M series chips are still slow... Are they gonna solve it?

[boss.king]

Because I look at real world use cases. A benchmark is just a set of numbers!

How are you measuring real world uses if not at least partly with numbers. If you don't quantify anything then it's meaningless. Ideally, you'd try to find relevant reviews or benchmarks that test the tools/apps that you use.

im sorry what? because i dont need benchmarks or reviews?
I have 14 days return policy to see how it helps me in my life and in my business
I dont care what some reviewers are testing..i didnt even see Maya users reviewers so?!

So you buy something going in blind and then complain when it doesn't live up to an expectation you set up for yourself? Seems like a recipe for sadness.

Actually, i don't believe even benchmarks, not anymore. Take a look at this thread i've opened:

https://forums.macrumors.com/threads/m1-handbrake-performance-very-disappointing.2345720/

I'm not really clear what I'm meant to be getting from that thread. Did Apple make any specific claims about Handbrake performance?

 

[Damian83]

How are you measuring real world uses if not at least partly with numbers. If you don't quantify anything then it's meaningless. Ideally, you'd try to find relevant reviews or benchmarks that test the tools/apps that you use.


So you buy something going in blind and then complain when it doesn't live up to an expectation you set up for yourself? Seems like a recipe for sadness.


I'm not really clear what I'm meant to be getting from that thread. Did Apple make any specific claims about Handbrake performance?

Point of that thread its that m1 its faster than my imac in maybe 28/30 benchmarks, then ive run just ONE "resources-greedy" app and after several tests its slower by around 50% than it, instead of being around 50% faster (like in most benchmarks). a huge difference...

 

[boss.king]

Point of that thread its that m1 its faster than my imac in maybe 28/30 benchmarks, then ive run just ONE "resources-greedy" app and after several tests its slower by around 50% than it, instead of being around 50% faster (like in most benchmarks). a huge difference...

But that doesn't make the chip bad, the software just seems poorly optimized, something that a handbrake benchmark would show. I don't get how this makes an argument against benchmarking, you just have find benchmarks for the thing you actually want measured.

 

[GuruZac]

I have the high-end 10nm 13” 2020 i5 MBP (with 4 USB-C ports) and that thing is really quite for that type of stuff.

It is currently used as a server now (because Apple trade-in value is so ****) and I haven’t heard the fans on that thing while acting as a server.

I don’t know the exact spec of his, but I know for sure he bought the cheapest one. It’s 2 thunderbolt port model and he’s not one who cares or even pays attention to RAM, specs, performance, so maybe that is the difference.

 

[stevemiller]

here's what I'll grant the OP... the people who are specifically in the market for a 3090 calibre graphics solution (and they know who they are) almost certainly won't be satisfied with apple silicon for their needs right now. apple invited this criticism by bringing 3090s into the conversation. sure there are technical metrics where it can be somewhat comparable - but at the end of the day M1 is not a gaming platform at this point, nor is it a path-tracing beast, and those are likely the overwhelming use cases for people who'd be in the market for a 3080 or 3090 system right now.

that said, I'm pretty sure nothing about apple's chips are slow. they are balancing performance (which is legitimately great in some areas, and more middle of the road in others) with considerations like heat, energy consumption, battery life, fan noise, etc.

for a lot of people's needs, that balance is likely an appealing one. I'd still love RT cores to speed up my 3d work, but this machine gives me more than adequate performance in a package that is generally more pleasant to use when I think about it overall. we'll see where they go in the future, but I think their general strategy for right now is plenty good.

 

[kschendel]

These whiny threads crack me up. If you need 3090 overall performance, why on earth would you look at an m1 or m2? If you need the multi-core performance of a 5950X or 12900K, buy one of those CPU's, and the 850w power supply you'll need when adding the 3090 into the mix. Don't complain that an M1 isn't a 5950X; a Tesla isn't a Ferrari either.

As has been said multiple times here and elsewhere, Apple Silicon is about a balance among CPU performance, graphics performance, power usage, heat output, and physical size. (An M2 chip is just a little bit smaller than a 3090!) Wanting Apple Silicon to best every metric and be cheaper than what Apple is charging, is just silly whinging.

 

[Unregistered 4U]

the people who are specifically in the market for a 3090 calibre graphics solution (and they know who they are) almost certainly won't be satisfied with apple silicon for their needs right now. apple invited this criticism by bringing 3090s into the conversation.

Not just right now, EVER. 3090’s are proprietary and they use proprietary methods. Anyone that depends on apps that make use of those proprietary NVidia methods will likely always be depending on them. And those will never run faster on m1.

And Apple invites criticism just by existing.

 

[sunny5]

If you still denying the fact, here you go.

 

[Jumpthesnark]

Doesn't really change the fact that Apple Silicon chip's performance is poorer than PC side.

Which doesn't matter to me in the least. I had to use PCs in the past and the awful experience of using Windows on a daily basis was enough of a reminder that I wouldn't touch a PC again on a bet.

I'm not sitting next to someone on a PC who's doing the same task as I am on a Mac and measuring which of us can do a CPU- or GPU-intensive task fastest. It's not a race. And if it were a race, we'd have to include the amount of time that I save by using the MacOS as compared to the time that is wasted fighting with Windows to get simple tasks done. Leave me alone, Clippy! 📎

The whole "which processor is faster at which limited benchmark" line of thinking serves nothing. It reminds me of the chip speed competition from the 1990s, which was entirely 100% a marketing ploy used to sell new computers, and for bragging rights among people who thought that such benchmarks were meaningful. It's still mostly used by gamers as a way to justify new purchases. Real performance comes from the integration between the OS and the chip, not a number on a spreadsheet.

 

[Damian83]

But that doesn't make the chip bad, the software just seems poorly optimized, something that a handbrake benchmark would show. I don't get how this makes an argument against benchmarking, you just have find benchmarks for the thing you actually want measured.

1) benchmarks should test "real-life" performances, not things used only by pros or not used at all
2) video encoding its something more common than most of the benchmarks included in geekbench, as almost everyone of us shrinked a video at least once in the life, while dont knows what, at least 25% of the benchmarks are.
3) if i buy a car thats 0-100 in 5s and vmax 250 im 99,9% sure it has that performance or very close to. i dont get then, in real life, 0-100 in 10s and vmax 130, only because the declared values was obtained in a downhill vacuum tunnel.
4) we are talking of numbers, so words like "bad", "good" doesnt fit here.

 

[boss.king]

1) benchmarks should test "real-life" performances, not things used only by pros or not used at all

What do you even mean by this? A benchmark can measure whatever you want.

2) video encoding its something more common than most of the benchmarks included in geekbench, as almost everyone of us shrinked a video at least once in the life, while dont knows what, at least 25% of the benchmarks are.

Who said anything about using Geekbench? A benchmark is just a standardized test that gives a common point of reference. If you're getting tripped up on the word benchmark, just substitute test or comparison. If you run the same Handbrake encode with the same settings, etc, across a few different machines and measure how long it takes, that's a rudimentary benchmark of Handbrake performance.

3) if i buy a car thats 0-100 in 5s and vmax 250 im 99,9% sure it has that performance or very close to. i dont get then, in real life, 0-100 in 10s and vmax 130, only because the declared values was obtained in a downhill vacuum tunnel.

Let's not delve into car analogies, please. Your underlying premise is wrong, but I believe I've already addressed the reason in the previous point.

4) we are talking of numbers, so words like "bad", "good" doesnt fit here.

So you don't want to measure with numbers, but you also don't want to discuss relative terms like good or bad. How are you supposed to describe performance, then?

 

[s.g.w]

Well that was 10 minutes of my life I am not getting back.

Summary.. Computer users needing beefy CPU/GPU...... Mac users <> PC users

 

[stevemiller]

Not just right now, EVER. 3090’s are proprietary and they use proprietary methods. Anyone that depends on apps that make use of those proprietary NVidia methods will likely always be depending on them. And those will never run faster on m1.

And Apple invites criticism just by existing.

ok I'll bite.

sure... apple isn't going to be supporting cuda or optix since they're proprietary. if your software only supports that implementation... of course. but metal may eventually gain more and more similar features over time. and their in house gpus will hopefully get more capable too. I won't be surprised if they're always still behind Nvidia's top offerings, but again thats catering to a pretty specific market.

Citing my own use case. I can absolutely render faster on a 3080 than my m1 max. but final outputs usually go to a render farm anyway. in terms of the vast majority of my workflow, the Mac stays quiet, cool, and I can move away from my desk without any performance degradation. I don't feel slowed down by the M1 system the way my 2018 MacBook Pro did. and boy was I ever considering getting a PC tower at that point.

I mean its also ok if you're like "nope the Mac won't work for me ever." everyone's got different needs, but I think Apple's serving a much better cross section of users now than they were previously.

 

[Damian83]

What do you even mean by this? A benchmark can measure whatever you want.


Who said anything about using Geekbench? A benchmark is just a standardized test that gives a common point of reference. If you're getting tripped up on the word benchmark, just substitute test or comparison. If you run the same Handbrake encode with the same settings, etc, across a few different machines and measure how long it takes, that's a rudimentary benchmark of Handbrake performance.


Let's not delve into car analogies, please. Your underlying premise is wrong, but I believe I've already addressed the reason in the previous point.


So you don't want to measure with numbers, but you also don't want to discuss relative terms like good or bad. How are you supposed to describe performance, then?

for benchmark i intend geekbench as its the most used, and considering it includes several benchmark types, it seems to be a trustable source for compare machines. but u get then a lot disappointed when a relatively common task such as software encoding has nothing to do with these benchamarks. if i tested ONE random app, and its results are opposite of the geekbench then i can image how many other apps will have same results... resuming, facts are: apple says m1 mba its faster than my "old" intel imac, geekbench confirms it. i say (i know) its fake

 

[Unregistered 4U]

but metal may eventually gain more and more similar features over time. and their in house gpus will hopefully get more capable too. I won't be surprised if they're always still behind Nvidia's top offerings, but again thats catering to a pretty specific market.

And that’s the point. Nvidia knew what they were doing by locking folks into their methods. With that huge library of already solved and endlessly tweaked (by the community) solutions, there’s no way anyone deeply invested in that workflow is ever going to uproot that legacy work just because the M1 chip is more efficient.

I mean its also ok if you're like "nope the Mac won't work for me ever." everyone's got different needs, but I think Apple's serving a much better cross section of users now than they were previously.

I think I’m only saying that anyone actually using Nvidia’s proprietary methods are NEVER going to be satisfied with any Apple Silicon solution for one main reason… it’s not Nvidia. Fortunately for Apple, there are millions each day that will start out and will use Apple Silicon over the upcoming years. The library of Apple Silicon solutions will grow, the community will hone and tweak those for performance and those future folks will find themselves doing impressive things on laptop form factors that run for hours.

 

[Unregistered 4U]

When Qualcomm’s talking about who they’re gonna be faster than, they don’t say Intel, they say Apple.

 

[aevan]

OP, you lost me…. with the title of the thread.

 

[aevan]

Doesn't really change the fact that Apple Silicon chip's performance is poorer than PC side.

You’re comparing different things. Apple Silicon is the fastest chip in it’s class. Yes, you can find faster PC chips, but they require a power plant to run. And Apple still hasn’t transitioned their most powerful Mac to Apple Silicon.

Nice try.

 

[Cognizant.]

You’re comparing different things. Apple Silicon is the fastest chip in it’s class. Yes, you can find faster PC chips, but they require a power plant to run. And Apple still hasn’t transitioned their most powerful Mac to Apple Silicon.

Nice try.

There are no chips like this in terms of power to performance ratio. It's honestly mind boggling. When "efficiency" is not a concern for the Mac Pro like it is for other devices, it's going to be interesting to see what Apple does with it.

 

[Damian83]

There are no chips like this in terms of power to performance ratio. It's honestly mind boggling. When "efficiency" is not a concern for the Mac Pro like it is for other devices, it's going to be interesting to see what Apple does with it.

I dont think someone that buy apple stuff, have "problems" with utility bills

 

[Cognizant.]

I dont think someone that buy apple stuff, have "problems" with utility bills

That's utter nonsense. People pay the same or more for equivalent PCs. These are tools we need for life or work or whatever. It's still nice to cut costs wherever we can, especially in this economy. Battery life is also extremely important.

 

[GuruZac]

You’re comparing different things. Apple Silicon is the fastest chip in it’s class. Yes, you can find faster PC chips, but they require a power plant to run. And Apple still hasn’t transitioned their most powerful Mac to Apple Silicon.

Nice try.

👏👏👏

 

[nastysailboat]

the click bate on here is unreal

 

[boss.king]

for benchmark i intend geekbench as its the most used, and considering it includes several benchmark types, it seems to be a trustable source for compare machines. but u get then a lot disappointed when a relatively common task such as software encoding has nothing to do with these benchamarks. if i tested ONE random app, and its results are opposite of the geekbench then i can image how many other apps will have same results... resuming, facts are: apple says m1 mba its faster than my "old" intel imac, geekbench confirms it. i say (i know) its fake

Dude. No. That’s not how measuring works. It’s not how benchmarking works. It’s not how anything works.

 

[cocoua]

Despite Apple's claims, M1 Ultra GPU is not as powerful as RTX 3090 - VideoCardz.com

No, Apple M1 GPU is not as powerful as RTX 3090 Apple may need to reconsider its M1 silicon marketing. The company was claiming that their newest chip called M1 Ultra equipped with 64-core GPU is as good as high-end desktop GPU while consuming 200W less power. As proven today, this claim...

videocardz.com

Apple’s charts set the M1 Ultra up for an RTX 3090 fight it could never win

Better performance per watt doesn’t equal better overall performance.

www.theverge.com

M1 series especially M1 Ultra still suffer its GPU performance and it turns out Apple lied about the performance. Yes, the optimization is still a problem but people still believe that M1 series are still performing poorly. Do M1 Ultra really perform just like RTX 3090? How about now? If this continue in 2022 and later, it will be a huge problem. I'm still seeing videos of poor M1 series performance due to lack of optimizations.

Power per watt is the only advantage for M1 series so far but dont we really care more about the performance? At this point, I'm already worrying about the future of M series.

it is true that there are better performance PC than Macs, as always it's been, BUT the cost is also quite higher in every aspects, larger and heavier devices, much more power consumption, and similar or even higher retail price.

So Macs are the best balance devices right now by far. And most balance performance/price/features. The whole industry has changed and Apple's leading.

Just think this: Apple did the change as soon as a the "slightest" difference could be notice, so further improvements are coming. Modern ARM SoCS are an emerging tendency and they has come to replace x86 in long term.

The lap between M1 Vs M2 in performance and power consumption has not been seen in X86 since pentium era (as last decade the performance was upgraded by adding multicores with extra heat and power consumption, but single core improvements were 10% each year), and is yet to see the improvement in the Max/Ultra areas.