The yearly M-series release cadence

[crazy dave]

Hmm, I still think replacing the 27” iMac - a 2000 USD semi-Pro computer - with a Mac Studio and Apple Display Pro-level combo which runs to 3800 USD is a cash grab aimed towards independent creators who need power.

If you’re in that bracket your choices are either a Mac Mini with the M4 Pro chip, or a Mac Studio with the M4 Max.

While I grant you the Apple Studio display at $1600 is extremely expensive, the nice thing about the mini and the Studio is you can pair them with any display - and you can still get the 24" iMac with the base Mx chip* if you'd like. Overall, that's a substantial savings unless you have to have the Studio display for some reason and never want to upgrade the mini without also upgrading the display.

Sure, it'd be nice to get a Mx Pro iMac with a bigger screen too, but back before Apple Silicon, the "headless iMac" Studio/Pro mini is what people were begging Apple to release because the 27" iMac and iMac pro were considered thermally limited and expensive/clunky as you HAD to get a screen and computer in one (yes there were people who also loved them, I loved mine). There were so many threads and complaints about that on this forum. Again, if you WANT an Apple Display + Studio and never want to upgrade one without the other then, sure, a comparable iMac makes sense, would be more cost effective, and certainly save a bit of space. I remember trying to explain to people at the time, that for those who want an Apple screen to go with their computer and don't upgrade often, the old iMac was actually good value. But for a lot of people in that market segment, that isn't want they want and this solution is almost certainly cheaper for them. Basically this forum is never happy.

*and I'd argue if GPU power isn't essential, then the base M1-5 processors deliver as much (if not more) CPU performance as the equivalent Intel CPUs that would've gone into the base $2000 iMac, making the $600 mini an excellent deal (or 24" iMac if the screen works for you)

 

[Bodhitree]

Back in the early Intel days, my stepfather (who was the original designer and Mac enthusiast in the family) used to be in that bracket of “those who want an Apple screen to go with their computer and don’t upgrade often”. He was a big fan of the 27” iMac, even though he did get hit twice by the thermal failures that used to happen.

Personally I’m very happy with my 24” M1 iMac 16/512, it still does everything I need swiftly and securely.

 

[snakes-]

I miss efficiency instead of speed. Even the M5 chip having high temps and consumes more power as the M1.

 

[thenewperson]

I miss efficiency instead of speed. Even the M5 chip having high temps and consumes more power as the M1.

You mean the lower power usage? M5 is more efficient than M1.

 

[deconstruct60]

You mean the lower power usage? M5 is more efficient than M1.

Peak load or average load usage? M5 probably races to sleep faster. At M1 performance limitation levels it is more efficient.

 

[crazy dave]

Peak load or average load usage? M5 probably races to sleep faster. At M1 performance limitation levels it is more efficient.

I'm not quite sure what you mean by that last comment, but looking at data I don't really see this. Unfortunately I don't have great M1 efficiency data myself, but the M3 had a similar power design philosophy to the M1. It wasn't until the M4 that Apple allowed the base M chips to get really more power hungry, but Apple still reined it in when it made sense to do so (e.g. the M4 Airs). An example using Notebookcheck CB R24 data* is in the thumbnail below (click to enlarge). You can see in the multicore tests that Apple never really lets the new chips get significantly less efficient than the old design (<= 10%) even under higher power draws and under the same or similar power draws is significantly more efficient (extra and better E-cores being the main benefactors here). And the efficiency of the ST test is pretty stable across the generations. I *think* Geekerwan has generally shown improvements in efficiency in ST SPEC tests across generations, but I'd have to double check.

*the tested M5 MBP had unfortunately double the RAM of the M3/M4 devices, which may mean its power usage may be slightly higher than it otherwise would be

 

[deconstruct60]

I'm not quite sure what you mean by that last comment, but looking at data I don't really see this. Unfortunately ...It wasn't until the M4 that Apple allowed the base M chips to get really more power hungry,

If M5 is more power hungry than M4 then that would be the point. A 'peak load" is looking for the corner case, not avoiding it. Yes, Apple scales the consumption to the workload to save power in many normal usage conditions.

And the efficiency of the ST test is pretty stable across the generations. I *think* Geekerwan has generally shown improvements in efficiency in ST SPEC tests across generations, but I'd have to double check.

How could running a single thread be a peak load? (cores with no work that have been put to sleep aren't going to consume much power). Running all the cores (or at least as many as possible) all at the same time, very likely consumes more power.

 

[crazy dave]

If M5 is more power hungry than M4 then that would be the point.

Efficiency is about work done per unit of power. Despite the base M4 and M5 being more power hungry while doing work, they are as or more efficient than the base M1-M3 thanks mostly to the additional (improved) E-cores allowing them to do the work at least concomitantly faster (and of course improvements to the P-cores and new nodes) - the M4 dipped a bit, but the M5 has brought efficiency back to parity even under the fullest load and when not under the fullest load, the M4 (and likely the M5) is far more efficient.

It is fine to say that one would prefer the base M chip remain at lower power envelopes only, but it is incorrect to say that Apple has sacrificed efficiency for speed. At worst you could say that for non-Air Macs they sacrificed additional efficiency improvements for speed, but even in those devices that Apple lets the M4/M5 run unconstrained (or at least less constrained), those chips are just as efficient as before. And obviously the M4/(probably M5) MBA Airs and probably iPad Pros have gotten far more efficient due to the thermal constraints limiting power.

A 'peak load" is looking for the corner case, not avoiding it.

I'm not sure what you mean by this.

Yes, Apple scales the consumption to the workload to save power in many normal usage conditions.

Yes ...

How could running a single thread be a peak load? (cores with no work that have been put to sleep aren't going to consume much power). Running all the cores (or at least as many as possible) all at the same time, very likely consumes more power.

From your original post it wasn't clear to me which issue you were referring to. However, my point was it didn't really matter, efficiency under full load or ST hasn't really changed or has improved over the generations, not gotten worse.