M2 Low power mode and the future.

Jimmyjames

Site Champ
Posts
634
Reaction score
708
Reluctant as I am to post something from them, MaxTech just posted a video that made me think...



TLDR: the M2 has powerful efficiency cores...insanely powerful given the power usage. The video shows the M2 Air in "Low Power Mode" beating the latest Dell XPS 13 on some tasks.

I'm curious about these cores and would love the input of the many more knowledgeble members on here. What I wonder is: could these cores be the basis for a future P core? Could they be "scaled up" to give a future chip even more power and efficiency? Or are they a different kind of design with tradeoffs that would prevent todays M2 E core becoming a future M3 P core?
 

Colstan

Site Champ
Posts
822
Reaction score
1,124
Since you're looking for feedback from folks here, I figured I'd chime in with a grab bag of other items not worth their own topic, but could be of interest for discussion, so hopefully some knowledgeable people will have more to say on these topics.

Speaking of which, we have reached peak hype, because Uncle Gurman and Vadim from Max Tech have now combined forces to create the incompetence singularity that will destroy the entire solar system. Somehow, they managed to bloviate for 45 minutes.


Gurman is doing some damage control here and claiming that the M1 Mac Pro was ready to launch months ago, but Apple decided to wait until later this year for the M2 "Extreme". He says this despite the folks working on the Asahi Linux project saying that the M1 die wasn't designed for this, which mirrors what our own @Cmaier has said on the subject. Also, the Mac mini will not get a redesign, so now he is peddling non-news as news.

Tangentially related, apparently Apple has evicted the last Intel component from the Mac with the M2 MacBook Air, if iFixit's teardown is any indication. This is particularly amusing, because whenever Apple representatives are asked about Intel, they say that they have a "great relationship" with Intel. The unspoken part is that it's great because they no longer have to use their products.

Also, some chip guy who worked for Apple is is leaving for Samsung, so I guess it's big news whenever somebody changes jobs. This only applies to Apple, of course, not the other way around.
 

jbailey

Power User
Posts
164
Reaction score
180
I'm not going to watch but apparently he also said that the M2 is a stopgap whatever that means. It seems to be playing on the idea that Apple can just up and switch nodes whenever TSMC releases a press release that says N3 is going to be delayed. In this case moving the M2 from N3 to N5P.
 

Cmaier

Site Master
Staff Member
Site Donor
Posts
5,213
Reaction score
8,257
Reluctant as I am to post something from them, MaxTech just posted a video that made me think...



TLDR: the M2 has powerful efficiency cores...insanely powerful given the power usage. The video shows the M2 Air in "Low Power Mode" beating the latest Dell XPS 13 on some tasks.

I'm curious about these cores and would love the input of the many more knowledgeble members on here. What I wonder is: could these cores be the basis for a future P core? Could they be "scaled up" to give a future chip even more power and efficiency? Or are they a different kind of design with tradeoffs that would prevent todays M2 E core becoming a future M3 P core?


I think it more likely that someday apple might go to three levels of cores, with performance cores, intermediate cores, and efficiency cores, and that the current efficiency cores would make decent intermediate cores. That said, while i find that MORE likely than them scaling up the efficiency cores into p-cores, i DON’T find it VERY likely.

You really wouldn’t scale up - one of the advantages apple has is that their cores are custom designed for where they want them to be on the performance/power knee curve. You can’t just “scale up” an efficiency core and end up with an optimal performance core.

What is happening, though, is that Apple is taking techniques it has learned about making efficient performance cores, and translating them into its efficiency cores.
 

Cmaier

Site Master
Staff Member
Site Donor
Posts
5,213
Reaction score
8,257
Tangentially related, apparently Apple has evicted the last Intel component from the Mac with the M2 MacBook Air, if iFixit's teardown is any indication. This is particularly amusing, because whenever Apple representatives are asked about Intel, they say that they have a "great relationship" with Intel. The unspoken part is that it's great because they no longer have to use their products.

I think there may still be a 80186 core in the HDMI chip, though I’m not sure whether Intel has anything to do with that.
 

Colstan

Site Champ
Posts
822
Reaction score
1,124
I think there may still be a 80186 core in the HDMI chip, though I’m not sure whether Intel has anything to do with that.
I was surprised when I learned that Intel only stopped producing the 80186 in 2007. Those old embedded dogs really hang on. It reminds me of hospitals salvaging ancient computers to replace chips inside old medical equipment that can't easily be replaced.
 

Cmaier

Site Master
Staff Member
Site Donor
Posts
5,213
Reaction score
8,257
I was surprised when I learned that Intel only stopped producing the 80186 in 2007. Those old embedded dogs really hang on. It reminds me of hospitals salvaging ancient computers to replace chips inside old medical equipment that can't easily be replaced.

I’m dealing with 8051 assembly code and chip designs for work. These things hold on forever, seems like.
 

Nycturne

Elite Member
Posts
1,109
Reaction score
1,417
I was surprised when I learned that Intel only stopped producing the 80186 in 2007. Those old embedded dogs really hang on. It reminds me of hospitals salvaging ancient computers to replace chips inside old medical equipment that can't easily be replaced.

PPC-based microcontrollers are still purchasable but look to be getting to EOL as well. We almost used one of these back in college for our final project when we needed more oomph.

It was an earlier chip, still branded Motorola at the time though.

 

Jimmyjames

Site Champ
Posts
634
Reaction score
708
I think it more likely that someday apple might go to three levels of cores, with performance cores, intermediate cores, and efficiency cores, and that the current efficiency cores would make decent intermediate cores. That said, while i find that MORE likely than them scaling up the efficiency cores into p-cores, i DON’T find it VERY likely.

You really wouldn’t scale up - one of the advantages apple has is that their cores are custom designed for where they want them to be on the performance/power knee curve. You can’t just “scale up” an efficiency core and end up with an optimal performance core.

What is happening, though, is that Apple is taking techniques it has learned about making efficient performance cores, and translating them into its efficiency cores.
Exactly what I was looking for. Many thanks.
 

Jimmyjames

Site Champ
Posts
634
Reaction score
708
I'm not going to watch but apparently he also said that the M2 is a stopgap whatever that means. It seems to be playing on the idea that Apple can just up and switch nodes whenever TSMC releases a press release that says N3 is going to be delayed. In this case moving the M2 from N3 to N5P.
Yes, he is very confused overall. I just had a "discussion" with him in which he said I couldn't compare the M2 to the A15 in a positive way because it's different, but that he could compare the M2 to the A15 in a negative way because it's the same!
 

jbailey

Power User
Posts
164
Reaction score
180
I’m dealing with 8051 assembly code and chip designs for work. These things hold on forever, seems like.
Lovely chip. Can't really write C code for it because it really doesn't support a real stack. Pretty much either BASIC, Forth, or ASM. Some awful hacks to get performance out of it too.
 

Cmaier

Site Master
Staff Member
Site Donor
Posts
5,213
Reaction score
8,257
Lovely chip. Can't really write C code for it because it really doesn't support a real stack. Pretty much either BASIC, Forth, or ASM. Some awful hacks to get performance out of it too.
Not sure if it is lovely or not, but it sure is weird. Memory mapped registers. Multiple register banks kind of like SPARC. Accumulator (sometimes). Special function registers (also memory mapped). The physical constraints at the time somebody thought this all up were quite something; if you sat down with a clean sheet of paper today you wouldn’t end up with this thing :)
 

jbailey

Power User
Posts
164
Reaction score
180
Not sure if it is lovely or not, but it sure is weird. Memory mapped registers. Multiple register banks kind of like SPARC. Accumulator (sometimes). Special function registers (also memory mapped). The physical constraints at the time somebody thought this all up were quite something; if you sat down with a clean sheet of paper today you wouldn’t end up with this thing :)
With all the proliferation chips it gets even weirder—different chips have different features. The stack grows up and only works in internal RAM which is sometimes as little as 128 bytes. So you can't really push parameters on the stack. Some of the internal RAM is memory mapped onto the registers and there are some bitwise registers there too that I don't remember the details on. If you are lucky enough to have 256 bytes of internal RAM, then you can use the extra 128 bytes for the stack since you can only access it indirectly anyway. I don't remember what happens if you overrun the stack—something bad if I remember correctly.

One of the strangest CPU architectures I've ever worked with.
 

Citysnaps

Elite Member
Staff Member
Site Donor
Posts
3,601
Reaction score
8,819
Main Camera
iPhone
Lovely chip. Can't really write C code for it because it really doesn't support a real stack. Pretty much either BASIC, Forth, or ASM. Some awful hacks to get performance out of it too.
8051... a blast from the past!

I designed two defense/commercial wideband all digital radios using signal signal processing techniques ( wideband analog IF/RF -> high speed high dynamic range A/D converters -> tune/downconvert to baseband via NCO and multipliers -> FIR filter and decimate .... all (except A/Ds) in an ASIC) and used an 8051 as a controller to interface to a larger system. The 8051 was the quickest way to get the product out the door.
 
Last edited:

Cmaier

Site Master
Staff Member
Site Donor
Posts
5,213
Reaction score
8,257
You know, now that I think about it, I think the TI 9900 had memory mapped registers, which were actually part of the memory and not a seperate flip-flop-styled register file. I could be misremembering, but that was probably the first system I ever coded in assembler for and it’s ringing a bell. Anyway…
 

Yoused

up
Posts
5,511
Reaction score
8,685
Location
knee deep in the road apples of the 4 horsemen
Man, TI had some of the most awful advertising

9900_ad.jpg
 

Cmaier

Site Master
Staff Member
Site Donor
Posts
5,213
Reaction score
8,257
Man, TI had some of the most awful advertising

9900_ad.jpg
It worked on me :)

(First computer I owned was a TI-99/4A)


By the way, I have PDFs of a ton of old magazines from the 1980s and late 1970s, Byte and the like, and there are so many ads like that…
 

Yoused

up
Posts
5,511
Reaction score
8,685
Location
knee deep in the road apples of the 4 horsemen
By the way, I have PDFs of a ton of old magazines from the 1980s and late 1970s, Byte and the like, and there are so many ads like that…

When I was a teenager, I wrote to TI about their calculators and they sent me back a poster size ad with numerous graphical representations of metric system quantities, the largest image being "Body temperature is exactly 37°C". I leave what that was to your imagination (it was from a very grainy photo).
 

casperes1996

Power User
Posts
170
Reaction score
139
I think there may still be a 80186 core in the HDMI chip, though I’m not sure whether Intel has anything to do with that.
Made by Intel or not licensing agreements need to be in place with them to make an 80186, right? Unless AMD or VIA makes it I guess, since they already have that an x86 license - But even then they could make an x86(_64) but not an 80186, right?

But yeah it's wild with how the old chips stick around. One of the very recent space rovers (probably Mars) was powered by a PowerPC G3 IIRC. Lots of monitors with old 80186, 80286 like chips too.
I don't remember what happens if you overrun the stack—something bad if I remember correctly.
I think "something bad" is a safe bet regardless of what you remember :p
 

Citysnaps

Elite Member
Staff Member
Site Donor
Posts
3,601
Reaction score
8,819
Main Camera
iPhone
I worked for TI for four years. The experience was quite different coming from the startup culture of Palo Alto (California).
 
Top Bottom
1 2