Apple Launches The M2 Pro & M2 Max + New Mac Mini With M2 / M2 Pro

[name99]

Those are still in the 5nm node right? It will be interesting to see a comparison against the future AMD mobile APUs in the same node later this year. Last year 6nm closed the gap a lot in the power/performance category. Now we will have a much better AMD64/ARM architecture comparison.

Yes. Which means they are somewhat behind where Apple is already (they are basically souped up N5 A15's whereas N4 A16 has already shipped in iPhones); and they are way behind the stuff that's coming up on N3.

It's very unclear quite where Apple is relative to where they "should" be and analysis is substantially conditioned by emotion.
One viewpoint is: after GW3 and Manu left, the whole thing fell apart and Apple from now on is basically just like any other vendor, optimized for mobile, but with minor unexciting improvements every year.
A variant of this theory doesn't blame the exiting engineers but says that's just the state of technology, that Apple has reached the far frontiers, and all improvements now (in process, and in SoC design) are so difficult they are slow.

Personally I see things very differently, that covid basically screwed over everyone's plans (this also includes, to be fair, AMD, Intel, and nV). Apple probably had a particular schedule in mind that included "SoC-next" designed for N3 and designed to ship as the A16. But then TSMC had to delay N3 (possibly internal failure, but I'm more inclined to think covid slowed down a plan that would otherwise have worked) and Apple had to ship the A16 as basically an A15 on N4 so with slight frequency boost.
Point is: in my worldview,
- M2 Pro and Max are somewhat behind schedule. They were probably supposed to have been based on the N3 design, but at some point it looked like that was too much of a delay.
- We still have the big chips (or to put it differently, the "desktop" chips) to look forward to, and my hope remains that those are the genuine N3 chips, fabbed on N3 using the SoC that was designed for N3 and not just a grown-up A15 or even A16. Obviously there's somewhat more slack in the schedule for those (Apple obviously wants to ship the Mac Pro, but can get away with delaying it to WWDC, and announcing these new designs gives them some media oxygen for a few more months.

If you look at the schedule we have
Sept 2020 A14
Nov 2020 M1
Oct 2021 M1 Pro Max
March 2022 M1 Ultra

Sept 2021 A15
June 2022 M2
Jan 2023 M2 Pro Max

Sept 2022 A16 (on N4)

To me this looks like a schedule that's constantly trying to delay things to see what of two options (plan A or a fallback plan B) are feasible, not a long-term strategy playing out as designed.
I think there's scope (given these schedules) for an M3 (if necessary based on the A16) say around to May, if Apple feels they need more PR, while still leaving a fair amount of time for the desktop designs by WWDC (June).
The real interest going forward, I think, is whether the Desktop line will (in some sense) split off from the mobile line. (Not drastically so, but think of how Intel separates Xeons from i5/7/9.) And whether Apple will simply drop desktops on N5/N4 and move them to N3 so that they now lead (rather than massively lag) mobile.

[pWe00Iri3e7Z9lHOX2Qx]

599$
need to spend 200$ more to have 16gb ram
need to to spend 200$ more to have 512gb ssd

This is always my struggle with Mac hardware. The base price is always enticing, but the upgrades are absolutely brutal. 16GB RAM / 512GB SSD is the bare minimum for any new machine for me. $599 sounds great. $999 for the minimum usable spec for me...not so much. And god forbid you want another 8GB of RAM beyond the 16GB tier, because that's another $200. $1,400 for a Mac mini with a 1TB SSD and 24GB of RAM sounds nuts.

[milkylainen]

People hating on Apple for soldered or expensive upgrade. But have they looked at recent laptop PC development?
If it's not soldered it's CAMM (Dell cr*p becoming a JEDEC standard) modules costing an arm and a leg.
So no. It's not just Apple. The entire industry is looking for gouging update methods.

[pWe00Iri3e7Z9lHOX2Qx]

lenovo also abuse also on ram and ssd

The critical difference is that if you want Windows or Linux, there's an enormous variety offered by myriad OEMs. It's still fairly easy to find machines that probably meet your other needs that also have user replaceable RAM and storage. Here's a fun apples to potatoes example.

I have a 14" M1 Pro MBP for work. Base model, 512GB SSD and 16GB of RAM for $2000. If I want 64GB of RAM, the machine costs $3,300. Yes, it costs $1,300 more to go from 16GB of RAM to 64GB, because it also requires the "Max" SOC. That $3,300 is still with a measly 512GB SSD.

I recently bought a new Asus ROG Strix G15 with an 8 core Ryzen 9 5980HX and a beefy Radeon RX 6800M for $1,100. 16GB of RAM and 512GB SSD, so density wise similar to the $2000 MBP. I splurged a bit on a CL20 kit, but upgrading it to 64GB of RAM only cost me $168. So for $1,268 I had a very speedy laptop with 8 CPU cores, a fast 12GB GPU, and 64GB of RAM. Not to mention there are 2 user replaceable NVMe slots.

Is that somewhat silly comparison? Sure. But the Strix can legitimately last 9-10 hours for basic desktop usage on battery which is nuts for a gaming laptop, so it's not like I'm comparing some DTR that only lasts 4 hours away from the plug.

tl;dr: Soldered RAM / storage is bad for us as consumers. That's true whether it is Apple or Lenovo or anyone else doing it.

[mdedetrich]

Me smiling at my 2TB SSD for less the price of that 512...

This is not normal SODIMM memory, its soldered next to the core logic and iirc its LPDDR5 so its much faster. In fact of the main reasons why the new Apple ARM laptops are so fast is that their onboard memory can run at much higher frequencies and bandwidth then what you can get in a typical x86 laptop.

This is slowly changing in x86 laptop world, newer laptops are coming out which have soldered memory and a new standard has been invented to replace SODIMM. The issue with SODIMM is that because its physically so far from the CPU, you are not able to maintain signal integrity for high performance memory (and yes we are hitting this point).

The critical difference is that if you want Windows or Linux, there's an enormous variety offered by myriad OEMs. It's still fairly easy to find machines that probably meet your other needs that also have user replaceable RAM and storage. Here's a fun apples to potatoes example.

I have a 14" M1 Pro MBP for work. Base model, 512GB SSD and 16GB of RAM for $2000. If I want 64GB of RAM, the machine costs $3,300. Yes, it costs $1,300 more to go from 16GB of RAM to 64GB, because it also requires the "Max" SOC. That $3,300 is still with a measly 512GB SSD.

I recently bought a new Asus ROG Strix G15 with an 8 core Ryzen 9 5980HX and a beefy Radeon RX 6800M for $1,100. 16GB of RAM and 512GB SSD, so density wise similar to the $2000 MBP. I splurged a bit on a CL20 kit, but upgrading it to 64GB of RAM only cost me $168. So for $1,268 I had a very speedy laptop with 8 CPU cores, a fast 12GB GPU, and 64GB of RAM. Not to mention there are 2 user replaceable NVMe slots.

Is that somewhat silly comparison? Sure. But the Strix can legitimately last 9-10 hours for basic desktop usage on battery which is nuts for a gaming laptop, so it's not like I'm comparing some DTR that only lasts 4 hours away from the plug.

tl;dr: Soldered RAM / storage is bad for us as consumers. That's true whether it is Apple or Lenovo or anyone else doing it.

As I explained, there is a reason why the memory is being soldered. Its because the performance of memory has become so high that you can no longer deliver it with SODIMM anymore.

Due to this the 16GB RAM on your Asus ROG strix is no way close to the performance of the memory you would find on a M1/M2, so its far from a fair comparison.

There is a newer standard for upgradable memory on a laptop which is meant to replace SODIMM, some Dell laptops have it but I can't remember the name of it. What the new standard does is it amongst other things, it places the memory much closer to the CPU and reduces the length of the traces but even that improvement is probably going to get outcompeted by soldered memory.

[t1r0nama]

599$
need to spend 200$ more to have 16gb ram
need to to spend 200$ more to have 512gb ssd

Here i bought 2x16 GB DDR4 3200 ram AND Crucial p5 plus 1 TB PCIe GEN 4 ssd for 168$. Apples prices are ridiculous.

[pWe00Iri3e7Z9lHOX2Qx]

This is not normal SODIMM memory, its soldered next to the core logic and iirc its LPDDR5 so its much faster. In fact of the main reasons why the new Apple ARM laptops are so fast is that their onboard memory can run at much higher frequencies and bandwidth then what you can get in a typical x86 laptop.

I'm well aware of this. The fact is that the glorious memory bandwidth does fuck all for most people's day to day experiences with these machines. My M1 Pro MBP doesn't feel any faster than an old Dual Xeon Haswell workstation that's even further gimped by using a 1TB SN550 DRAMless SSD. Both machines have the same security goop on them and the CPU side of the M1 Pro is more than twice as fast in single threaded performance. For myself and many others, having much more RAM is more of an enabler than having much more memory bandwidth but far less RAM.

I'd like my MBP even more with a pair of DDR5 SODIMMs and a pair of NVMe slots. The real driving force behind companies going soldered for both isn't performance, it's for revenue / ASP / margin. It's not like Apple just started soldering memory and storage when they moved to their own SOCs / unified memory. Lenovo was an early adopter of this on the PC side. And many now have machines with 1 channel soldered and 1 free.

[mdedetrich]

I'm well aware of this. The fact is that the glorious memory bandwidth does fuck all for most people's day to day experiences with these machines. My M1 Pro MBP doesn't feel any faster than an old Dual Xeon Haswell workstation that's even further gimped by using a 1TB SN550 DRAMless SSD. Both machines have the same security goop on them and the CPU side of the M1 Pro is more than twice as fast in single threaded performance. For myself and many others, having much more RAM is more of an enabler than having much more memory bandwidth but far less RAM.

I have an M1 max 14 inch and it compiles code 7-10x faster than my previous AMD 8 core/16 thread Thinkpad. I think you really under-estimate how important memory bandwidth is for certain applications.

Also do note that the M1 max/pro are designed to be for professional use. If you don't need that power there is the Macbook Air which still packs significant punch at a lower price point (and is completely passively cooled to boot!)

I'd like my MBP even more with a pair of DDR5 SODIMMs and a pair of NVMe slots. The real driving force behind companies going soldered for both isn't performance, it's for revenue / ASP / margin. It's not like Apple just started soldering memory and storage when they moved to their own SOCs / unified memory. Lenovo was an early adopter of this on the PC side. And many now have machines with 1 channel soldered and 1 free.

You are wrong in the case of Mac, it really is performance. I don't know if you realize, but a significant portion of Macbook pro users are either developers (they definitely need that power) or creatives that just like on the linked page at https://www.apple.com/newsroom/2023/...ro-and-m2-max/ are doing 3D rendering/video editing on 4/18/12k etc etc).

Of course Apple's business model puts a bigger markup on their hardware compared to their competitors, but I would argue that this is the first time for a while where the money you spend is actually worth it because if you compare performance along with efficiency, there is really no competition in the laptop space.

[pWe00Iri3e7Z9lHOX2Qx]

You are wrong in the case of Mac, it really is performance.

You realize that Apple has been soldering all the memory and storage on it's laptops for many years right? Long before there was a performance advantage. It all started to make more money. I'm not blaming them. They are a public company whose primary goal is to make as much money as humanly possible. But again, soldering memory and storage sucks for consumers.

[t1r0nama]

I have an M1 max 14 inch and it compiles code 7-10x faster than my previous AMD 8 core/16 thread Thinkpad.

I have compared M1 back to my i9 9900k and intel simply eat alive that M1 in code compilation. Even apple itself says that M1 max is only 70% faster then M1. I simply do not believe that M1 max compiles code 7-10x faster then any AMD zen processor with 8 core/16 thread, of course if you did not disable 3/4 of that amd cpu