draw massively more power.  Then if you consider the M2 Max that would require 8 dim slots!  

That's irrelevant I did not say that the sole reason they have soldered memory and storage is to make more money obviously it has technicall benefits as well

I said that THEY SAVE MONEY by doing so (what made them do it is irrelevant as far as this argument is concerned) and that the money saved does not "trickle-down" to the consumer which is another proof that "trickle-down economics" propagated by the right wing politicians is BS that never gonna realistically happen and what will happen is the rich becoming richer. 

[papajo]

Oh and 

https://youtube.com/clip/UgkxWSJSDb23XhYQHMU-ARK8R8if2My5aDEA

Well in MOST (especially for what's considered "mid range" and above and/or "gaming" ) facets of PC hardware e.g GPUs  and accessories e.g keyboards lol 

Also in the audiovisual industry 

 

market in general  is getting worse and worse it used to be that companies compete with each other on who is going to make the best products and sell to more people by giving the greatest value that will add to their fame.

Now its like how to get a kickstarter so that you get money in advance to half ass something you know you can sell to some ignorant or gullible people at a huge markup with the right marketing campaign

Even stupid stuff such as WEIGHT PLATES e,g for dumbbells have become overpriced AF because of people making fitness youtube video... 

 

Its casted metal (the cheapest kind of metal everything casted is cheaper ) and probably with impurities and recycled (because it doesnt have to be any specific alloy or have any specific other trades other than not brake by a sneeze and being heavy )  yet they sell as if they were jewlery or something lol.

[ProprietaryModules]

Wanted to toss in a few corrections and shed some more info on how the SSD side of things works on Apple Silicon (and also T2 macs by extension) since there was a lot of info missing or glossed over or not exactly accurate. This info can explain a lot about why macs are the way they are, why people struggled with upgrades on M1 Studio/M2 Studio Macs, and such. This will be a long post with a lot of info, but it will paint a clear picture of whats *actually* happening under the hood. 

Disclosure: I do work at an Apple Authorized Service Provider as my main job and therefore get a deeper understanding of how Apple's repair system works, and I do some board level Rossmann-style repairs on the side as well for money and spare change (Apple pays Service Providers peanuts, but thats a topic for another day).

For starters, It was said in the video that Apples SSD's utilize standard NVMe in order to chat with the NAND modules. They do not use NVMe and have not since the introdiction of T2 model Macbooks. Apple uses their own proprietary standard that simply runs over PCIe, so that's one huge but super crucial difference as far non-apple module upgrades go. Source: https://www.notebookcheck.net/Mac-Studio-SSD-does-not-work-on-NVMe-top-level-ARM64-SSD-controller-in-M1-Ultra-makes-it-nearly-impossible-to-swap-out-or-add-raw-storage-modules.609363.0.html This alone is enough to hinder most upgrades, but it goes way way way deeper. The way they actually work and communicate is a lot more complicated than you think, which you can guess based on the size of this post. 

 

Now for some Apple Storage Lore:

When the 2010 MacBook Air launched, it came with a proprietary SSD connecter that had a 6+12 pin layout. Even though the connecter is proprietary, it uses standard SATA AHCI, making it very easy to upgrade years down the line to mSata when it finally became abundant. The reason for this proprietary connecter is actually very simple, at the time when the MacBook was being developed, mSata was *just* announced in Mid 2009, which was already too late as far as the prototyping stages go, Plus at the time basically nothing utilized that standard. they used this specific connecter until 2012, when they redesigned it with a new 7+17 pin layout. it's still use the exact same signalling so I'm not sure why it was changed specifically, but it was also trivially easy to adapt to mSata and later Sata m.2. disconnect stayed until mid 2013 when Apple adopted a new 12+16 pin connector, which switch over to PCIe AHCI. This connector was in use from 2013-2019. even though at this time M.2 was announced, it wouldn't see actual usage until around 2015 in laptops so once again Apple simply designed its own standard because nothing viable and proven existed at the prototyping stages. Eventually with the release of high sierra, all those 12+16 pin drives were upgraded to utilize PCIe NVMe in order to facilitate the usage of the newly release APFS FIle System, which also was officially released with High Sierra. While upgrades of these 12+16 drives are possible, there are a lot of quirks with adaptor reliability, kernel panic issues after sleep/hibernate on 2013/2014 due to unfixed firmware bug, as well as a variety of wacky thermal and power draw differences between different brands of m.2 NVMe drives. Pain in the ass, but not the end of the world as far as drive stuff goes.

An example on a 2015/2017 Macbook Air with an M.2 adapter and Sabrent M.2 SSD. 

Moving on to 2016/2017, which is where things start to get interesting. There was 2 types of Machines released around this era, the Touchbar and Non-Touchbar variants. The more interesting one is the Touchbar, which we will get to in a bit. The Non-Touchbar introdiced a new connector type for SSD's yet again, a 22+34 Pin connector which sort of vaguely resembled a miniature PCIe slot in a way, but functionally was the same thing as the previous 2015 model. More than likely the SSD form factor was changed to physically fit into the design of these new logic board shapes, though this is me speculating. Ultimately its a minor note anyways because it was short lived and the Touchbar Mac is way more unique. The Touchbar Macbooks is where the SSD soldering stuff *truly* started. Yes I know 12" exists but that was for actual miniaturization reasons anyways given the logic board is smaller than a phone. Touchbar Mac intriduced the T1 chip, which was a small Coprocessor based on an Apple Watch S2 SIP. It was responsible for controlling Touchbar, handling Apple Pay, storing TouchID Data in Secure Enclave (SEP). Here is a block diagram of a T1 equipped Mac.


Take note of the SPI ROM chip with the Intel UEFI stuff, this will be important later. SPI ROM chip is functionally the equivalent of your computers Bios/UEFI, but with 0 user controls. It stores info like machines serial #, Intel ME Region stuff, Trusted Execution stuff, Machine Identifiers, and other such info that is needed for a Mac to actually Chime (post) and turn on. In this scenario, dead SSD wont wipe out anything SPI ROM related (foreshadowing!). This Block diagram also explains why the 2016/2017 Touchbar Macs are the only ones that had the Lifeboat connector on them, which is a connector used by Apple Techs like myself to do data recovery if the Logic Board is dead. It connects to a little black box that effectively connects directly to the PCie x4 link and acts as a PCIe to USB-c adapter and looks like this. 
On these models, it *is* possible to replace the SSD if it dies, but you effectively need to remove the NAND modules, controller, as well as a few other components and add a specialty PCB that solders directly to the PCIe X4 solder pads so you can utilize any tiny M.2 NVMe SSD's. 

In Late 2017, Apple introdiced the iMac Pro with the T2 Chip, and the T2 equipped Macbooks and Mac Mini. This is where the fun *really* starts, so pay attention. T2 macs are going to lay the basis for Apple Silicon macs, so all of this needs to be explained.  Lets start with a block diagram of a T2 Mac:

We can see a lot of new changes here, all of which I will explain. The T2 chip, which is basically a customized Apple A10 CPU, controls all sorts of things like TouchID, Touchbar, Audio, Facetime Camera, Intel eSPI, Power Managmement, etc. Its so complex that it runs its own OS called BridgeOS, which is basically a really scaled down version of iOS. Much like an iphone, it also has DFU mode (Device Firmware Upgrade, this will matter later) which allows it to be restored using another Mac. Notice the mising SPI ROM chip? we will get to the UEFI side of things, but basically the T2 chip boots from the chip that stores iBoot (apples own bootstraping stuff) which then boots BridgeOS. The whole operation of T2 chip is basically that of an iphone thats been frankensteined into the Logic Board of an Intel Mac, Its like 2 computers running at once.  On top of all this, the NAND used on T2 models onwards is...... very very different from literally anything else on the market. They are NOT standard NANDs like what youll find on a normal M.2 drive. The only other product that uses similar NAND are other Apple devices like iPhones and iPads.

Important notes about these NANDs:
- Each NAND is made by either Samsung, SK Hynix, Kioxia, Sandisk, and each one of them has a custom ARM-based controller thats also runs its own miniature OS as firmware, and acts as an intermediary between the PCIe x1 connection and T2 chip as wear as other NAND management stuff like wear levelling.

- These NANDs are CUSTOM DESIGNS! They CANNOT be found anywhere outside of Apple devices, and they CANNOT be purchased (apart from Mac PRo SSSD kit, more on that later)
- Each Logic Board has a pre-defined number of NANDs on them based on the chosen storage option from factory, and all the NANDS are paired together at a firmware level, meaning each set of NANDs behave as a singular unit after firmware is installed.
- These firmwares, which are specific to not only the NAND OEM, but also the NAND's capacity, cannot be interchanged with each other. Lets say your mac has 2x Kioxia 128gb (256gb Mac). We will call this firmware K128. If you wanted to swap one NAND with Samsung 128, which we will call firmware S128, They will not work because K128 and S128 are not shared. Same applies for every configuration (2x128gb, 4x64gb, 4x128gb, 6x 180gb, 4x1TB, etc). Theres more complications with swapping in 2x S128 NANDs which youll see later on...
- These firmwares also carry important data like TBW stats, power on time, SMART, etc. 
- These NANDs striped all data across them, similar to Raid0 (though its not *actually* Raid). This includes OS data, User Data, UEFI (below), and other stuff
- In the diagram above, notice how the intel UEFI is missing? Well, thats because it is split across all NANDs! They moved it from its little SPI ROM chip onto each NAND. This has a very fun effect where any dead NAND chip = machine wont even power on, because part of the critical Intel UEFI is litearlly missing. This UEFI striping will matter later as well. 

If you find and poke through schematics for these T2 Macs, you'll find sections similar to this:
This is from a 15" 2018 model Macbook Pro.

Dont be overwhelmed by the Schematic stuff! It's not super difficult. The columns indicate Internal Part Number, Part Quantity, part Info, Location on Logic Board, Part Importance to rest of system, and BOM (bill of materials) Group, which is just to figure out part costs and sourcing logistics and such for other internal teams. The prefix for each NAND stands for the type of controller and firmware running on chip (S3E, S4E, S5E) and suffix for each NAND (TS,WD,HY,SM) just stands for Toshiba,Western Digital, Hynix, and Samsung respectively. Also, notice the wacky capacities for some of them? 85GB chips, 170GB chips, etc. There's all sorts of these for various models. Now, lets refer to the total number of occupied chips as FLP (Functional Landing Pad). The S4E_1TB_WD option would be 6x WD 170GB NANDs paired together, giving us an FLP of 6 on that specific config. There's 2 other terms that will matter, NFLP and TLP. NFLP is Non-Functional Landing Pads, which are just locations on the board that are unoccupied and are missing their respective components. TLP refers to the Total Landing Pads, or the total number of spots NANDs could theoretically go. TLP is calculated by adding FLP's and NFLP's. Since the 2018 15" model has 8 TLP's, it means that depending on which storage config you have, you get either 4,6, or 8 NANDs total, with the remaining unoccupied Landing Pads becoming NFLP's. 

Here is a chart of Landing Pads for each T2 unit: 

But before we can do anything, there are also a few rules to these T2 Mac NAND swaps:
1: 128gb/256gb/512gb T2 Macs cannot exceed 512gb of storage. This is because the T2 SoC only has 1GB onboard Mem. 1TB models and up have 2GB Mem T2 variant. T2 itself also cannot be swapped for another for a multitude of other reasons. You're stuck, sorry. 
2: All NANDs must be moved as a set. You cannot mix and match NAND quantities, NAND brands, or NAND Capacities.
3: Macs will Larger FLP's can swap in NAND sets that come from Same or Lower FLP quantities, but cannot accept NAND sets from higher FLP configurations. Ex. Mac Mini with 4x FLP's can accept NAND sets from other macs with 4/3/2 FLP's, but cannot accept NAND sets from 6/8 FLP units, while Mac Pro 2019 can accept NAND sets from all other models provided other rules are followed.
4: NAND's must fill each Ports Landing Pads in order(Landings 0,1,2,3 on port 00. Refer to Block Diagram above!), and NAND sets must fill all landings on the 00 port in the block diagram before the 01 ports landing pads can be used at all. Example, swapping over 3 FLP's must be put into landings 0,1,2 on port 00 to work, 0+2+3 will not work, and putting those 3 FLP's into 01 port also will not work, they must go into port 00 since 00 isnt saturated. Swapping over 6 FLP's will fill port 00's 0,1,2,3 landings, as well as port 01's 0,1 landings. 

So, we can finally get into how some repairs are done. Basically, you have 3 choices: Used Donor NANDs, Used Donor Programmed NANDs, New Donor NANDs. 
- Used NANDs are basically pulled off of a board that has a compatible FLP number, and they are moved into each landing on each port accordingly (port00 nand0 to port00 nand0 on second board, port 00 nand1 to port00 nand1, etc). Upside is that it retains all data about NAND (TBW's, SMART, power-on hours, etc), downside is its used NANDs which means possibility of more worn down NANDs is there
- Used Programmed NANDs are basically same-size same-brand NANDs pulled from multiple boards, and using special programming jig, are flashed with pulled NAND firmware from another set of NANDs of same size and OEM, and each firmware is also paired in a striped manner (mystery NAND0 flashed with FW from good NAND0 firmware image, mystery NAND1 flashed with pulled FW  image from good NAND1, etc). Upside is that sourcing nands is easier and you can "create" a paired set, downside is that all the old firmware, including TBW and SMART and power-on and such is wiped and replaced with Fake data, so you have no clue how *actually* worn down your NANDs are. Pray they arent awful and dont die soon!
- New Donor NANDs is basically buying the Mac Pro SSD Upgrade kit, and pulling the NANDs onto the target board. Those Kits are pretty much always 8 FLP's meaning it will only wiork with other macs that support 8 FLP's (15"/16"/MacPro2019/iMacPro). 13" models cannot utilize this because of rule 3 above.

Assuming you follow all the rules above and successfully do the repair, now you just have to put the Target Mac into DFU mode, and restore it using Apple Configurator 2. On T2 models, this will wipe out all the data on NANDs and reinstall fresh Intel UEFI firmware across all FLP's, as well as installing the most current version of BridgeOS which allows T2 to work. After this, your machine boots up, goes into Internet Recovery mode in oder to Activate itself and fetch its Activation Profile (which sets things like iCloud Lock statis, if its under Device Enrollment Program which is remote management for enterprise clients, Wifi Region, and other parameters for that model). After this, assuming its not iCloud locked, just reinstall MacOS! And NOW you have working mac, unless another NAND goes of course, in which case you repeat process. 

Now, FINALLY onto Apple SIlicon. Luckily this section is short because Apple Silicon obeys almost all the rules and limits and quirks above, but with a few notable differences:
- Rule 1 of NAND swaps does not exist. The other rules apply for NAND swaps still. 
- Rule 4 becomes very important specifically for upgrades on the Mac Studio, and the recently released Apple Silicon Mac Pro. the reason being that port 00 and port 01 are literal physical port for the NAND carrier cards, and their physical placement inside the machine matters for this reason. This was never specifically documented on The Intel Mac Pro and iMac Pro specifically because both of those shipped with port 00 and port 01 populated with storage carrier cards, meaning that because of the Nando swap rules above, you can technically utilize basically every other capacity after DFU restore. This is why YouTuber and publications were having so much trouble getting consistently working upgrades on Mac Studio, or even storage swaps in general. They kept putting stuff in the wrong ports, or they purchased a studio that had all of Port01 set as NFLP from the factory (2TB and below iirc)! If you want to utilize both, you basically have to invest into the 4 TB model and up, otherwise you will be limited to single slot on the M1 max/M1 ultra Mac studio. More than likely the exact same stuff applies to The M2 version of the studio, as well as the recently released M2 Mac Pro. 

- T2 NANDs CANNOT be utilized on Apple Silicon due to electrical reasons! On Apple SIlicon macs, the NAND chips got power efficiency improvements. The old T2 chips operate at 2.7volts while the APple SIlicon NANDs operate on 2.5v, meaning they arent intercompatible. You can only source them from other Apple Silicon models
- Because Apple Silicon doesnt have anything Intel, all the striped Intel UEFI nonsense is gone, but instead replaced with iBoot for Apple Silicon bootstrapping. DFU restore basically reinstalls iBoot as well as Mac OS APFS Snapshot at the same time, similar to iPhone DFU restore. 

Here's a block diagram for reference:

So yeah, this is why we cannot realistically expect Apple Silicon and T2 macs to be upgradable in any form, even with consumer NANDs, and also some of the decisions surrounding pricing, part choices, why upgrading the removable storage models is a pain. This super-custom design is like a huge factor with regards to part pricing, since RnD on stuff like this is not exactly cheap, even for Apple. Not saying I condone this level of ridiculousness, but I can see why standard consumer, or even enterprise grade SSD pricing likely doesnt scale here. 

Now' I'll leave you with this nugget. The Schematic for the M1Max 16" Macbook Pro references NAND sockets, meaning that during prototyping, they were routinely swapping NANDs in and out. With that said, these types sockets are not small at all and require extra hardware to basically pressure fit the NAND to the Board, so the odds of this being a "They Couldve added tiny sockets!!!" scenario is extremely small and likely unrealistic.

 

Hopefully this adds a lot of important context about how it works on Apples side of things, which is.... literally like nothing else out there.
 

Huge shoutout to iBoff RCC on Youtube for creating this amazing 1 hour long video explaining all of the above in a visual sense! I used a lot of screenshots from their video because honestly its way easier than creating my own and they did a fantastic job explaining a lot of details anyways. I added a bunch of Apple Service Provider context as well as my own screenshots from my own info stash as well. 

And of course, Daily reminder to run Time Machine backups routinely!

 

 

 

 

[hishnash]

4 hours ago, Blademaster91 said:

That isn't the definition of right to repair at all, and is what is pretty much what companies like Apple do now, you have to have a certified repair shop with specialized tools fix the laptop or phone.

Not saying cerfired im saying anyone being able to buy the parts and secmatics. Also cerfried repair vendors with appel are not permitted to do any board level operations all they can do is the same that you can do with the self repair program. 

 

 

4 hours ago, Blademaster91 said:

It isn't realistic for the consumer to spend thousands on a soldering station and SMD rework tools.

Is it realistic for consumer to spend $ of dollars on the tools needed to do some car repairs no?  Right to repair is not make everything dumb so that you an reaper it without skill to tools its make it possible for people who have the skills and tools without needing approval form the HW vendor to do repairs.   Its not about change HW design (apart from replacing some forms of non removable glue resin) its about access to parts, access to documentation and access to firewere tooling so that professionals can be professionals and do repairs.   

 

 

4 hours ago, Blademaster91 said:

solder everything down when RAM or an SSD could be easily swapped

No if you need memory bandwidth you cant do a socketed ram without at minimum massively more power draw and slower speeds!  This is the laws of physics adding a socket massively reduces signal quality making a worce product.   And using solder does not impact the ability to repair as long as you can get parts etc.  

 

4 hours ago, Blademaster91 said:

IMO it should be the same as vehicles in terms of right to repair, standardized parts the consumer can buy that aren't locked down by the manufacturer.

But that is not the case, if you want to fully disable a car and put it back together you need a LOT of expansive tools, stopping an engine, doing this this and putting everything back so its done properly is not somthign you can do with a $5 screwdriver on your kitchen table while eating dinner.  Yes a consumer is expected to reapiace the wheel with a spare but even then just a spare, to do a proper weal replacement you like need a leather costly jig to check for alignment issues. 

If your idea of right to repair is to dumb down all products so every capaisto and every restore is socketed and nice and large withs standard values and and maybe no screws just but plastic clips then you want some very horrible devices that will break much more often and will be much much worce than current products. 

[hishnash]

4 hours ago, papajo said:

well those go hand to hand repair/upgrade its essentially the right of being able to do WTF you want with the  you paid with your hard earned money for. 

No, these are very differnt things.  Yes you should be able to do what you paid for but its OK if when you paid for it it was clear that you cant upgrade it then you opted to buy that product. 

Right to repair is about having laws in place that make it so companies cant sell products unless they also sell parts and provide schematics and firmware to enable repair.   If you mix in right to upgrade into that you get into the situation were your making the product worse, eg using socketed memory on a modern laptop would draw more power or run slower. But making it so that apple need to document which DRAM chips they support and that they cant sign contracts with factories making it impossible for those factories to sell on those chips to others is right to repair and does not make the product worce in any way.
 

[hishnash]

@ProprietaryModules Thank you for writing all of this up!

 

1 hour ago, ProprietaryModules said:

Important notes about these NANDs:
- Each NAND is made by either Samsung, SK Hynix, Kioxia, Sandisk, and each one of them has a custom ARM-based controller thats also runs its own miniature OS as firmware, and acts as an intermediary between the PCIe x1 connection and T2 chip as wear as other NAND management stuff like wear levelling.

- These NANDs are CUSTOM DESIGNS! They CANNOT be found anywhere outside of Apple devices, and they CANNOT be purchased (apart from Mac PRo SSSD kit, more on that later)

This is a key point with each module having its own controler and each module talking directly over PCIe these are very different from the NAND dies you would find in any other product. And are going to cost a good chunk more to make. 

 

[WolframaticAlpha]

12 hours ago, hishnash said:

Apple does just as much R&D they are not a dell

Talked about percentage of profit/revenue for Apple vs Intel.

 

Intel spent close to 18 billion on r&d in 2022 while apple spent 24.

 

 

Intel had a gross profit of 27 billion while apple had one of 95.

 

Percentage for Intel is significantly higher hence. That's why they have high profit margins.

[WolframaticAlpha]

13 hours ago, hishnash said:

Apple spends much more on R&D than NV

You cant simply say that so and so spends more than so and so without considering their sizes. Apple's market cap is close to 20x that of Intel and about 3x that of nvidia. No shit, they have to spend more. Their fraction of spending less though, because R&D cost of OSes chassis designs etc is objectively less than that of fabbing chips. Also not sure why you would want to compare to Nvidia, because their margins are often the subject of ridicule as well.

[naylor2006]

I just sold my Macbook Pro 2020 13", it was one of the last Intel ones and therefore had eGPU support. I know this is super weird, I have been PC gaming since the 90's (yeah im nearly 40 now) but for the last 7 years or so I daily drove this MacBook and PC gaming took a back seat in favour of my PS4 and PS5 etc.

 

But these Intel Mac were unique, sure it wasnt going to last forever but I could still enjoy CS:GO on my Mac hooked up to an eGPU, could run everything on high at 1080p 150+ fps on an external monitor..., it kinda provided me the best of both worlds....if you just wanna play one game which I did/do. Been playing HL games since they came out but lately ive gotten into more video editing and music production which Ive always found so much easier on Mac, I cant afford two systems.

 

The whole M1 M2 thing has pushed me back to PC now, my Mac held such a high value still I was able to build a pretty good gaming machine along with the GPU I already had. I just thought each time I buy a Mac im dropping nearly 2k and there's nothing I can do to improve it, now without the ability to use a eGPU is further restricts any kind of upgrade path, I know its not strictly the same thing as this topic but for me its yet another restriction and i'd have to be stupid to spend enough on another one to play CS:GO with the onboard gpu's now.

 

Dont get me wrong M chips are incredible but with further restrictions and the price of getting a half decent model you actually want is crazy, I cant complain though, actually plugging my GPU into a proper PCI-E port has tripled its performance over bottling it down thunderbolt. I actually feel stupid for running it like that for so long!

 

I accept my case is slightly different but I felt compelled to post from over here in the UK. 

[AndreiArgeanu]

14 hours ago, ProprietaryModules said:

*snip*

Interesting. And are there any benefits to this storage architecture to the end user? Performance or security? Is it possible for this level of performance/security to be achieved with standard nand packages and a controller?

[dogwitch]

15 minutes ago, AndreiArgeanu said:

Interesting. And are there any benefits to this storage architecture to the end user? Performance or security? Is it possible for this level of performance/security to be achieved with standard nand packages and a controller?

other then some edge case things with performance.(most users would never need)

its just  greed.

[papajo]

5 hours ago, dogwitch said:

other then some edge case things with performance.(most users would never need)

its just  greed.

I doubt that there are even edge cases where they overperform 

 

If I had to bet I would say that there are some edge cases where they underperform but they keep this "different" copy paste with some minor details changed protocol for greed. 

[SkyHound0202]

I won't necessarily call it a bad thing if a "starting at" price is low enough that makes a previously unobtainable tech (in terms of price and/or performance) available and accessible to more people, while also informing the customers what caveat ("catch") it may have.

 

For some (if not most) people, their biggest concern is not performance, but price. To them, solving the problem of owning a device take precedence over the actual performance of the device.