I don't think that there will be an App that uses the M1 GPU, now that Apple had dropped even the silent support for OpenCL with the last update to MacOS 11.4. There are just too few M1 Macs on E@H that justify a development just for them.
How much are the LATeah1083 WUs slower than the LATeah1082?
It is also in my observation that on my 2010 Intel MBP, the FRGP5 app is even slower churning out WU (around 23h on LATeah1082 awarding 693 credits per WU) as opposed to GW app (averaging 16h awarding 1000 per WU). I wouldn’t even wanna try LATeah1083/4 dataset on my dinosaur.
I'm biased, I admit. But not developing for Apple Silicon Macs seems short sighted. Apple has been releasing yearly updated Apple Silicon since the A4, and are now rolling them into the Mac. The only Apple Silicon Macs released so far are the low power versions. The high power versions are coming starting this month. Even the high power versions will be much more energy efficient than the intel/amd + gpu computers. I am probably not alone in that I am judicious in choosing to crunch Einstein units because of the high cost of electricity. I think my high power Apple Silicon Mac along with a growing number of other enthusiasts will be able to more affordably further Einstein searches if the gpu software is provided.
I'm biased, I admit. But not developing for Apple Silicon Macs seems short sighted. Apple has been releasing yearly updated Apple Silicon since the A4, and are now rolling them into the Mac. The only Apple Silicon Macs released so far are the low power versions. The high power versions are coming starting this month. Even the high power versions will be much more energy efficient than the intel/amd + gpu computers. I am probably not alone in that I am judicious in choosing to crunch Einstein units because of the high cost of electricity. I think my high power Apple Silicon Mac along with a growing number of other enthusiasts will be able to more affordably further Einstein searches if the gpu software is provided.
you misunderstood. Einstein does have an app that uses the M1 CPU.
Bernd said they would not have an app for the M1 "GPU". which does not support OpenCL. Einstein's apps are all opencl based and they are not going to expend the necessary resources to learn/test/deploy an app in Metal for a handful of users at best. and as far as I know, no other BOINC project supports their GPU either. be mad at Apple for dropping OpenCL support, not at projects who dont want to use Apple's Metal.
you misunderstood. Einstein does have an app that uses the M1 CPU.
Bernd said they would not have an app for the M1 "GPU". which does not support OpenCL. Einstein's apps are all opencl based and they are not going to expend the necessary resources to learn/test/deploy an app in Metal for a handful of users at best. and as far as I know, no other BOINC project supports their GPU either. be mad at Apple for dropping OpenCL support, not at projects who dont want to use Apple's Metal.
Thanks for replying! I see I was not clear - I was speaking about gpu, not cpu.
Hey, I'm not mad at anybody! I just enjoy contributing to the einstein@home project. I've been doing it for years, and invested quite a bit into buying Macs and outfitting them with supported gpu's. However, my Mac towers have just been too expensive to run since I retired. I still fire one up now and then, and also use my iMac Pro periodically for einstein@home, but unfortunately the electricity costs are more than I can afford. The electric company not only puts surcharges on my usage, but also sends me a nasty letter about my using more electricity than 75% of my neighbors. (Of course I think the electric company should think I'm one of their best customers...)
As far as "handful of users at best" it's a chicken and egg situation - more and more people will be buying Macs using Apple Silicon, and less and less of them will be interested in supporting einstein@home and other BOINC projects. The CPU payoff is not interesting enough to use our Macs, whereas the GPU payoff is. Apple did drop OpenCL support - it's a dead end. IMO "write once, run everywhere" just supports a dismal future. Maybe einstein@home can ask Apple to support them in making a gpu app for the Mac? Apple likes to support causes.
Anyways, I will keep monitoring einstein@home and other BOINC projects, hoping to see that one day my Apple Silicon based Macs can make a good contribution.
I am also interested in this matter. A month ago I resigned from my old Win machine and completely migrated to MacBook Air (M1, 8GB RAM).
I set processing to 1/8 of processors and 50% of CPU to avoid thermal issues and I'm surprised that Gamma-ray pulsar search #5 1.12 tasks are completed in about 2h40m instead of over 8 hours as my old laptop did.
If something will change in supporting OpenCL or Metal in M1-GPU, please let me know. I'm opened for checking test tasks.
I am also interested. I have a new MacBook Pro with an M1 Max 64GB Ram. I keep the thing throttled down to 25% to keep the fans from spinning up too much. It still manages to complete a task in 2 ½ hours or so.
I would like to see what it would do if the GPU cores were used. I don't know how many MAC users are doing processing for the various projects but I expect that more of them will eventually transition to newer models with the Apple M# chips.
Refusing to support the GPUs on those chips seems like a waste of potential processing resources.
I am also interested. I have a new MacBook Pro with an M1 Max 64GB Ram. I keep the thing throttled down to 25% to keep the fans from spinning up too much. It still manages to complete a task in 2 ½ hours or so.
I would like to see what it would do if the GPU cores were used. I don't know how many MAC users are doing processing for the various projects but I expect that more of them will eventually transition to newer models with the Apple M# chips.
Refusing to support the GPUs on those chips seems like a waste of potential processing resources.
Just my two bits worth.
Thanks
Einstein isn't refusing to support it. they literally can't since Apple removed openCL support.
if you want to crunch on the M1 GPU, write to Apple that they should bring openCL support back.
Another vote for more
)
Another vote for more M1 apps...
The current app with the longer LATeah1083 WUs come with paltry credit awards compared to the 1082 variant...
I don't think that there will
)
I don't think that there will be an App that uses the M1 GPU, now that Apple had dropped even the silent support for OpenCL with the last update to MacOS 11.4. There are just too few M1 Macs on E@H that justify a development just for them.
How much are the LATeah1083 WUs slower than the LATeah1082?
BM
LATeah1082 ~ 1h
)
LATeah1082 ~ 1h 20m
LATeah1083 ~ 2h 53m
LATeah1084 ~ 2h 42m
It is also in my observation that on my 2010 Intel MBP, the FRGP5 app is even slower churning out WU (around 23h on LATeah1082 awarding 693 credits per WU) as opposed to GW app (averaging 16h awarding 1000 per WU). I wouldn’t even wanna try LATeah1083/4 dataset on my dinosaur.
I'm biased, I admit. But not
)
I'm biased, I admit. But not developing for Apple Silicon Macs seems short sighted. Apple has been releasing yearly updated Apple Silicon since the A4, and are now rolling them into the Mac. The only Apple Silicon Macs released so far are the low power versions. The high power versions are coming starting this month. Even the high power versions will be much more energy efficient than the intel/amd + gpu computers. I am probably not alone in that I am judicious in choosing to crunch Einstein units because of the high cost of electricity. I think my high power Apple Silicon Mac along with a growing number of other enthusiasts will be able to more affordably further Einstein searches if the gpu software is provided.
jd wrote: I'm biased, I
)
you misunderstood. Einstein does have an app that uses the M1 CPU.
Bernd said they would not have an app for the M1 "GPU". which does not support OpenCL. Einstein's apps are all opencl based and they are not going to expend the necessary resources to learn/test/deploy an app in Metal for a handful of users at best. and as far as I know, no other BOINC project supports their GPU either. be mad at Apple for dropping OpenCL support, not at projects who dont want to use Apple's Metal.
_________________________________________________________________________
Ian&Steve C. wrote: you
)
Thanks for replying! I see I was not clear - I was speaking about gpu, not cpu.
Hey, I'm not mad at anybody! I just enjoy contributing to the einstein@home project. I've been doing it for years, and invested quite a bit into buying Macs and outfitting them with supported gpu's. However, my Mac towers have just been too expensive to run since I retired. I still fire one up now and then, and also use my iMac Pro periodically for einstein@home, but unfortunately the electricity costs are more than I can afford. The electric company not only puts surcharges on my usage, but also sends me a nasty letter about my using more electricity than 75% of my neighbors. (Of course I think the electric company should think I'm one of their best customers...)
As far as "handful of users at best" it's a chicken and egg situation - more and more people will be buying Macs using Apple Silicon, and less and less of them will be interested in supporting einstein@home and other BOINC projects. The CPU payoff is not interesting enough to use our Macs, whereas the GPU payoff is. Apple did drop OpenCL support - it's a dead end. IMO "write once, run everywhere" just supports a dismal future. Maybe einstein@home can ask Apple to support them in making a gpu app for the Mac? Apple likes to support causes.
Anyways, I will keep monitoring einstein@home and other BOINC projects, hoping to see that one day my Apple Silicon based Macs can make a good contribution.
Hello everyone, I am also
)
Hello everyone,
I am also interested in this matter. A month ago I resigned from my old Win machine and completely migrated to MacBook Air (M1, 8GB RAM).
I set processing to 1/8 of processors and 50% of CPU to avoid thermal issues and I'm surprised that Gamma-ray pulsar search #5 1.12 tasks are completed in about 2h40m instead of over 8 hours as my old laptop did.
If something will change in supporting OpenCL or Metal in M1-GPU, please let me know. I'm opened for checking test tasks.
Hi Everyone, I am also
)
Hi Everyone,
I am also interested. I have a new MacBook Pro with an M1 Max 64GB Ram. I keep the thing throttled down to 25% to keep the fans from spinning up too much. It still manages to complete a task in 2 ½ hours or so.
I would like to see what it would do if the GPU cores were used. I don't know how many MAC users are doing processing for the various projects but I expect that more of them will eventually transition to newer models with the Apple M# chips.
Refusing to support the GPUs on those chips seems like a waste of potential processing resources.
Just my two bits worth.
Thanks
Ted Lewis wrote: Hi
)
Einstein isn't refusing to support it. they literally can't since Apple removed openCL support.
if you want to crunch on the M1 GPU, write to Apple that they should bring openCL support back.
also the pool of mac users is very small.
_________________________________________________________________________
Hi, all. I'm even more
)
Hi, all.
I'm even more interested in this topic now that I've ordered a Mac Studio.
Perhaps the software needed from Apple can be obtained:
9to5mac says "Blender 3.1 update adds Metal GPU rendering for M1 Macs"
blender.org says "Cycles now has a Metal GPU backend, contributed by Apple."
https://www.blender.org/download/releases/3-1/
Could that be our missing link?
-- Dick