First M2 benchmarks

[EugW]

Mac14,7 - Geekbench

Benchmark results for a Mac14,7 with an Apple M2 processor.

browser.geekbench.com

M2 3.5 GHz

Single-core: 1919
Multi-core: 8928

I expect to see multi-core scores above 9000 by next week. With the 8K ProRes acceleration and 24 GB RAM, this is getting awfully close to the base M1 Pro.

 

[EugW]

Mac14,7 - Geekbench

Benchmark results for a Mac14,7 with an Apple M2 processor.

browser.geekbench.com

1927/8805

 

[leman]

Sounds about right. Not too bad for a passively cooled laptop either.

 

[Bodhitree]

Compared to the 1700 / 7200 scores the base M1 was posting, thats a respectable gain.

 

[EugW]

Sounds about right. Not too bad for a passively cooled laptop either.

My passively cooled m3 gets 1600 multi-core.

However, I suspect this M2 score is the 13" MacBook Pro.

 

[leman]

However, I suspect this M2 score is the 13" MacBook Pro.

Ah, yeah, good point.

 

[mi7chy]

Comparing with M1 it looks like difference in M2 single-core performance is from CPU frequency bump from 3.2GHz to 3.5GHz rather than IPC improvement while multi-core benefits from higher sustained all-core frequency on improved TSMC 5nm node. In summary, TSMC improved 5nm gains 9% in single-core and 16.6% in multi-core before thermal throttling that's yet to be tested.

M1
Single-core 1752 (3.5 / 3.2 = 1.09375 or 109.375%, 1752 * 1.09375 = 1916.25)
Multi-core 7654

M2
Single-core 1919
Multi-core 8928 (8928 / 7654 = 1.1664 or 116.6%)

 

[JimmyjamesEU]

Comparing with M1 it looks like difference in M2 single-core performance is from CPU frequency bump from 3.2GHz to 3.5GHz rather than IPC improvement while multi-core benefits from higher sustained all-core frequency on improved 5nm node. In summary, TSMC improved 5nm gains 9% in single-core and 16.6% in multi-core before thermal throttling that's yet to be tested.

M1
Single-core 1752 (3.5 / 3.2 = 1.09375 or 109.375%, 1752 * 1.09375 = 1916.25)
Multi-core 7654

M2
Single-core 1919
Multi-core 8928 (8928 / 7654 = 1.1664 or 116.6%)

They increased performance while increasing efficiency. Take a day off.

 

[EugW]

Ah, yeah, good point.

The Air should score about the same though, since Geekbench is such a short test.

That Mac14,7 name is curious though. It could just be that the listing doesn't play well with Geekbench, but FWIW:

M1 MacBook Air: MacBookAir10,1
M1 MacBook Pro: MacBookPro17,1
M1 Mac mini: Macmini9,1
M1 iMac: iMac21,1 & iMac21,2

 

[JimmyjamesEU]

The Air should score about the same though, since Geekbench is such a short test.

That Mac14,7 name is curious though. It could just be that the listing doesn't play well with Geekbench, but FWIW:

M1 MacBook Air: MacBookAir10,1
M1 MacBook Pro: MacBookPro17,1
M1 Mac mini: Macmini9,1
M1 iMac: iMac21,1 & iMac21,2

Yeah, it's curious. I've been keeping an eye on geekbench browser. Currently if you search for Mac14,7 nothing comes up. Perhaps it takes a while for the results to be integrated?

 

[Kazgarth]

9% overclock = 9% score increase.

 

[EugW]

Yeah, it's curious. I've been keeping an eye on geekbench browser. Currently if you search for Mac14,7 nothing comes up. Perhaps it takes a while for the results to be integrated?

Yeah, whenever new entries for new Macs appear in Geekbench, you can't search by the listed name initially. It takes a while to start working. BTW, if you search by "M2" here, that doesn't work either. The first M2 bench in this thread had already been posted elsewhere, but I found the second one just by manually sifting through the recent benches at the Geekbench website.

 

[JimmyjamesEU]

9% overclock = 9% score increase.

Can we put this nonsense to bed right now please. Performance increases typically don't happen without a reduction in efficiency or at bests flat efficiency. These cores (A15) managed both a ~10% perf increase and a 17% single core increase in efficiency. Per Anandtech.

 

[Kazgarth]

Can we put this nonsense to bed right now please. Performance increases typically don't happen without a reduction in efficiency or at bests flat efficiency. These cores (A15) managed both a ~10% perf increase and a 17% single core increase in efficiency. Per Anandtech.

View attachment 2019545

The efficiency part done by TSMC with second gen 5nm improvments

 

[JMacHack]

Better SC than the 12700 and an efficiency gain, not bad. Not terribly great but a decent uplift.

I hope Apple gets more aggressive with the higher end models though.

People on the forums are already whining about single core speed and its not topping the charts. And SC is constant throughout the range.

 

[JimmyjamesEU]

The efficiency part done by TSMC with second gen 5nm

With all due respect, I’m gonna trust Andrei (ex anandtech, now Nuvia) over you.

 

[Love-hate 🍏 relationship]

They increased performance while increasing efficiency. Take a day off.

what makes u think they increased efficiency

 

[prime17569]

The Air should score about the same though, since Geekbench is such a short test.

That Mac14,7 name is curious though. It could just be that the listing doesn't play well with Geekbench, but FWIW:

M1 MacBook Air: MacBookAir10,1
M1 MacBook Pro: MacBookPro17,1
M1 Mac mini: Macmini9,1
M1 iMac: iMac21,1 & iMac21,2

To add, the Mac Studio is Mac13,1 (M1 Max) and Mac13,2 (M1 Ultra).

This twitter post suggests that Apple might be following a new model name format:

https://twitter.com/i/web/status/1534296010020855808

 

[Kazgarth]

With all due respect, I’m gonna trust Andrei (ex anandtech, now Nuvia) over you.

That's not me, that's TSMC's roadmap based on data published officially by the company itself.

 

[JimmyjamesEU]

what makes u think they increased efficiency

The m2 is a15 cores. We know these characteristics.

 

[leman]

Can we put this nonsense to bed right now please. Performance increases typically don't happen without a reduction in efficiency or at bests flat efficiency. These cores (A15) managed both a ~10% perf increase and a 17% single core increase in efficiency. Per Anandtech.

I don't see any issues with what @Kazgarth said? The point is that we see no IPC improvements, which makes sense given that we are dealign with Avalanche cores. Whether the efficiency has changed remains to be seen. Personally, I doubt that the power consumption of M2 CPU cores will be higher than that of M1 CPU cores.

As to the rest, we will likely see bigger improvements on workloads that benefit from the larger M2 caches, but Geekbench won't reflect that.

 

[leman]

what makes u think they increased efficiency

Higher performance without increasing power consumption = higher efficiency.

 

[Mr.Blacky]

Geekbench is nice to quickly see and compare CPU performance, but there are quite a few bull*** results.

Geekbench 5 Results - Geekbench

 

[JimmyjamesEU]

That's not me, that's TSMC's roadmap based on data published officially by the company itself.

Sure, but the framing is that it’s not that good. They just increased frequency! Great. I wonder what the other cpu designers are doing? I see a great deal of cpus increasing both perf and energy. Guess they must be dumb when they could have “just” increased clock speed and gotten a corresponding increase in efficiency.

 

[satcomer]

Well according to multiple sites that Geekbench is still INTEL Software in 2022! It's like these developers are living in 2016! No wonder it didn't score high on Arm system!