Native Apple Silicon / M1|M2|M3 Apps available

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250362713
RAC: 35188

Sorry, had to retract that

Sorry, had to retract that app version for further analysis. Results might be wrong.

BM

L. John Keller II
L. John Keller II
Joined: 12 Nov 20
Posts: 2
Credit: 4619193
RAC: 10596

Running on my Apple M1 iMac.

Running on my Apple M1 iMac, and it seems to be working perfectly. However, since installing the MacOS Sonoma 14.0 I have been unable to run the screensaver. Does anyone have a work around to run the Einstein at Home screensaver on MacOS 14?

jd
jd
Joined: 13 Mar 05
Posts: 36
Credit: 556804511
RAC: 83610

Hey all - Happy New

Hey all - Happy New Year!

I need some help with a puzzle. I have a Mac Studio Ultra M1 with 20 CPUs (4 efficiency, 16 regular) and 64 GB memory. I also have a MacBook Pro Max M1 with 10 CPUs (2 efficiency, 8 regular) with 32 GB memory. So on the Mac Studio, I run 8 boinc tasks and each takes about 2 hours 41 minutes, running each CPU at about 90%. My MacBook Pro runs 3 boinc tasks and each takes about 1 hour 9 minutes, running each CPU at about 100%. My Mac Studio Ultra will complete 3 boinc tasks concurrently in about 1 hour 30 minutes.

Both computers produce about 50,000 a day (100,000 total). So the puzzle is why does the Mac Ultra take so much longer than my MacBook Pro? I went through and tested the times on the Mac Studio Ultra for various CPU numbers, and for 1 CPU 1 hour, 2 CPUs 1 hour 16 minutes, and adding CPUs just increases the time for each one. (It turns out 8 CPUs is the most efficient.)

Any ideas or suggestions on why my MacBook Pro M1 Max beats my Mac Studio M1 Ultra? 

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 3060
Credit: 4961714353
RAC: 1396017

jd wrote: Hey all - Happy

jd wrote:

Hey all - Happy New Year!

I need some help with a puzzle. I have a Mac Studio Ultra M1 with 20 CPUs (4 efficiency, 16 regular) and 64 GB memory. I also have a MacBook Pro Max M1 with 10 CPUs (2 efficiency, 8 regular) with 32 GB memory. So on the Mac Studio, I run 8 boinc tasks and each takes about 2 hours 41 minutes, running each CPU at about 90%. My MacBook Pro runs 3 boinc tasks and each takes about 1 hour 9 minutes, running each CPU at about 100%. My Mac Studio Ultra will complete 3 boinc tasks concurrently in about 1 hour 30 minutes.

Both computers produce about 50,000 a day (100,000 total). So the puzzle is why does the Mac Ultra take so much longer than my MacBook Pro? I went through and tested the times on the Mac Studio Ultra for various CPU numbers, and for 1 CPU 1 hour, 2 CPUs 1 hour 16 minutes, and adding CPUs just increases the time for each one. (It turns out 8 CPUs is the most efficient.)

Any ideas or suggestions on why my MacBook Pro M1 Max beats my Mac Studio M1 Ultra? 

Hi JD!

I'll give this a try, but I know nothing about a Mac, Studio, Pro, or otherwise so I apologize in advance if I'm way off beat.

Your CPU processing speed, the GHz rating if you will, plays a BIG role as does the video display which I surmise is integrated into both of your CPUs in your Macs.  Then the next important thing is the speed of the memory of each Mac: i.e. is it running at 2133 MHz  or 3200 MHz, etc. in each computer.  Finally, I don't know what projects you are running on each of your computers.  Are they both running the same, or different projects?  Each individual project has multiple projects within themselves that have different sizes and length of run times required depending on what your computers specs are.  Since both computers are running Einstein as the only project, which project within Einstein are you running?  Or are you running multiple projects at the same time, that also plays a role in how well each computer does.

That was a long winded attempt to say I can't give you an answer unless you give us more information.  Sorry.

George

Proud member of the Old Farts Association

Sandro
Sandro
Joined: 15 Jul 23
Posts: 9
Credit: 28840579
RAC: 94613

George, I imagine that most

George, I imagine that most of us with Apple Silicone are running the ONLY M1-native Einstein application - Gamma-ray pulsar search #5 1.14. To my knowledge, Einstein is the only project within the entire BOINC system with an M1-native application. Everything else runs under emulation. Fun stuff. :) 

My M1 MacBook Pro at 50% CPU utilization takes 1h 54m per task. This is in line with your MacBook numbers, JD. No idea why your Studio Ultra would take 2.5x as long per task. 

You mentioned that adding CPUs increases the time it takes for an individual task to complete. Is it possible that a task isn't limited to a single processor? 

jd
jd
Joined: 13 Mar 05
Posts: 36
Credit: 556804511
RAC: 83610

GWGeorge007 wrote: Hi

GWGeorge007 wrote:

Hi JD!

I'll give this a try, but I know nothing about a Mac, Studio, Pro, or otherwise so I apologize in advance if I'm way off beat.

Your CPU processing speed, the GHz rating if you will, plays a BIG role as does the video display which I surmise is integrated into both of your CPUs in your Macs.  Then the next important thing is the speed of the memory of each Mac: i.e. is it running at 2133 MHz  or 3200 MHz, etc. in each computer.  Finally, I don't know what projects you are running on each of your computers.  Are they both running the same, or different projects?  Each individual project has multiple projects within themselves that have different sizes and length of run times required depending on what your computers specs are.  Since both computers are running Einstein as the only project, which project within Einstein are you running?  Or are you running multiple projects at the same time, that also plays a role in how well each computer does.

That was a long winded attempt to say I can't give you an answer unless you give us more information.  Sorry.

Sandro wrote:

George, I imagine that most of us with Apple Silicone are running the ONLY M1-native Einstein application - Gamma-ray pulsar search #5 1.14. To my knowledge, Einstein is the only project within the entire BOINC system with an M1-native application. Everything else runs under emulation. Fun stuff. :) 

My M1 MacBook Pro at 50% CPU utilization takes 1h 54m per task. This is in line with your MacBook numbers, JD. No idea why your Studio Ultra would take 2.5x as long per task. 

You mentioned that adding CPUs increases the time it takes for an individual task to complete. Is it possible that a task isn't limited to a single processor? 

Hey George and Sandro - thanks for looking at this and replying! 

So I was reviewing your comments and reviewing my einstein@home activity and .... to my surprise, the boinc app seemed to crash (never happened before), and when I restarted it, a few of the in progress tasks were missing and after a bit of time, I realized it was computing faster! The completion time change from 2 hours 40 minutes (or more per task) to 1 hour 11 minutes!!!! Ok, something fishy is going on here.... but I'm not looking at a gift horse in the mouth! Did something happen behind the scenes? It seems likely. Could it be that I had the only Mac Ultra running einstein@home and there was a configuration issue that has now been fixed????

George - I looked into the specs for the M1 Ultra and the M1 Max more closely. The Mac Studio Ultra is basically 2 MacBook Pro M1 Max chips interconnected magically together. The individual CPU cores are the same, just more in the Ultra. The Memory speed is the same, though the bandwidth in the Ultra is twice the Max. So I expect the CPU performance to be the same in both computers, it's just more cores in the Ultra. When I study the Activity Monitor, it certainly appears that 1 core is assigned to 1 task.  On both computers, boinc is only running einstein@home, which is because I'm interested in the einstein@home science, plus it runs natively on my Mac, which makes for a good performance/power load. Running Intel based apps on Apple Silicon limits performance. I had tried running multiple einstein@home projects, but the inefficiencies of running Intel projects eventually made me switch to only the native Apple Silicon project (Gamma-ray pulsar search #5 1.14).

Sandro - Nice to see that our MacBook times line up! I'm not sure about CPU sharing of einstein@home tasks, but when studying the Activity Monitor it sure looks like a task has it's own CPU.

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3945
Credit: 46604112642
RAC: 64211778

Sandro wrote: George, I

Sandro wrote:

George, I imagine that most of us with Apple Silicone are running the ONLY M1-native Einstein application - Gamma-ray pulsar search #5 1.14. To my knowledge, Einstein is the only project within the entire BOINC system with an M1-native application. Everything else runs under emulation. Fun stuff. :)

I thought PrimeGrid had some M1 apps? looking at their apps list, they have several Apple M apps : https://www.primegrid.com/apps.php. looks like their apps even use the GPU.

also Asteroids@home is close to releasing their Apple Silicon native app. but it is slower than the Rosetta emulated app since there are not the same kind of optimizations for Arm. they are working on a NEON optimized app however. keep an eye out for that.

_________________________________________________________________________

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250362713
RAC: 35188

We're testing a new Apple

We're testing a new Apple Silicon App versions for BRP4, both CPU and CPU/GPU. As validation against other BRP4 apps is pretty poor, we are testing this as a separate BOINC Application "Binary Radio Pulsar Search (Arecibo,GBT,arm64) (BRP4A)". Although the plan class is named "Apple_M-opencl" (so that it works with current clients), the code of the GPU version is actually our first Metal code.

BM

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250362713
RAC: 35188

Hm. The GPU version seems to

Hm. The GPU version seems to crash on older macOS versions. Working on it.

BM

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250362713
RAC: 35188

The GPU app requires macOS

The GPU app requires macOS Ventura. Set.

BM

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.