Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

orthorim

Suspended
Feb 27, 2008
733
350
Apple will successfully scale the huge performance per watt lead into more high performance... M2 Pro is incoming...
 

bobcomer

macrumors 601
May 18, 2015
4,949
3,699
Old systems using software relying on x86 arch might as well be emulated on whatever architecture because they will never "feel" the performance penalty for emulation.
I don't think you truly realize just how much of a hit a pure software emulator is -- there is VERY much a penalty. An M1 is not fast enough to emulate an early pentium PC running Windows XP, not really even close, and that's a 20-25 year old machine, and it wouldn't use DOSBox, it needs a full x86 emulator. I tried, it didn't work near well enough to be useful.
 

bobcomer

macrumors 601
May 18, 2015
4,949
3,699
Are you suggesting that comparing performance between different architectures is not possible? That's a very radical viewpoint.
You can, but is it a useful measure -- I'm not so sure it means anything at all. Remember these things are to run software, that's their only purpose, and the amount and quality mean way more than just performance in a benchmark.
 

Shirasaki

macrumors P6
May 16, 2015
16,263
11,764
Why would you need to do a complete rewrite? As a dev, I find what you say extremely puzzling.
Your previous post gives me the impression that to take advantage of new hardware, a complete rewrite is necessary. It’s not the same thing as providing compatibility.
 

quarkysg

macrumors 65816
Oct 12, 2019
1,247
841
The only problem Apple has is that fewer and fewer people will need all this CPU horsepower. Most ppl just wanna browse the web and do some Google sheets.
That's viewing what a computer can do with what's possible today. We are still quite a ways away from a true personal digital assistant that understand what we say and can act on what we intend for a computer system to do ... think Jarvis in the Avenger movies. And this is just one of the many applications.
 

Shirasaki

macrumors P6
May 16, 2015
16,263
11,764
The thing is that for a lot of software Rosetta2 emulation/translation is allready good enough. As long as you don’t do stuff that requires long cpu times (like rendering) you can not tell if the software runs rosetta or is native.
What I’m talking about is not macOS apps. I’m talking about full fledge emulation of any x86 hardware with full performance running Windows. This virtual machine would function like any other, but instead of relying on hardware virtualisation, it completely uses software. M1 is not there yet.
 

Shirasaki

macrumors P6
May 16, 2015
16,263
11,764
Can you elaborate this last statement? I think the software transition has been a tremendous success. Majority of libraries, developer toolchains and popular commercial apps were fully native by spring 2021. There is still some lack of support for Apple proprietary technology but these things obviously take longer.
To me it is not enough. The remote control software I’m using doesn’t natively support apple silicon yet, and who knows when it will be there. VS Code somehow triggers “this app will not be compatible with future version of macOS” every time I launch despite being native to apple silicon. Battle.net is Intel only atm, and StarCraft 1 and 2 prolly will never receive apple silicon port. Those “major apps” you are talking about only counts for a tiny bit of macOS software library. Way more macOS apps are gonna be left behind than those that do get support, especially games.
And anyway, what libraries are you talking about? Even Intel own high-performance numeric libraries such as embree are natively supported on M1 hardware.
I’m talking about applications, not libraries.
 

Spindel

macrumors 6502a
Oct 5, 2020
521
655
Opening those apps remains painfully slow for me on the new MBP 16
Even if it's fun when apps just pop up as you click on them that really isn't what counts. It's if they work well when you actually use them.

For my personal experience Office before it was native ran just as good from a UX perspective, biggest difference I noticed when the AS native versions came was that their RAM usage halved.

OneDrive running trough rosetta works just as bad as it did on Intel macs.

MS Teams, except for eating RAM (as it did on my intel mac) runs without any noticeable performance hit.

MS Remote Desktop worked with no noticeable difference except more RAM usage before the native version launched.

AutoCAD runs smoothly with the caveat that the Mac version of AutoCAD has never been in feature parity with the Windows version (but that does not effect the work I need AutoCAD for). Performance wise it actually runs better then AutoCAD on my work issued windows laptop.

Civ 6, the only game I really play, runs really well with fast round times even in late game on huge maps. Runs better than on any other system I personally have run the game on.

The logitech camera settings app works without a fuzz.

Firefox ran well over rosetta before the native version launched (not that I use it that much since I mainly use Safari)

And so on...

The common thread I've noticed, with the applications I use, is that RAM usage is the most noticeable difference when a native version launches. For some reason all Intel programs seem to use more (and I mean a lot) RAM compared to their native counterparts. If this is an artifact of Rosetta or just x86 specific I really can't tell.


I've had my M1 Mini since around 20 November last year and during this year of ownership I personally can stand by that with the software I use/have used I can not tell if it's native or not by just using the application. Even with rosetta overhead programs run fast and smooth (well except maybe a bit slower startup times that you noted).
 
  • Like
Reactions: Christoph_SK

leman

macrumors Core
Original poster
Oct 14, 2008
19,521
19,678
To me it is not enough. The remote control software I’m using doesn’t natively support apple silicon yet, and who knows when it will be there. VS Code somehow triggers “this app will not be compatible with future version of macOS” every time I launch despite being native to apple silicon. Battle.net is Intel only atm, and StarCraft 1 and 2 prolly will never receive apple silicon port. Those “major apps” you are talking about only counts for a tiny bit of macOS software library. Way more macOS apps are gonna be left behind than those that do get support, especially games.

The case of individual apps is up to the app maintainer. This is just the question of developer focus and willing to maintain their software. And it is clear that not all apps will have the same quality. In this sense, moving to ARM is not that different from any other major macOS update — some apps take longer to fix the bugs and some never get updated.



You can, but is it a useful measure -- I'm not so sure it means anything at all. Remember these things are to run software, that's their only purpose, and the amount and quality mean way more than just performance in a benchmark.

The most basic useful measure is to do the same work on both machines and compare the time/energy used. For the workloads useful to me (tested on projects I work with), already the consumer M1 was as fast (or faster) than Coffee Lake i9 mobile. I expect my new M1 Max to be almost twice as fast. And benchmarks such as SPEC include these types of workloads as part of their suite. Their 502.gcc_r test for example is good proxy for software build performance and correlates highly with real-world tests.

Your previous post gives me the impression that to take advantage of new hardware, a complete rewrite is necessary. It’s not the same thing as providing compatibility.

I was referring to new Intel's paradigm of using low-powered cores for throughput. This is very different from Apple's approach where E-cores have an auxiliary function. In Alder Lake, E-cores are a real sustained performance driver. So if you want your app to work well with these CPUs, you better make sure that your code runs well on both P-cores and E-cores, properly scales with the number of hardware threads and does not do anything that would confuse the OS thread scheduler. Intel is doing exactly what AMD was accused of years ago — moving goalposts by throwing more cores at the problem.
 

Serban55

Suspended
Oct 18, 2020
2,153
4,344
Remember that this Alder Lake is a step forward for at least next 2-3 years...so dont expect much from Intel until 2024 while Apple....well in just 1 year and a half "

Future Apple Silicon Macs Will Reportedly Use 3nm Chips With Up to 40 Cores​

 

NotTooLate

macrumors 6502
Jun 9, 2020
444
891
Why not compare to the i5 and i7 12th gen cpus? The i5 pulls 125watts and still beats the m1 in single core. There will be 700 dollar laptops out soon that will go head to head with 3k Mac books
Your pov is so weird would it not be the same for the 4K windows laptops ? A laptop is more then it’s SoC , but putting that aside , let’s see those alder lakes in a 700$ laptops , the cooling and GPU needed to make it competitive not to mention screen will push it way over 700 , those desktop water cooled cpus that are now out are nothing to be proud of, when apple release their own halo desktop product we can look at the alder lake posts with a laugh.
 

grilledcheesesandwich

macrumors member
Jun 10, 2021
64
251
Your pov is so weird would it not be the same for the 4K windows laptops ? A laptop is more then it’s SoC , but putting that aside , let’s see those alder lakes in a 700$ laptops , the cooling and GPU needed to make it competitive not to mention screen will push it way over 700 , those desktop water cooled cpus that are now out are nothing to be proud of, when apple release their own halo desktop product we can look at the alder lake posts with a laugh.
Well, you'll have to laugh at Raptor Lake or Meteor Lake posts by then.
 

JMacHack

Suspended
Mar 16, 2017
1,965
2,424
You know, I may actually build an Alder Lake i5 PC for Linux experiments. There’s a mobo with some features that I really like (for 11th gen, but I expect a z690 version to have feature parity) the cpu seems to be decent cost wise vs Ryzen 3.

Though it’s gonna be in an old PowerMac G4 MDD case, so I’m concerned about the heat. I may try to underclock it.

The power consumption figures are astonishing though. It makes me wonder why Intel bothered with e-cores at all. 10nm is roughly equivalent to tsmc 7nm, and Zen consumes half the power in benchmarks, for comparable performance. Maybe AMD has a pure architecture advantage?
 

Serban55

Suspended
Oct 18, 2020
2,153
4,344
You know, I may actually build an Alder Lake i5 PC for Linux experiments. There’s a mobo with some features that I really like (for 11th gen, but I expect a z690 version to have feature parity) the cpu seems to be decent cost wise vs Ryzen 3.

Though it’s gonna be in an old PowerMac G4 MDD case, so I’m concerned about the heat. I may try to underclock it.

The power consumption figures are astonishing though. It makes me wonder why Intel bothered with e-cores at all. 10nm is roughly equivalent to tsmc 7nm, and Zen consumes half the power in benchmarks, for comparable performance. Maybe AMD has a pure architecture advantage?
Agree...for PC Intel still is an option...but for laptops if you can go with mac, its superb, if you dont, amd is a better overall package
 

dmr727

macrumors G4
Dec 29, 2007
10,668
5,770
NYC
Though it’s gonna be in an old PowerMac G4 MDD case, so I’m concerned about the heat. I may try to underclock it.

If you're gonna be modding the case anyway, fit a radiator on the top of that bad boy and water cool it. :)
 

zarathu

macrumors 6502a
May 14, 2003
652
362
Opening x86 apps remains painfully slow for me on the new MBP 16
None of my x86 apps like VLC player open slowly on my M1Pro. Are you sure that you don’t have the dreaded memory leak problem? Check Activity Monitor and see if something has grabbed all your available RAM.
 

Rigby

macrumors 603
Aug 5, 2008
6,257
10,215
San Jose, CA
The desktop i5 is 150W at Turbo, 125W at base and it’s MSRP is $264-$289 depending on igpu or no. It beats the M1 max by 5% in single core and by 0.7% in multicore. To accomplish this feat, it is burning 300-500% more energy. A 90W Alder Lake laptop chip isn’t likely to win any prizes against the M1 Max.
According to what benchmark? In Cinebench R23, the M1 Max doesn't come close to the i5 (~1600 single/12,000 multi vs ~1900 single/17,000 multi).
 

Rigby

macrumors 603
Aug 5, 2008
6,257
10,215
San Jose, CA
It has always been about realistic performance in a reasonable package. In this regard Apple is definitely the king of the hill and they will stay this way for the foreseeable future. It's really laughable that mobile Apple CPU with under 40W basically has the same performance as latest x86 desktop behemoth in workloads such as scientific compute and software development.

But hey, Intel is faster in Cinebench! That's gonna count for something, right?
Cinebench is a good benchmark for raw CPU performance. Application benchmarks are often influenced by things like SSD speed, which depend more on the platform than the CPU. The M1 also has a distinct advantage in memory bandwidth due to the fact that the RAM is integrated on the SoC, but that also limits the amount of RAM it can have and makes after-sale upgrades impossible.
 

crazy dave

macrumors 65816
Sep 9, 2010
1,453
1,229
According to what benchmark? In Cinebench R23, the M1 Max doesn't come close to the i5 (~1600 single/12,000 multi vs ~1900 single/17,000 multi).

The benchmarks he quoted and SPEC and quite a number of others. I already said why CB23 is a bad benchmark either on macOS or M1 not sure which if it’s bad coding for the OS or processor.

Cinebench is a good benchmark for raw CPU performance.

No it isn’t. Not on Macs.
 

Rigby

macrumors 603
Aug 5, 2008
6,257
10,215
San Jose, CA
The benchmarks he quoted and SPEC and quite a number of others. I already said why CB23 is a bad benchmark either on macOS or M1 not sure which if it’s bad coding for the OS or processor.
You said that not all cores may run, but even if true that wouldn't explain the significantly lower single-thread performance of the M1 in Cinebench.

No it isn’t. Not on Macs.
If you just start discounting benchmarks you don't like, you won't get a real picture of what's going in.
 

crazy dave

macrumors 65816
Sep 9, 2010
1,453
1,229
Not really blind faith, considering Tiger Lake remains much more respectable than Rocket Lake, a dismal showing by every performance metric. The existential problems for Intel lie in the desktop & server markets, not mobile computing. Intel will need to successfully iterate with Raptor Lake & Meteor Lake over the next couple years to regain market share from AMD, I imagine.

I gotta admit. I don’t know why Intel released Rocket Lake. Especially with the confusion over AVX512 here and Rocket Lake’s characteristics and the fact that Alder Lake desktop came out so soon afterwards. At some point during development they must’ve known it wasn’t working and yeah even if they learned lessons as some tech outlets put forwards releasing it was a weird choice.
 
  • Like
Reactions: JMacHack
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.