Apple: M1 vs. M2

[dada_dave]

Hmmm. I have quite bullish on Apple Silicon and I'm not changing my mind but...

these are a little disappointing. I mean they are fine, even good for laptops but I'm not sure this is gonna cut it for desktops. Yes Intel/AMD gobble power, but they are iterating and producing really good SC perf. Nvidia is really delivering huge amounts of compute with the 4090. I don't know if Apple can match that tbh. I just saw a test where the 4090 is delivering 300+fps 4k av1 or hevc encode. That's incredible. The M1 Max up until recently was the best for this that I found and it gets around 100fps. Nvidia have just demolished Apple's encoders. I wonder in my more doubtful moments if this isn't a repeat of 2013: Apple making the wrong gpu bet.

As a fan of desktops I'm a little concerned. We just aren't seeing any kind of progress. I was thinking earlier, since 2012 we've seen 4 reasonably powerful desktop macs. I don't count the mini or the iMacs. The 2013 Mac Pro (trash can), the 2017(?) iMac Pro, 2019 Mac Pro and 2022 Studio.

In 10 years, 4 desktops. That's not good enough. It really seems like there is no appetite for desktops within Apple, and no consistent endeavour. Pros want consistency and commitment to a platform. I just don't see that at the moment when it comes to desktops.

I would be interested if any gpu results leaked also. Haven't seen any.

(Yes this is a rambling post, but I'm struggling to keep the faith and these numbers do not help)

These leaks may not be real or representative of the final product. It’s important to caveat that. And if they are real they could indicate that the Pro/Max is on the same N4 node as the regular M2. Which would be unfortunate but not dire. However, things are moving more slowly than I’d hoped though and indeed more slowly than Apple had forecasted.

 

[Jimmyjames]

These leaks may not be real or representative of the final product. It’s important to caveat that. And if they are real they could indicate that the Pro/Max is on the same N4 node as the regular M2. Which would be unfortunate but not dire. However, things are moving more slowly than I’d hoped though and indeed more slowly than Apple had forecasted.

All correct.

While it’s not necessarily their fault, it might be their problem. As the only maker of Mac hardware, they have an obligation to produce a range of computers, including desktops.

In order to convince people to stay within the ecosystem, and that Apple Silicon is worth believing in, they should be massively outperforming x86. I felt that was the case when the M1 arrived, for laptops at least. I thought great desktops would follow. I don’t feel their offering is that impressive now. Certainly in pure performance. I also am losing faith in their ability to iterate at the required rate.

 

[dada_dave]

All correct.

While it’s not necessarily their fault, it might be their problem. As the only maker of Mac hardware, they have an obligation to produce a range of computers, including desktops.

In order to convince people to stay within the ecosystem, and that Apple Silicon is worth believing in, they should be massively outperforming x86. I felt that was the case when the M1 arrived, for laptops at least. I thought great desktops would follow. I don’t feel their offering is that impressive now. Certainly in pure performance. I also am losing faith in their ability to iterate at the required rate.

That’s fair.

 

[theorist9]

All correct.

While it’s not necessarily their fault, it might be their problem. As the only maker of Mac hardware, they have an obligation to produce a range of computers, including desktops.

In order to convince people to stay within the ecosystem, and that Apple Silicon is worth believing in, they should be massively outperforming x86. I felt that was the case when the M1 arrived, for laptops at least. I thought great desktops would follow. I don’t feel their offering is that impressive now. Certainly in pure performance. I also am losing faith in their ability to iterate at the required rate.

I think we need to wait for M3 to assess their progress. We don't have enough info. yet.

Like you, I'm also a fan of desktops. I'm most affected by SC CPU performance, and there the Mac desktops have thus far, on average, compared favorably to high-end PC's, particularly when you factor in their efficiency and its benefits (esp. quiet operation).

For instance, when the M2 was released (granted, it's not yet in a desktop), its GB5 SC score was within 10% of Intel's fastest enthusiast desktop chip, the i9-12900KS, which had been released just two months before. [According to GB's benchmark charts, the M2 (released June 2022) and i9-12900KS (released April 2022) have GB5 scores of 1899 and 2083, respectively.] Getting that close in a much quieter and more efficient machine seems like a win. Having said that, I would like to see Apple push SC speeds somewhat more for its desktop machines, if they can do it while maintaining quiet operation.

Where we see much more of a gap, particularly at the high-end, is in GPU performance. I assume they're working actively on that, but we shall see.

 

[Jimmyjames]

I think we need to wait for M3 to assess their progress. We don't have enough info. yet.

Like you, I'm also a fan of desktops. I'm most affected by SC CPU performance, and there the Mac desktops have thus far, on average, compared favorably to high-end PC's, particularly when you factor in their efficiency and its benefits (esp. quiet operation).

For instance, when the M2 was released, its GB5 SC score was within 10% of Intel's fastest enthusiast desktop chip, the i9-12900KS, which had been released just two months before. [According to GB's benchmark charts, the M2 (released June 2022) and i9-12900KS (released April 2022) have GB5 scores of 1899 and 2083, respectively.] Getting that close in a much quieter and more efficient machine seems like a win. Having said that, I would like to see Apple push SC speeds somewhat more for its desktop machines, if they can do it while maintaining quiet operation.

Where we see much more of a gap, particularly at the high-end, is in GPU performance. I assume they're working actively on that, but we shall see.

Totally fair points.

 

[Yoused]

The more I look at those numbers, the wronger they look. Single core should absolutely not be ~2% lower than base M2, and the claim is 12 core, which makes no mathematical sense at all for the score, even in it is 8P + 4E. No way would Apple be releasing a device that is so weak.

 

[Cmaier]

The more I look at those numbers, the wronger they look. Single core should absolutely not be ~2% lower than base M2, and the claim is 12 core, which makes no mathematical sense at all for the score, even in it is 8P + 4E. No way would Apple be releasing a device that is so weak.

hard to know what’s going on, even assuming the number is real. Could be a pre-production chip, could be firmware is set to debug, could be that not all cores are enabled. who knows.

 

[theorist9]

What puzzles me is why these results (if real) were uploaded at all. I can understand someone wanting to run GB, but what do they get out of taking the extra step to upload it to GB's site (other than possible exposure and termination)? If they're leaking to a person, they can potentially get favors or attention. But an anonymous upload seems curious--especially in this case, when there is nothing notable about the result.

 

[dada_dave]

What puzzles me is why these results (if real) were uploaded at all. I can understand someone wanting to run GB, but what do they get out of taking the extra step to upload it to GB's site (other than possible exposure and termination)? If they're leaking to a person, they can potentially get favors or attention. But an anonymous upload seems curious--especially in this case, when there is nothing notable about the result.

Often times it is supposedly by accident. I’ve never run it myself but I’ve been given to understand that uploading results is a simple button click (or it’s on by default?) and thus a lot of the pre-release results are by someone who screwed up.* That’s what I’ve read elsewhere.

*Provided that they are genuine of course, which as you mentioned as well they not always are.

 

[Andropov]

The more I look at those numbers, the wronger they look. Single core should absolutely not be ~2% lower than base M2, and the claim is 12 core, which makes no mathematical sense at all for the score, even in it is 8P + 4E. No way would Apple be releasing a device that is so weak.

Multicore crypto score is half of the M1 Pro/Max. Something is definitely wrong. Integer and FP multicores were up ~14 and 19% each IIRC.

 

[theorist9]

Often times it is supposedly by accident. I’ve never run it myself but I’ve been given to understand that uploading results is a simple button click (or it’s on by default?) and thus a lot of the pre-release results are by someone who screwed up.* That’s what I’ve read elsewhere.

*Provided that they are genuine of course, which as you mentioned as well they not always are.

Ah, that's it, the scores run with the free version are uploaded automatically, which would explain it. I thought you needed to do an extra step to upload them, but that's not the case. To avoid the automatic upload, you need to get the paid version.

 

[exoticspice1]

The performance gains are quite reasonable, assuming this isn’t N3. If they don’t jump in performance when they get to N3, that would indicate a problem.

How is this performance reasonable? M2 Max bumped up the clock speeds to 3.54Ghz. so the improvements are mainly from that plus the additional 2 e cores.

M1 Max was on 3.2Ghz. hopefully they get sorted by 3nm M3.

 

[leman]

I am a bit puzzled that the single-core scores for these "M2 Max" (of which there are now several) are consistently slightly lower than the base M2 although the frequency is slightly higher? If these are all from the same machine, it could be that the final scores will be better. GB5 of 1950/15000 would be not too shabby for a compact laptop.

 

[Cmaier]

I am a bit puzzled that the single-core scores for these "M2 Max" (of which there are now several) are consistently slightly lower than the base M2 although the frequency is slightly higher? If these are all from the same machine, it could be that the final scores will be better. GB5 of 1950/15000 would be not too shabby for a compact laptop.

it’s not unusual for pre-production scores to be a little lower than the final product. Often there are things that need to be adjusted in firmware or the OS to deal with the new processor, or the particular chip has known bugs that are being worked around, etc. I worked on a chip that didn’t have one of its caches enabled unless the “BIOS” was changed, for example.

In the end, if this is the real chip, it means it was fabbed on the same process as M2, and we will see performance comparable to M2 for SC and scaling in MC based on the number of cores. Of course, there is an additional scale factor if they choose to actually ship with an increased clock, but we don’t really know what they’ll do.

 

[Yoused]

Someone on ars observed that that reported model number seemed off

… Also, Mac 14,6 is an unexpected model number: there’s already a Mac 14,7 (the 13” M2 MacBook Pro), and the existing Mac Studio models are designated Mac 13,1 and 13,2.

 

[theorist9]

Someone on ars observed that that reported model number seemed off

Well, FWIW, there is this guy, who claims to have found these model numbers in Apple code. Though it's not much better-supported than the GB result:

Apple is preparing three new M2 Macs [u] | AppleInsider

References to three new M2-based Macs, along with model identifiers and even code names have reportedly been found within unspecified Apple code.

appleinsider.com

 

[Yoused]

well, if 14,7 is the base M2, then at least 2 of those numbers appear to be irrelevant

 

[theorist9]

Here's another one the same guy ("iro") posted a couple hours later. Not that we should believe this one either, but it doesn't have the issue with the MC crypto score that @Andropov mentioned.

I checked iro's GB posting history, and they did post what appear to be legit scores for the M1 Ultra on March 8, which is 10 days before it was released (https://browser.geekbench.com/user/iro and https://browser.geekbench.com/v5/cpu/13345054); this may have been a review sample. But they also have some strange postings, like a 2019 Intel Mac Pro with an i9-13900KF (https://browser.geekbench.com/v5/cpu/18486658). Not sure what that means--that they can spoof postings, or that they also build Hackintoshes (?)

https://browser.geekbench.com/v5/cpu/18988586

 

[Buntschwalbe]

But they also have some strange postings, like a 2019 Intel Mac Pro with an i9-13900KF (https://browser.geekbench.com/v5/cpu/18486658). Not sure what that means--that they can spoof postings, or that they also build Hackintoshes (?)

The i9-13900KF is definitely a Hackintosh, as is stated under Motherboard. Acidanthera is the project from the builders of OpenCore.

 

[Andropov]

Here's another one the same guy ("iro") posted a couple hours later. Not that we should believe this one either, but it doesn't have the issue with the MC crypto score that @Andropov mentioned.

I checked iro's GB posting history, and they did post what appear to be legit scores for the M1 Ultra on March 8, which is 10 days before it was released (https://browser.geekbench.com/user/iro and https://browser.geekbench.com/v5/cpu/13345054); this may have been a review sample. But they also have some strange postings, like a 2019 Intel Mac Pro with an i9-13900KF (https://browser.geekbench.com/v5/cpu/18486658). Not sure what that means--that they can spoof postings, or that they also build Hackintoshes (?)

https://browser.geekbench.com/v5/cpu/18988586

Yeah, the i9-13900KF is definitely a hackintosh.

The new benchmark (1889 Single Core and 14586 Multi Core) makes a lot more sense. It's interesting to see that (if true) they've gone with 4 E-cores this time. I like those cores a lot, they don't often get the praise they deserve.