ATP Inspired me!
This ATP Podcast: https://mastodon.social/@atpfm/116971474099995854
Remote
DavidSchaub
@dschaub@mstdn.social
mastodon 4.7.2I'm just this guy, you know.
0 Followers
0 Following
46 Posts
Joined November 04, 2022
Website:
@atpfm@mastodon.social
@siracusa@mastodon.social
The RAM drought is over!
Replying to
@atpfm@mastodon.social@marcoarment@mastodon.social
Will Apple ever be a leader in AI models… probably not.
Does that matter… absolutely not.
AI is a commodity and a feature, not a platform. Apple will get enough AI, and companies that are nothing but AI will fail.
So I feel, Apple failed at something that doesn’t matter much, so nothing bad for Cook there.
Here's an exciting RAM Chart of Apple's cheapest consumer laptop base SKUs. Graphing the inflation-corrected price over time against the logarithmic scaled RAM capacities.
My main conclusion is: laptops are pretty darn affordable these days, and the MacBook Neo is a great value.
Replying to
@siracusa@mastodon.social
I think the causation is wrong there.
it isn’t “Tim is sticking around” that is the problem. It is “the Board and Ternus’ likely decision to continue to kiss Trump’s rear” that is.
Anything Tim does is all Ternus’ fault… ignore the shell game.
Replying to
Replying to
@Eric_Neue@indieweb.social @ianRobinson@mastodon.social @atpfm@mastodon.social
Beyond latency concerns, this doesn't solve the Thunderbolt bandwidth problem (being 1/4 of max PCIe slot bandwidth).
What I think Apple needs to do is:
-
Add "bonded multiple cable connections" to the Thunderbolt spec to allow 2-4 Thunderbolt cables between a Mac and a PCIe card slot chassis.
-
Then ship such a chassis for $1999 :)
Replying to
The tradeoff is: a 16-lane PCIe card always has the potential for 4x the bandwidth as ThunderBolt (which has traditionally using only 4-lanes of PCIe).
I'm sure there are lots (a few hundred?) of niche users which Apple is kicking off the platform, who use:
- high end networking;
- high end video / audio;
- high end storage;
- who knows what other shockingly small niches out there.
I'm sure Apple's making the right business choice.
Replying to
@darthnull@infosec.exchange
Yup, it is entirely plausible product, but I doubt Apple wants to make it.
Spec wise it is just an AppleTV Pro.
Maybe if the Neo does well, Apple will consider it later.
Apple doing a volume play with the MacBook Neo is awesome, either way.
Replying to
@dluz@hachyderm.io @atpfm@mastodon.social @marcoarment@mastodon.social
I would add, the M-Pro can't easily go away because it's primary value is:
- price, and
- power/thermals.
The M5 being powerful is immaterial. There are price points and power/thermal envelopes in Apple's product matrix that need filling.
The M3 was embarrassing; but otherwise, the Pro is so much cheaper than the Max, it likely deserves to keep existing.
The only threat to the M5 Pro, Max, and Ultra is: If the M6 lineup is ready as soon as the rumours suggest.
Replying to
@leonoverweel@mastodon.social
You are absolutely correct that slopes are sub-optimal.
It would still be wrong as it treats each year as a single atomic point of time.
Next one I build, I'll consider that change.
Replying to
@atlauren@mastodon.social @atpfm@mastodon.social
Oh, absolutely, the A18 Pro were almost certainly planned for the Neo from earlier in the SoC’s development, even if the Neo wasn’t fully greenlit.
5 GPU A18 Pros are still cheaper than 6 GPU A18 Pros.
I would still expect that only a portion of the Neos’ SoCs were “binned ones with one fewer GPU working from the iPhone”
Replying to
@siracusa@mastodon.social @aetles@mastodon.social
The LC and LC II felt too compromised.
The LC III was fine, and the LC 475 was great for the price. All that ‘040 cache.
I never really got the SE/30 love, as my first Mac (LC 520, basically an all in one LC III) blew it away in most ways, at 1/3rd the price. Living 4 years in the past makes computers seem even more amazing.
Replying to
@atlauren@mastodon.social @atpfm@mastodon.social
Cook is a figure head. Ternus will be in charge unless the board fires him.
@atpfm@mastodon.social
@siracusa@mastodon.social
Here's a new fun RAM chart, including the MacBook Neo.
Replying to
@illustro@mastodon.social
Anyway, I think Apple's trajectory is good, so the M6, M7, and M8 should be fine.
Replying to
@atlauren@mastodon.social @atpfm@mastodon.social
I don't know how Apple's bylaws work, but I would be surprised if the Executive Chairman actually has more voting power than the rest of the board.
The board made Cook Executive Chairman for appearances. He is not needed.
If it comes down to a conflict between the two, the board would be derelict to side with Cook.
Ternus absolutely has the actual power in this situation, so Ternus is the one who is responsible, unless the board gets rid of him.
Replying to
@atpfm@mastodon.social
A $399 Mac Neo with an A19 Pro and 12GB of RAM (even looking like an AppleTV), would be a great.
"Ultra" means "more expensive than Pro". That is all it will ever mean.
I don't think the MB Pro will be replaced with the MB Ultra. The higher priced M6 MBU will come out, and then after (or simultaneously) the lower priced M6 MBP will come out.
I also think it would be cool if Apple shipped a bigger 18" MacBook too.
Something like this M6 MB Line Rumours & Wishcasts:
Apple's Macbook Neo looks great for the price.
I would have accepted a poorer display for 12 GB of RAM, and the USB-C ports need some work.
Great that the most important wishcast is done... FOUR to go:
Has anyone heard any real data on the configuration of Apple's M5 chiplets?
I'm left guessing:
Maybe I closed my eyes too quickly due to how ugly the design was and how horrible the buttons felt?
If Apple is going to extend their higher-end laptop line-up, maybe they should REALLY extend it:
Here's a table of the possible price points of the M6 lineup, including a wishcast 18" model.
Replying to
@marcoshuerta@vmst.io
I expect it will be fine for light users, which is what the MacBook Neo is for.
Hopefully the MacBook Neo does as well as it should, and the 2nd Gen will come with 12 GB of RAM.
Does anyone know how Apple's chiplet fusion dies are configured?
I'm tempted to guess that they are:
- the CPU cores die;
- the GPU and Additional memory controllers die.
So we have:
- 2 bins of the same main die design across the M-Pro and M-Max;
- 3 or 4 separate GPU / Memory Controller die designs across lineup.
... which is really the proper way of doing how Apple chopped the M1-Pro vs M1-Max?
Maybe possible die designs like this?
Replying to
@atpfm@mastodon.social@caseyliss@mastodon.social
Regarding these "IPMI in a box" remote server devices...
The popular product recently has been:
https://jetkvm.com
Here's a Geerling video about is:
https://www.youtube.com/watch?v=6pYfHedjjgw
Replying to
@nwd@mastodon.online @siracusa@mastodon.social@carozas@mstdn.social
I roughy answered that in this post:
https://mstdn.social/@dschaub/111813544514647792
... around a TB of memory, which is obviously not needed.
It isn't that the slope needs to continue at that rate, it is just that it can't stay flat forever.
Replying to
@atpfm@mastodon.social @siracusa@mastodon.social
I'm not sure if your comparison is sufficient.
"Unified Memory" is mostly marketing.
I feel there are 3 models (2 & 3 blur together):
1) dGPU: CPU and GPU have separate pools of separate physical memory. [all dGPU]
2) iGPU (not unified): CPU and GPU have driver-dependant partially separate (sometimes configurable) pools of memory on the same physical memory [x86 iGPU]
3) iGPU (unified): CPU and GPU have a shared pool of memory on the same physical memory. [Apple Silicon]
Replying to
@eugenekim@mastodon.world @atpfm@mastodon.social @siracusa@mastodon.social
Welcome to the new normal.
I'm sure in the 2030s we'll be complaining about 16GB not being enough, but that's next decade's problem.
Replying to
@onyxraven@hachyderm.io
That's really hard.
At any time there is a huge range consumers can pay for RAM, due to type, functionality, density, speed, latency, etc...
We have no idea what kind of deals Apple gets for its RAM.
The chips are fairly standard, but DDR5X hasn't been really consumer RAM either.
The error bars might be wider than the chart.
I feel the opposite of corruption is bureaucracy, but everyone keeps saying they hate bureaucracy.
So which do people want more:
Replying to
@atpfm@mastodon.social @siracusa@mastodon.social
In theory, Apple's iGPU memory use could be more efficient than x86 iGPU memory use... but I have no idea how real that might be, or how dependent on any specific driver versions.
All it really means is: Apple controlling the OS and the hardware lets them make a cleaner implementation of what x86 & Windows also do.
Replying to
@illustro@mastodon.social @atpfm@mastodon.social
Apple's monolithic designs made sense, and scaled up to well to the Max.
With the M5 Pro/Max, Apple is now on heterogenous chiplets, but Apple's not scaling multiples of them (I think it is just a CPU die and a GPU die).
We don't know how they'll scale up to Ultra (or even Quadra... my preferred term), without designing NEW chiplets.
With the M7 and M8 we'll see if Apple gets to multiple CPU dies and multiple GPU dies to allow true efficient scaling.
With the Mac Pro being discontinued...
There are users with use cases that the Mac Studio cannot do.
Do you think:
- Apple would make an Apple branded external ThunderBolt PCIe card chassis?
- Apple would extend the Thunderbolt spec to allow bonded parallel Thunderbolt connections to mitigate how slow TB is vs PCIe (i.e. 1/4 the speed)?
Replying to
@joethephish@mastodon.gamedev.place @atpfm@mastodon.social @siracusa@mastodon.social
Given that most Apple Intelligence features are supposed to work (if not well) on 8GB Mac models, I think we can trust that that isn't the case?