Benchmarks : 2021 M1Max not as good as 2015 Core i7 for 68K/PPC emulation?

Anything about Mac emulation that does not belong in the above categories.

Moderators: Cat_7, Ronald P. Regensburg

Post Reply
yserra
Student Driver
Posts: 11
Joined: Sun Mar 15, 2020 8:14 am

Benchmarks : 2021 M1Max not as good as 2015 Core i7 for 68K/PPC emulation?

Post by yserra »

Hi all,
I tried a few benchmarks under BasiliskII and Sheepshaver, with the latest universal binaries on a 2021 MacBook Pro 14-inch M1 Max, and compared to a 2015 MacBook Pro 13-inch Core i7 with the same executables and the same MacOS 7/MacOS 9 images. Apparently, the results are really less good on the M1 Max (except for the FPU emulation under Sheepshaver).
Is it normal?
Is it something to be improved?
Did I make something wrong?
All this is really not important. I'm just wondering.

On Sheepshaver (PowerPC, JIT), with Norton System Info 6.0.5,

System Rating:
2015 MBP-13 Core i7 = 2405
2021 MBP-14 M1 Max = 596

CPU:
2015 MBP-13 Core i7 = 2449
2021 MBP-14 M1 Max = 566

Video:
2015 MBP-13 Core i7 = 1974
2021 MBP-14 M1 Max = 610

Disk:
2015 MBP-13 Core i7 = 6293
2021 MBP-14 M1 Max = 2193

FPU :
2015 MBP-13 Core i7 = 650
2021 MBP-14 M1 Max = 1498

On BasiliskII (68040, JIT), with Speedometer 4.02,

CPU:
2015 MBP-13 Core i7 = 61.94
2021 MBP-14 M1 Max = 24.41

Graf:
2015 MBP-13 Core i7 = 36.88
2021 MBP-14 M1 Max = 23.55

Disk:
2015 MBP-13 Core i7 = 5.57
2021 MBP-14 M1 Max = 7.00

Math:
2015 MBP-13 Core i7 = 1485.2
2021 MBP-14 M1 Max = 283.70

PR (?):
2015 MBP-13 Core i7 = 19.79
2021 MBP-14 M1 Max = 17.23
User avatar
Ronald P. Regensburg
Expert User
Posts: 7821
Joined: Thu Feb 09, 2006 10:24 pm
Location: Amsterdam, Netherlands

Re: Benchmarks : 2021 M1Max not as good as 2015 Core i7 for 68K/PPC emulation?

Post by Ronald P. Regensburg »

The M1 M1 Pro and M1 Max chips cannot be compared with the usual benchmark software, not with other processors and not with each other. That is because of the complex core allocation used with these chips. A recent post about the impossibility to really compare an M1 and an M1 Pro can be found here: https://eclecticlight.co/2021/11/26/how ... llocation/
The article also links to a previous article about this subject: https://eclecticlight.co/2021/11/17/how ... geekbench/

And BTW: On Apple Silicon JIT compiler is not available in BasiliskII. JIT settings will be ignored.
yserra
Student Driver
Posts: 11
Joined: Sun Mar 15, 2020 8:14 am

Re: Benchmarks : 2021 M1Max not as good as 2015 Core i7 for 68K/PPC emulation?

Post by yserra »

Thank you for this answer Ronald.
Should I understand speed of Sheepshaver/BasiliskII is not going to be increased by code in the future on M1x CPUs, since any ARM optimization would be broken by any Apple Silicon update?
User avatar
Ronald P. Regensburg
Expert User
Posts: 7821
Joined: Thu Feb 09, 2006 10:24 pm
Location: Amsterdam, Netherlands

Re: Benchmarks : 2021 M1Max not as good as 2015 Core i7 for 68K/PPC emulation?

Post by Ronald P. Regensburg »

I am sorry, but I do not understand your question. I am grateful that the developer (kanjitalk755) made BasiliskII and SheepShaver available on M1 at all. Maybe he can answer your question. Anyway, the future in PPC emulation is for QEMU, not SheepShaver.
yserra
Student Driver
Posts: 11
Joined: Sun Mar 15, 2020 8:14 am

Re: Benchmarks : 2021 M1Max not as good as 2015 Core i7 for 68K/PPC emulation?

Post by yserra »

Accept my apologies, english is not my native tongue.

My question was: can we expect any code optimization that could improve speed of BasiliskII/Sheepshaver on M1 Mac?

Or is this a stupid question since the BasiliskII/Sheepshaver code already provides the maximum performance that a M1/Pro/Max could provide for a 68K/a PPC emulation?
(Because, for example, it would not be possible to use multiple parallel cores of Apple Silicon to improve the emulation of a single-core CPU like the 68K or PPC. And even if a coder would find the magical trick to emulate a super fast 68K from the ten M1 Pro cores, his effort would be ruined as soon as Apple launch a new machine with a different number of cores...)
(yes, all this was what my previous question was about...).

Regarding QEMU, is there a plug-and-play product that is as simple as Sheepshaver? Last time I tried, it was a bit complicated to use it for emulating a Power Mac...
User avatar
Ronald P. Regensburg
Expert User
Posts: 7821
Joined: Thu Feb 09, 2006 10:24 pm
Location: Amsterdam, Netherlands

Re: Benchmarks : 2021 M1Max not as good as 2015 Core i7 for 68K/PPC emulation?

Post by Ronald P. Regensburg »

I do not know if there will be more optimisation in the future. As I understand, it also needs more documentation by Apple about how core allocation works on Apple Silicon and how it can be influenced.

Anyway, I see no need right now. Both BasiliskII and SheepShaver still feel faster on my M1 iMac than my old real hardware Macs back in those days.


QEMU is indeed not as easy to set up as SheepShaver, but it is the future for emulation.
yserra
Student Driver
Posts: 11
Joined: Sun Mar 15, 2020 8:14 am

Re: Benchmarks : 2021 M1Max not as good as 2015 Core i7 for 68K/PPC emulation?

Post by yserra »

Thank you for this answer.
User avatar
Ronald P. Regensburg
Expert User
Posts: 7821
Joined: Thu Feb 09, 2006 10:24 pm
Location: Amsterdam, Netherlands

Re: Benchmarks : 2021 M1Max not as good as 2015 Core i7 for 68K/PPC emulation?

Post by Ronald P. Regensburg »

If you are interested, here is another article about comparing M1 processor performances: https://eclecticlight.co/2021/12/01/com ... 3-p-and-e/
Post Reply