Intel vs Apple Silicon

Hi, just done comparisons on my current iMac 3.6Ghz 8-Core i9 32Gb vs a Mac Studio M1 Max 32Gb… Xojo 2022r3

Xojo application launch time: iMac: 2m 2s, Mac Studio: 1m 7s
Open project: iMac: 59s, Mac Studio: 1m 3s
Analyse project: iMac: 1m 55s, Mac Studio: 2m 9s
Save project: iMac: 22s, Mac Studio, 27s
Build project: iMac 4m 30s, Mac Studio 4m 37s

I guess we’re not seeing the benefits of Apple Silicon just yet then.

Thought I’d post this just in case (like me) you’re considering getting a more powerful Mac for Xojo use.

2 Likes

Do you use plugins?
if yes, do you use the Plugin Manager from Tim?

if yes, it can unpack plugins, so they load quicker on macOS.

1 Like

I do Christian :wink: Yours and others. No aware of Plugin Manager or what it does - will investigate.

Since new macOS likes to do malware checks on the dylibs of the plugins, unpacking them can help with the startup speed.

Having disabled plug-ins that aren’t used - this makes little (if any) difference to compile times. Just, as you say, application launch time.

Are your tests conducted with native apps on M1 ?

Contrary to what you posted, Geekbench shows a sensible speed gain on M1 side.

Yes, all native on the Mac Studio. I was expecting more, hence the post. Launching Xojo is twice as quick as the iMac, but once in Xojo performance is almost exactly the same as the iMac. Underwhelmed.

So, you got Mac Studio in two versions, one for Intel and the other for M1 ?

My experience is that the speed gain needs to be measured to be ‘seen’
To my brain, the experience of my new M2 machine versus my 2015 intel is very similar.
Buyers remorse

1 Like

Not quite sure what you mean there. Sorry. iMac i9 with Xojo installed vs new Mac Studio also with Xojo installed performing like for like tasks. Lot of money for no benefit. Tests prove in fact, that the i9 iMac still has the edge over the Mac Studio wih M1 Max cpu.

This really shouldn’t be a big surprise. The single core clock speeds are largely the same. Xojo is a single core app. Until CPUs make a quantum leap in core clock, or Xojo works better with multiple cores, there’s really not much performance to gain.

5 Likes

If for some reason an app is using emulation, then it would make sense to see slower results with M1.

Yet, Geekbench shows definitely more speed on M1 side. When I bought my M1 Mac Mini, I compared with the iMac 27" i7: it was twice as fast.

As Thom says, it really depends on how well an application can take advantage of multiple cores. I’ve been banging on about it for some time as that’s where we’re seeing performance gains on the Mac.

On Wintel, it appears that AMD and Intel are still pushing for single core performance increases and I’ve read good things about Intel’s latest chips. But of course it goes without saying, that Intel remains power hungry processors.

The lack of noise and that the ARM chips don’t make the computers hot alone is very nice. I’ve been doing fractals yesterday. The old Intel MacBook Air would get hot and loud. The M1 Air isn’t bothered by anything.

4 Likes

Not really, according to Geekbench Browser the i9 has a Single Core Speed of 1222 and the M1 Max of 1791 (latest uploaded test results).
There must be some other reasons for the result of David, maybe different macOS or Xcode versions, running background tasks, connected devices etc.

1 Like

What about a panel of other applications ?

Your results can be correct for Xojo, but as Thom says earlier…

A lot of us have been saying this for a long time.

Unfortunately, Xojo’s solution doesn’t really allow you to take advantage of multiple cores and doesn’t work for mobile targets.

1 Like

What is Xojo doing after a the progress bar for Cmd-K is finished? For my project it takes about 25 - 30 seconds to compile. Then Xojo meditates with the beachball for 40 - 45 seconds:

CompileFeedbackList.MakeFeedbackTab%A1o%ooA1oo (in Xojo) + 5992 [0x1084219c0]
+ 2208 StudioMainWindow.StudioMainWindow.ShowCompileErrors%%o<StudioMainWindow.StudioMainWindow>A1ob (in Xojo) + 136 [0x10904e1f4]
+ 2208 StudioResultsPane.StudioResultsPane.AttachErrorList%%o<StudioResultsPane.StudioResultsPane>A1ob (in Xojo) + 136 [0x10c778884]
+ 2193 ErrorPane.ErrorPane.AddErrorList%%o<ErrorPane.ErrorPane>A1ob (in Xojo) + 752 [0x10c7afc50]
+ ! 2174 CompileFeedback.SearchResult%o%oo (in Xojo) + 4120 [0x107aaa80c]
+ ! : 1731 RBProject.FindSourceUnitOwner%o%oo (in Xojo) + 536 [0x105511a7c]
+ ! : | 429 RBWindow.Contains%b%oo (in Xojo) + 272 [0x10b422324]
+ ! : | + 374 RBWindow.GetAllChildren%%oA1ob (in Xojo) + 524 [0x10b442950]
+ ! : | + ! 205 RBControl.GetAllChildren%%oA1ob (in Xojo) + 52 [0x1051130e4]
+ ! : | + ! : 187 RuntimeStackCheck (in XojoFramework) + 92 [0x111529424]
+ ! : | + ! : | 177 StackContext::framePtr() const (in XojoFramework) + 44 [0x11156ab5c]
+ ! : | + ! : | + 172 unw_init_local (in libunwind.dylib) + 204 [0x1bf973d44]
+ ! : | + ! : | + ! 169 libunwind::UnwindCursor<libunwind::LocalAddressSpace, libunwind::Registers_arm64>::setInfoBasedOnIPRegister(bool) (in libunwind.dylib) + 124 [0x1bf973dd0]
+ ! : | + ! : | + ! : 77 dyld4::APIs::_dyld_find_unwind_sections(void*, dyld_unwind_sections*) (in dyld) + 172 [0x1b409a778]
+ ! : | + ! : | + ! : | 75 dyld3::MachOLoaded::findSectionContent(char const*, char const*, unsigned long long&, bool) const (in dyld) + 136 [0x1b40b89ec]
+ ! : | + ! : | + ! : | + 70 dyld3::MachOFile::forEachSection(void (dyld3::MachOFile::SectionInfo const&, bool, bool&) block_pointer) const (in dyld) + 192 [0x1b40b2548]
+ ! : | + ! : | + ! : | + ! 69 dyld3::MachOFile::forEachLoadCommand(Diagnostics&, void (load_command con

I’m sure that app’s that can use multiple cores would be much faster. Sadly I don’t have the time to test other apps, I just wanted to highlight my findings using Xojo and make others aware.

1 Like

I understand.

That said, I have a project who process files from a master folder and produce an html code in a TextArea (a Desktop).
For processing nearly 500 sub folders, my 2014 i5 is faster than my 2020 (2021 ?) m1 MacBook Pro.
OK, two or three more seconds on m1: I do not really care, but yes Xojo on m1 is not that fast than Intel CPUs… (on a < 10 seconds whole process time).
Once I set arm compile in the IDE, I had the feeling m1 compile is faster, but the difference (even 2 seconds) is ridiculously small.

I have to say, nearly 40 years ago, I have seen the difference between standard code and optimized code, on an Apple IIGS (GS/OS 4 vs GS/OS 5 on the same machine)… It was tremendous; IMHO worth the time spent. And it was not only from the boot time, usage experience was faster too !

1 Like