M2 Pro and M2 Max

[theorist9]

First gpu geekbench score I've seen
86805

Mac14,6 - Geekbench

Benchmark results for a Mac14,6 with an Apple M2 Max processor.

browser.geekbench.com

The M2's Metal score was 30,600 with 10 GPU cores so, with linear scaling, you'd expect the M2 Max's to be 92,000 and 116,000 for the 30-core and 38-core GPU models, respectively.

 

[theorist9]

As for the die shots provided by Apple, how trustworthy do we think they are in reflecting the true nature of the chip? I'd like to see one actually de-lidded.

Here you go...

 

[casperes1996]

Does anyone know, does the single core score for an Intel processor run entirely at "Turbo" speed, being so short?

For Geekebench, you're almost entirely seeing Turbo, yes. Some fluctuations may occur and Tau (allowed turbo-time) is adjustable by OEMs/motherboard makers (in the past especially, a lot of desktop motherboards set Tau to infinity so they could turbo indefinitely. Technically against Intel guidelines, but allowed), so YMMV from device to device, but broadly speaking, yeah it's turbo most of the way

 

[Colstan]

The 16-inch M2 Max also gets slightly higher performance in Geekbench Metal than the 14-inch M2 Max.

 

[Cmaier]

In some nondescript building leased by Qualcomm now, the Nuvia guys are starting over from scratch again…

 

[casperes1996]

In some nondescript building leased by Qualcomm now, the Nuvia guys are starting over from scratch again…

You don't think Qualcomm will just settle the legal dispute and agree to new terms with SoftBank/ARM? That seems like it'd be less costly than starting fresh

 

[Andropov]

A M2 Pro has surfaced with 52691 points in Geekbench Compute: https://browser.geekbench.com/v5/compute/6261915

That's more than 40% faster than M1 Pro, in a single generation. The M2 Max hasn't improved by that much relative to the M1 Max... (~25% if the benchmarks are true). May have something to do with what we've been talking about Geekbench compute being too short?

In some nondescript building leased by Qualcomm now, the Nuvia guys are starting over from scratch again…

Would they be expecting any different? I mean, the M2 Pro/Max are shaping up to be *great* improvements over last year's SoCs, but... not exactly unexpected.

 

[Cmaier]

You don't think Qualcomm will just settle the legal dispute and agree to new terms with SoftBank/ARM? That seems like it'd be less costly than starting fresh

I was referring more to the fact that indications were that they might match M1 speeds. They may have taken some heart from recent speculation that M2 was going to be not great - I mean, look at the other place at all the geniuses who insist M2 sucks because all the people who know what they’re doing left for Nuvia.

Given Nuvia’s delays, whatever product they’ve been working on all this time would seem likely to already be far behind. They should be targeting M3 at this point, which seems a big ask.

 

[casperes1996]

I was referring more to the fact that indications were that they might match M1 speeds. They may have taken some heart from recent speculation that M2 was going to be not great - I mean, look at the other place at all the geniuses who insist M2 sucks because all the people who know what they’re doing left for Nuvia.

Given Nuvia’s delays, whatever product they’ve been working on all this time would seem likely to already be far behind. They should be targeting M3 at this point, which seems a big ask.

Oh I see, I misunderstood. Well, once the legal stuff with ARM is sorted out they can at least iterate on the ideas there instead of having to design something new. I'm not even sure how the legal stuff there would work cause you can't forget things you've learned from working on that, so while you may not be able to take things from that design over if they lost the legal argument with ARM and didn't sign a new license, they would still have the knowledge in their bran from it.

But in a way they don't really have to compete with either M2 or M3. The chips in Android phones are somewhat consistently 2 years behind Apple and I don't think any Android user will switch to iPhone purely because of this. Likewise, even if A chips were slower then Qualcomm and Samsung, I wouldn't jump to Android. I think much the same is true for a large portion of the PC/Mac space.

 

[casperes1996]

I was referring more to the fact that indications were that they might match M1 speeds. They may have taken some heart from recent speculation that M2 was going to be not great - I mean, look at the other place at all the geniuses who insist M2 sucks because all the people who know what they’re doing left for Nuvia.

Also the tendencies have switched! They're now saying Apple should acquire AMD, because the new mobile Ryzen chips can supposedly compete with M1 Pro!

 

[Cmaier]

Oh I see, I misunderstood. Well, once the legal stuff with ARM is sorted out they can at least iterate on the ideas there instead of having to design something new. I'm not even sure how the legal stuff there would work cause you can't forget things you've learned from working on that, so while you may not be able to take things from that design over if they lost the legal argument with ARM and didn't sign a new license, they would still have the knowledge in their bran from it.

If they were to actually lose the lawsuit, they likely would have to “cleanroom” the design, with different employees.

But in a way they don't really have to compete with either M2 or M3. The chips in Android phones are somewhat consistently 2 years behind Apple and I don't think any Android user will switch to iPhone purely because of this. Likewise, even if A chips were slower then Qualcomm and Samsung, I wouldn't jump to Android. I think much the same is true for a large portion of the PC/Mac space.

Sure, they don’t have to compete. But the hype train says they are going to trounce Apple. And they’ve made vague remarks about beating apple, too. So good luck with that.

 

[Cmaier]

Also the tendencies have switched! They're now saying Apple should acquire AMD, because the new mobile Ryzen chips can supposedly compete with M1 Pro!

Apple already acquired many of the chip design team from AMD, who I presume are happy to now be working on mobile products that don’t require jet cooling and a fusion-reactor for a battery.

 

[casperes1996]

If they were to actually lose the lawsuit, they likely would have to “cleanroom” the design, with different employees.

Ouch. That would be very bad for their work. I assume then the team would be put on some "You cannot work on ARM based chips for x months at Qualcomm" timeout thing? (eloquently phrased, I know)

Apple already acquired many of the chip design team from AMD, who I presume are happy to now be working on mobile products that don’t require jet cooling and a fusion-reactor for a battery.

Yeah; Just a sentiment I'm seeing more and more at the other place - I'm sure it'll fade once the 7x4x Ryzen mobile chips have been out a while

 

[jbailey]

That's more than 40% faster than M1 Pro, in a single generation. The M2 Max hasn't improved by that much relative to the M1 Max... (~25% if the benchmarks are true). May have something to do with what we've been talking about Geekbench compute being too short?

I'm kind of fascinated with the idea that the M2 was a bad update but the M2 Pro is good. I see a 48% improvement from the M1 to the M2 on Metal compute. My M1 8-core GPU MacBook Air gets about 20700 and the M2 10-core GPU MacBook Air gets about 30600. Single core CPU is up about 12% and multi core is up about 19%. These numbers compare pretty well with the M2 Pro upgrade but were considered underwhelming.

 

[casperes1996]

I'm kind of fascinated with the idea that the M2 was a bad update but the M2 Pro is good. I see a 48% improvement from the M1 to the M2 on Metal compute. My M1 8-core GPU MacBook Air gets about 20700 and the M2 10-core GPU MacBook Air gets about 30600. Single core CPU is up about 12% and multi core is up about 19%. These numbers compare pretty well with the M2 Pro upgrade but were considered underwhelming.

I thought the same. I'm assuming it's just because there was a bigger jump from Intel MacBook Air the M1 MacBook Air in experience than from 16->16 Intel vs. M1 Pro, so people wanted a bigger jump from M1 to M2 Air but a more modest jump on Pro->Pro is acceptable? It may also be that people didn't expect the two extra efficiency cores in Pro and Max and just the optics of the silicon having had "more work" changes peoples' opinions? To some extend I think the rhetorics that "M2 is a stopgap solution" has just spread so much that it gets regurgitated without much consideration or understanding of chip development

 

[Jimmyjames]

I'm kind of fascinated with the idea that the M2 was a bad update but the M2 Pro is good. I see a 48% improvement from the M1 to the M2 on Metal compute. My M1 8-core GPU MacBook Air gets about 20700 and the M2 10-core GPU MacBook Air gets about 30600. Single core CPU is up about 12% and multi core is up about 19%. These numbers compare pretty well with the M2 Pro upgrade but were considered underwhelming.

Yeah, It's a couple of things I think.

The initial M1 release was such a shock to many, especially in terms of single core improvement that it set expectations really high. Of course it's completely unrealistic to expect generational leaps with every iteration, but many seem to expect that, or pretend to, especially if they have an axe to grind (aka many macrumors posters).

I do think that in terms of raw power, Intel and AMD have responded quite well. Obviously they are nowhere near the efficiency of ASi, but the "this is the fastest" is an easy sell to many.

So the M2 came at an awkward time. The M1 came out and was pretty much the fastest (single core), then x86 chips quickly caught up in terms of raw speed (at a cost) and the M2 was a reasonable, but not earth shattering (which is to be expected in many ways). To a casual observer, or a motivated one, it's easy to come to the conclusion that the M2 was a disappointing update.

I have a feeling that the M3 will be a much more impressive update, but we'll see.

 

[tomO2013]

I'm reminded again with the M2 Pro / Max why I moved to an ARM laptop.

I recently did a transatlantic flight. The power socket did not work in my seat so needed to use battery only.
As I was working on a deadline , there were numerous apps open (XCode, VSCode, Docker) .... (I also had photoshop open where I was hobby playing around with stuff between (very quick) compiles).
I ended the 8 hour flight with battery remaining.

Just Amazing.
Today I'm back to the office and my Dell Precision 5570. I'm reminded that on battery my Dell and MS Visual Studio performs like a dog and the battery drains faster than Cinebench threads sprout up over at the other place.

I wish that Microsoft had a similarly powered officially supported ARM laptop with ooomph and battery life!

 

[casperes1996]

I wish that Microsoft had a similarly powered officially supported ARM laptop with ooomph and battery life!

You probably won't see equivalent power for quite some time to come but their recent dev-kit does indicate an interest in taking their ARM based platforms further as far as I can see

 

[Nycturne]

I'm reminded again with the M2 Pro / Max why I moved to an ARM laptop.

Agreed. I have an M1 Max 16" for myself, and my employer provided us with 2019 16" MBPs for development. I am reminded pretty much daily that I'd rather be spending 8 hours on the M1 in Xcode and the all the tools than I would be spending 3-4 hours on the 2019 in the same toolset.

You probably won't see equivalent power for quite some time to come but their recent dev-kit does indicate an interest in taking their ARM based platforms further as far as I can see

The problem is that I can't tell if the dev kit is a hobby or not. And if third party devs can't really tell, they won't invest, and the whole thing just kinda rots for a while. Microsoft has a history of dabbling and then just going "Oh, the market didn't create a bunch of demand for this with us just dabbling in it? Guess it's time to axe it."

 

[casperes1996]

The problem is that I can't tell if the dev kit is a hobby or not. And if third party devs can't really tell, they won't invest, and the whole thing just kinda rots for a while. Microsoft has a history of dabbling and then just going "Oh, the market didn't create a bunch of demand for this with us just dabbling in it? Guess it's time to axe it."

True. Though I do see this dev kit as more serious than prior dabbling. In part because they've actually, finally, made ARM native versions of Visual Studio (not code) and other tools that previously saw no love. But time will tell.