October 18th Apple Event - Unleashed - Apple Silicon, MacBook Pro upgrades, HomePod mini, AirPods 3rd Generation

[BondiBlue]

Alright, I've updated the post with some of the information from the event. I can't wait to upgrade from my 2017 MacBook Pro. The new 16" looks like everything I've been waiting for, except the notch. I could have done without that. 

[Flannelist]

1 hour ago, saltycaramel said:

Linus is gonna love those number-less graphs. 

Oh there were numbers. Meaningless numbers. The vertical axis was literally labelled "relative performance," with marks of 100, 200 etc. And no indication of WTF that actually means. it was just the marketing team drawing lines on a graph that made them look good, with absolutely no actual hard data.

[leadeater]

14 minutes ago, GDRRiley said:

but it is still based on "relative performance" and it doesn't say SOC, full system whats the power consumption being measured?

Well you pretty much have to do relative performance otherwise you have to pick something like CB, or SPECPerf (which is an aggregate so relative anyway). Also it says M1 Pro/Max and Laptop Chip so it should just be the CPU/SoC power which is in line with what Intel CPUs use. Note Apple is using PL2 boost power not PL1, so kinda ehh but also fine.

 

Edit:

Nope MSI has the PL1 cranked to the max on the 10870H, 90W PL1 / 200W PL2. Not the most power efficient configuration lol.

[AlTech]

25 minutes ago, leadeater said:

snip

Their GPU performance claim graphs against PC laptop dGPUs was pretty sus tho. No mention of exactly what they were comparing against or in what workload.

 

I would still argue not disclosing the workload in any graph (even the CPU graphs you said were done properly this time) is trying to manipulate potential buyers and it's wrong when any company does it whether that's Nvidia, AMD, or in this case Apple.

 

Relative performance graphs without the context of workload are almost meaningless imho.

[TheReal1980]

People on here who show what everything cost and don't like it please understand, people who buy these more powerful models are most likely using them to make money, they are not buying them to play games.

The price will be repayed fast if you work on the computer so the price means less than you think, especially to creators who use it to earn a living.

[n0stalghia]

4 minutes ago, Flannelist said:

Oh there were numbers. Meaningless numbers. The vertical axis was literally labelled "relative performance," with marks of 100, 200 etc. And no indication of WTF that actually means. it was just the marketing team drawing lines on a graph that made them look good, with absolutely no actual hard data.

Y axis is percentage, with M1 sitting at 100% and the other chips having relative performance to that.

[TheReal1980]

7.2GB/sec SSD, is that good on a laptop these days?

[RedRound2]

Just now, TheReal1980 said:

People on here who show what everything cost and don't like it please understand, people who buy these more powerful models are most likely using them to make money, they are not buying them to play games.

The price will be repayed fast if you work on the computer so the proce means less than you think, eapecially to creators who use it to earn a living.

Ikr! People for some reason here think that the $400 SSD upgrade is something they should cough up for their mom on her birthday

[Flannelist]

Just now, n0stalghia said:

Y axis is percentage, with M1 sitting at 100% and the other chips having relative performance to that.

Yes, but performance doing what? What tests, which tasks? Which version of the M1 because even within the M1 Pro and Max there are variations of Ram and CPU/GPU cores.

[WereCat]

Just now, TheReal1980 said:

7.2GB/sec SSD, is that good on a laptop these days?

It's a sequential speed which tells us it's a PCI-e 4.0 speed... which as a spec alone is meaningless since almost all if not all PCI-e NVMe SSDs will advertise that kind of sequential speed on their spec sheet.

 

Only relevant for moving large files like videos for example. Which may or may not be relevant for many Mac users since a lot of them edit videos.

[Freakwise]

1 minute ago, TheReal1980 said:

7.2GB/sec SSD, is that good on a laptop these days?

pretty sure you can get away with half or even quarter of that speed if you doing basic things or even light photo/video editing

[TheReal1980]

3 minutes ago, RedRound2 said:

Ikr! People for some reason here think that the $400 SSD upgrade is something they should cough up for their mom on her birthday

Yeah people need to understand that. People who buy laptops just to play games on their free time is in the end more expensive if you don't use it for any work and try to earn your money back.

[jimmy3189]

This is what apple says they compared the GPU performance to:

Testing conducted by Apple in August and September 2021 using pre-production 16-inch MacBook Pro systems with Apple M1 Max, 10-core CPU, 32-core GPU and 64GB of RAM, and pre-production 16-inch MacBook Pro systems with Apple M1 Pro, 10-core CPU, 16-core GPU and 32GB of RAM. Performance measured using selected industry‑standard benchmarks. Discrete PC laptop graphics performance data from testing Lenovo Legion 5 (82JW0012US). High-end discrete PC laptop graphics performance data from testing MSI GE76 Raider (11UH-053). PC compact pro laptop performance data from testing Razer Blade 15 Advanced (RZ09-0409CE53-R3U1). Performance tests are conducted using specific computer systems and reflect the approximate performance of MacBook Pro.

They look pretty good to me!

Long time lurker first time poster

[leadeater]

11 minutes ago, AluminiumTech said:

Their GPU performance claim graphs against PC laptop dGPUs was pretty sus tho. No mention of exactly what they were comparing against or in what workload.

Well they are working on perfect GPU scaling with cores which won't happen so it's not 2x or 4x but the 2x will be likely very close and the 4x could be like 3.5x to 3.8x (maybe better).

 

Also keep in mind Apple compares their performance in their applications which the M1 was already extremely good at. If you go outside of this to more general purpose compute the gap will start to open between the M1 Pro/Max GPUs and desktop add in cards. Thing is that's not a big use case on Mac OS like it is on Windows and Linux.

 

11 minutes ago, AluminiumTech said:

Relative performance graphs without the context of workload are almost meaningless imho.

Yea but if you pick a workload then that could also be highly misleading too, re: Intel "Real World Performance". Unless it's something like SPECPerf it's automatically highly misleading.

[RedRound2]

1 minute ago, Flannelist said:

Yes, but performance doing what? What tests, which tasks? Which version of the M1 because even within the M1 Pro and Max there are variations of Ram and CPU/GPU cores.

M1 Pro and Max have the same CPU with latter having 32 core GPU instead of 16 core GPU on the M1 Pro

 

They have shown clear distinctions in the GPU performance between the Max and Pro. CPU however remains the same

 

Unlike Intel's graphs, I doubt the workload is cherry picked. Historically Apple hasn't mislead with their numbers. And I assume it is some wrokload with raw CPU crunching task. And definitely not something like ProRes

[jimmy3189]

3 minutes ago, Freakwise said:

pretty sure you can get away with half or even quarter of that speed if you doing basic things or even light photo/video editing

You can do light photo/video editing on a phone, don't buy a £3000/$3000 computer to do light photo or video editing.

[leadeater]

2 minutes ago, jimmy3189 said:

You can do light photo/video editing on a phone, don't buy a £3000/$3000 computer to do light photo or video editing.

And a MacBook Air with the M1 running Final Cut or iMovie will be faster than most things anyway.

[DrMacintosh]

I’m very impressed with the performance Apple claims to achieved with M1 Pro and Max. My problem with these devices is at the architecture level though. ARM severely limits the use case of these new Apple Silicon Macs. Sure they’re great if everything you do is on macOS, but as soon as you need to fire up a VM to do anything else, the experience is gone. 

[AlTech]

7 minutes ago, leadeater said:

Well they are working on perfect GPU scaling with cores which won't happen so it's not 2x or 4x but the 2x will be likely very close and the 4x could be like 3.5x to 3.8x (maybe better).

 

Also keep in mind Apple compares there performance in their applications which the M1 was already extremely good at. If you go outside of this to more general purpose compute the gap will start to open between the M1 Pro/Max GPUs and desktop add in cards. Thing is that's not a big use case on Mac OS like it is on Windows and Linux.

Okay but they did a comparison to dGPU PC laptops. People running dGPU PC Laptops aren't going to be running Apple's apps on Windows so the use case is even more important because it's one that Windows users should reasonably be able to replicate and be more relevant to WIndows users.

 

Apple saying it's faster than the best dGPU for PC laptops without providing context is completely inappropriate and wrong unless they have the context and details to back up their claims.

Quote

Yea but if you pick a workload then that could also be highly misleading too, re: Intel "Real World Performance". Unless it's something like SPECPerf it's automatically highly misleading.

I would say it's more misleading to not provide one. Because at least if one is provided then people can decide for themselves if the numbers are trustworthy or not. As soon as you remove the labels so people can't make the judgement call then the numbers become de facto untrustworthy because that ability to scrutinize them has been removed.

[Flannelist]

18 minutes ago, RedRound2 said:

M1 Pro and Max have the same CPU with latter having 32 core GPU instead of 16 core GPU on the M1 Pro

 

The Pro, has 2 variations in CPU and GPU cores as well as the variation in Unified memory. And even though the Max has the same number of CPU cores there's a 24 core GPU variant as well, and there are still 2 Memory configs available. All of those things matter if we're judging performance. And it's misleading to just put one line on a graph which represents all of those variants.

Edited October 18, 2021 by Flannelist
Added 24 GPU core Max variant

[RejZoR]

I wasn't quite expecting such massive capacities and core counts for laptops. Makes me wonder what kind of beast desktop Mac Pro's will be...

[WereCat]

11 minutes ago, RedRound2 said:

M1 Pro and Max have the same CPU with latter having 32 core GPU instead of 16 core GPU on the M1 Pro

 

They have shown clear distinctions in the GPU performance between the Max and Pro. CPU however remains the same

 

Unlike Intel's graphs, I doubt the workload is cherry picked. Historically Apple hasn't mislead with their numbers. And I assume it is some wrokload with raw CPU crunching task. And definitely not something like ProRes

There are definitely ways where they could skew the numbers massively while not lying at the same time.

I'll give you an example since there was that benchmark with Cinema4D for rendering... I don't have experience with that particular application but I use Blender and the rendering may work in a similar way.

 

Let's say you're doing a GPU rendering and the project takes 10GB of VRAM.

Now you take a GPU with enough VRAM and do the render and it finishes in let's say... 5 minutes.

Then you take the same GPU with the same power but with less VRAM than the project needs and what happens? You need to keep sending that to system memory and trough CPU and you get a big bottleneck while sending data back and forth since it can't all fit in the VRAM so the GPU has to do the project in pieces.

 

Now what Apple did is they only have a shared RAM between CPU and GPU and they said that you "DON'T NEED TO COPY THE DATA BETWEEN CPU AND GPU" so even if LPDDR5 is much slower than a GDDR6 VRAM on a dedicated card, their benchmark might have been requiring more memory than what the VRAM could fit on the other laptop (Radeon Pro 5600M or whatever card that was) so there is your performance difference.

 

It's likely the reason and it's not misleading because if you work with large projects like 20GB-40GB+ then good luck finding a dedicated graphics card (especially in a laptop) that can fit that in a VRAM.

 

Only thing that sucks is that if you actually need more system memory and you don't care about the shared memory for the GPU.... you need to cash out hard just to get 64GB...

[RedRound2]

5 minutes ago, Flannelist said:

 

The Pro, has 2 variations in CPU and GPU cores as well as the variation in Unified memory. And even though the Max has the same number of CPU and GPU cores, there are still 2 Memory configs available. All of those things matter if we're judging performance. And it's misleading to just put one line on a graph which represents all of those variants.

 

I didnt realize that there was a 8 core CPU and 14 core GPU version. But they were afterall talking specifically about the 10c/16c version

Anywho, just waiit like a week. You'll get all your answers. And I honestly have no doubt that these are going to outperform notebooks in the same category, just like how the M1 did last year despite almost everyone on this forum being skeptcial about the whole thing, including Linus

[saltycaramel]

One crucial thing about Magsafe 3 compared to Magsafe 1 and Magsafe 2: it is now a detachable usb-c cable. 

Good for the environment and for connecting it to powerbanks, 3rd party usb-c chargers, etc.

[gjsman]

Apple didn't mention that there are 14-GPU-core versions of the M1 Pro, and a 24-core-GPU version of the M1 Max.