Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

bobcomer

macrumors 601
May 18, 2015
4,949
3,699
You’re mistaken because you’re comparing it to CPU/GPUs that output a lot more heat and therefore Intel/AMD/Nvidia implemented throttling features in cases where the cooling solution couldn’t remove the heat fast enough.
Nope. The M1 isn't magical, it can build up heat and have trouble dissipating it too if the load is higher than the ability to dissipate.

Physics baby what is that?
What happens to your M1 Mini when it's sitting in 80F air and you're trying to run a WOA VM and an Ubuntu VM? It throttles, big surprise there. It's only got one fan, and it's not very high speed.
 

turbineseaplane

macrumors P6
Mar 19, 2008
17,409
40,207
Hardware of the Apple chips might be great, but without a great developer ecosystem and tools and API’s, the comparison to any console platform are basically pointless.

There’s a lot more to a great gaming system than just hardware.
 

leman

macrumors Core
Oct 14, 2008
19,522
19,679
Anyway, here you go, a quick clip of Metro Exodus, a recent game

Metro Exodus is coming to Mac this month, maybe even with native ARM support (was not too clear to me). Would be interesting to look at the benchmarks then!
 
  • Like
Reactions: TrueBlou

leman

macrumors Core
Oct 14, 2008
19,522
19,679
That's exactly what I mean and am actually quite shocked that someone would think that's a bad thing for a processor/PC to do. I wouldn't buy a modern processor any other way without some VERY special hardware and cooling to support it.

Ah well, you know how folks are. It became almost fashionable to talk about Apple computers throttling. Technical considerations, “how things work” or even common sense - who cares about those. Every single person on this forum is a CPU engineer after all ;)
 
Last edited:

leman

macrumors Core
Oct 14, 2008
19,522
19,679
Hardware of the Apple chips might be great, but without a great developer ecosystem and tools and API’s, the comparison to any console platform are basically pointless.

There’s a lot more to a great gaming system than just hardware.

Very true, and that’s why Apple has been working on improving their ecosystem. The documentation is still crap, but tools/API seem to come along quite nicely.
 
  • Like
Reactions: JMacHack

bobcomer

macrumors 601
May 18, 2015
4,949
3,699
Ah well, you know how folks are. It became aldosterone fashionable to talk about Apple computers throttling. Technical considerations, “how things work” or even common sense - who cares about those. Every single person on this forum is a CPU engineer after all ;)
Well said, and I have to admit I'm old school and not a Mac user professionally (only personally), so I don't really pick up some of the nuance around here. I remember the days where CPU's didn't have the ability to throttle based on heat at all. :)
 

JMacHack

Suspended
Mar 16, 2017
1,965
2,424
Name one showcase game for M1? None. You're playing 10+ year old PC games at like 720p. Not even comparable to the old 2013 PS4.
10 year old pc games are honestly 100% better than current gen in everything but graphical fidelity.

oh no, I will miss out on the glitch ridden, released-in-beta, gambling mechanic-ridden crapfest that is modern gaming, woe be me for I am a Mac user!
 

ArPe

macrumors 65816
May 31, 2020
1,281
3,325
Nope. The M1 isn't magical, it can build up heat and have trouble dissipating it too if the load is higher than the ability to dissipate.


What happens to your M1 Mini when it's sitting in 80F air and you're trying to run a WOA VM and an Ubuntu VM? It throttles, big surprise there. It's only got one fan, and it's not very high speed.

Make a video of it then along with the live stats. BTW, if it throttles 100Mhz that doesn’t count.
 
  • Like
Reactions: jdb8167

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
Nope. The M1 isn't magical, it can build up heat and have trouble dissipating it too if the load is higher than the ability to dissipate.


What happens to your M1 Mini when it's sitting in 80F air and you're trying to run a WOA VM and an Ubuntu VM? It throttles, big surprise there. It's only got one fan, and it's not very high speed.
I don’t know what happens in that situation since I haven’t seen anyone test it. Have you?

The tests that I’ve seen show the M1 hitting its peak GHz and staying there. If there are situations where that isn’t true I’d like to see them documented. I can certainly show throttling on my M1 MacBook Air so testing shouldn’t be that hard. If I owned a M1 mini I would do the test but I don’t.
 
  • Like
Reactions: souko

bobcomer

macrumors 601
May 18, 2015
4,949
3,699
Make a video of it then along with the live stats. BTW, if it throttles 100Mhz that doesn’t count.
I don't even have an M1 Mini to take a video of -- it's how all M1's work, it's part of the chip...
 

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
I don't even have an M1 Mini to take a video of -- it's how all M1's work, it's part of the chip...
So are you just guessing that it throttles in the situation you described? Or are you saying that if it didn’t throttle the M1 would run faster than 3.2GHz. Running at its designed power and clock isn’t throttling in my opinion.
 
  • Like
Reactions: souko

bobcomer

macrumors 601
May 18, 2015
4,949
3,699
I don’t know what happens in that situation since I haven’t seen anyone test it. Have you?
I haven't tested with an M1 Mini specifically, but I have used loads like that with MANY different kinds of computers and OS's.
The tests that I’ve seen show the M1 hitting its peak GHz and staying there. If there are situations where that isn’t true I’d like to see them documented. I can certainly show throttling on my M1 MacBook Air so testing shouldn’t be that hard. If I owned a M1 mini I would do the test but I don’t.
I don't really see why it would be that important to you. We know the M1 can throttle based on heat, and there is absolutely nothing different about the Mini that would change that. We also know MacOS isn't a RT OS.
 

bobcomer

macrumors 601
May 18, 2015
4,949
3,699
So are you just guessing that it throttles in the situation you described? Or are you saying that if it didn’t throttle the M1 would run faster than 3.2GHz. Running at its designed power and clock isn’t throttling in my opinion.
Nope, it's not a guess -- it's designed to do that. No, I'm not saying it will run faster than 3.2Ghz. (and don't just look at the system clock, look at the usage of the cores themselves, especially the hp cores.)
 

pmiles

macrumors 6502a
Dec 12, 2013
812
678
It's rather all relative... considering that most computers are never really tapped to their full potential, it's like having and extra 200 horses that never even enter the race. Compare today to the days of the Moon landing... computing power then compared to now... seems to me we should be doing much more with them than surfing social media and playing video games.

Much like the human brain... 10% used 90% wasted.
 
  • Like
Reactions: bobcomer

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
Nope, it's not a guess -- it's designed to do that. No, I'm not saying it will run faster than 3.2Ghz. (and don't just look at the system clock, look at the usage of the cores themselves, especially the hp cores.)
I would need some evidence that it’s true. I’m sorry I can’t just accept a random post on a rumor site public forums as evidence. I haven’t seen any test that shows all-core maxed out along with the a maxed out GPU so I can’t be positive it never throttles. But I have seen evidence that just all-core doesn’t throttle.
 

bobcomer

macrumors 601
May 18, 2015
4,949
3,699
I would need some evidence that it’s true. I’m sorry I can’t just accept a random post on a rumor site public forums as evidence. I haven’t seen any test that shows all-core maxed out along with the a maxed out GPU so I can’t be positive it never throttles. But I have seen evidence that just all-core doesn’t throttle.
No skin off my back what you believe.
 

leman

macrumors Core
Oct 14, 2008
19,522
19,679
I would need some evidence that it’s true. I’m sorry I can’t just accept a random post on a rumor site public forums as evidence. I haven’t seen any test that shows all-core maxed out along with the a maxed out GPU so I can’t be positive it never throttles. But I have seen evidence that just all-core doesn’t throttle.

The evidence is very simple: first, the chip reaches 100C under load and then just "magically" hover's there, second, all tested M1's so far show surprisingly similar peak power usage figures. These things alone make it very clear that there is an arbitrary power, frequency, and/or thermal limit. Without these limits we'd see some M1's go past 3.2 ghz, but they see locked in at that frequency.
 

ArPe

macrumors 65816
May 31, 2020
1,281
3,325
The evidence is very simple: first, the chip reaches 100C under load and then just "magically" hover's there, second, all tested M1's so far show surprisingly similar peak power usage figures. These things alone make it very clear that there is an arbitrary power, frequency, and/or thermal limit. Without these limits we'd see some M1's go past 3.2 ghz, but they see locked in at that frequency.

He’s not going to listen and he doesn’t even have a machine to check. It takes 5 mins to check.
 

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
The evidence is very simple: first, the chip reaches 100C under load and then just "magically" hover's there, second, all tested M1's so far show surprisingly similar peak power usage figures. These things alone make it very clear that there is an arbitrary power, frequency, and/or thermal limit. Without these limits we'd see some M1's go past 3.2 ghz, but they see locked in at that frequency.
What part of that denotes throttling? It is just hitting design power and heat. The M1 MBA throttles when hitting that 100C but the mini and MBP ramp up their fans. Again, I haven’t seen any tests that document all-core and max GPU and I don’t know of any synthetic benchmark that does that easily so it is possible that the performance cores throttle on a mini in that situation. I have seen numerous Cinebench benchmarks that show all-core with test results showing no throttling.
 

leman

macrumors Core
Oct 14, 2008
19,522
19,679
What part of that denotes throttling? It is just hitting design power and heat. The M1 MBA throttles when hitting that 100C but the mini and MBP ramp up their fans. Again, I haven’t seen any tests that document all-core and max GPU and I don’t know of any synthetic benchmark that does that easily so it is possible that the performance cores throttle on a mini in that situation. I have seen numerous Cinebench benchmarks that show all-core with test results showing no throttling.

Well, here we again have the two definitions of "throttling". We mean "throttling" as in a deliberate mechanism employed mechanism to manage the power and performance. You seem to talk about "throttling" as in "not reaching the expected performance due to thermal limitations". The M1 performs as expected in almost every scenario I've seen so far, so the second definition of "throttling" does not apply. But Apple definitively employs throttling (in wider sense) to manage the chip.
 
  • Like
Reactions: ArPe and jdb8167

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
Well, here we again have the two definitions of "throttling". We mean "throttling" as in a deliberate mechanism employed mechanism to manage the power and performance. You seem to talk about "throttling" as in "not reaching the expected performance due to thermal limitations". The M1 performs as expected in almost every scenario I've seen so far, so the second definition of "throttling" does not apply. But Apple definitively employs throttling (in wider sense) to manage the chip.
We started out by asking if the M1 could match or exceed the PS4. The discussion devolved to talking about the M1 not being that impressive since it throttles. So I’m using the term as it was originally used during this discussion.
 

Chozes

macrumors member
Oct 27, 2016
75
97
In practice the consoles run circles around their equivalent PC and Mac counterparts.
 

Krevnik

macrumors 601
Sep 8, 2003
4,101
1,312
Nope. The M1 isn't magical, it can build up heat and have trouble dissipating it too if the load is higher than the ability to dissipate.

This I don’t necessarily disagree with. It’s clear the MBA’s passive heatsink will saturate given enough time under load and cause problems. But the evidence so far doesn’t demonstrate that the load is higher than the fan’s ability to dissipate it in the 13” MBP and Mini.

What happens to your M1 Mini when it's sitting in 80F air and you're trying to run a WOA VM and an Ubuntu VM? It throttles, big surprise there. It's only got one fan, and it's not very high speed.

The fan isn’t running at high speeds because it doesn’t need to run at high speed. It’s the same cooling solution as in the 2018 Mini. The difference here is that the M1’s total wall consumption is much lower than the 2018. Max 40W at the wall, versus 122W. Max 133BTU/h vs 417BTU/h. The i5/i7 Minis will stabilize at around 65W for the CPU alone (although I’ve seen an i5 go higher), giving us a rule of thumb for what the cooling can actually dissipate. And the M1 won’t even draw that much from the wall at max clock speeds, so there’s headroom to keep the M1 from thermal throttling.

I was running some encodes (not using VideoToolbox, so it was all on the CPU cores, not the accelerators) during the recent heat wave here in the PNW. M1 Mini was sat in an ambient temperature of around 105F (~40C). SoC sensors reported peaks of ~70C. A couple core sensors did get into the 80s, with others in the high 70s, but clock speeds remained maxed out without spinning up the fan. Spinning up the fan manually would have brought the temps down noticeably, but since it wasn’t thermally throttling in a heat wave, I didn’t bother. My i7 though… that is another story.

Now, when Apple shrinks the Mini again, and puts a higher end SoC in there, you’d probably have a point about that one. Apple’s going to consume that headroom to make things smaller. They certainly ate up the headroom to make the MBA passively cooled. But if anything the M1 Mini as it is today is “over engineered” for the power it consumes.

I haven't tested with an M1 Mini specifically, but I have used loads like that with MANY different kinds of computers and OS's.

Then at least look at the data to date on the thing you are commenting on, because all the reviews have been pointing out the low power consumption under load of the M1 compared to the i5/i7 it replaced in the Mini. Apple posted power consumption numbers for both models (which tend to be surprisingly conservative compared to real world measurements of those two models). Folks have done tests around clock speed and thermals. And none of it lines up with your claim above.

Well, here we again have the two definitions of "throttling". We mean "throttling" as in a deliberate mechanism employed mechanism to manage the power and performance. You seem to talk about "throttling" as in "not reaching the expected performance due to thermal limitations". The M1 performs as expected in almost every scenario I've seen so far, so the second definition of "throttling" does not apply. But Apple definitively employs throttling (in wider sense) to manage the chip.

Unfortunately, the common usage is the second definition for a forum like this one, while the first is mostly in certain engineering circles (not common even on my dev team for example). Based on bobcomer’s usage of the term, the second definition is being leaned into in passages like what I quote above. So it’s fair to continue the conversation in that context IMO.

——

Getting somewhat back on topic though. One thing that I am surprised hasn’t been brought up is how systems like the PS4 use tricks to gain performance that aren’t (yet) common in PC gaming, and are dealing with lower resolutions like 1080p compared to a Mac. Although that will change, and these tricks would help Mac gamers especially due to the high resolution displays Apple embeds in their laptops.

Internal scaling and dynamic resolution were pretty common on both PS4 and Xbox One IIRC. I remember discussion around the dynamic resolution tech for Halo 5 a while back for sure. Being able to render at something as low as 720p and upscale to 1080p or 4K for output is certainly going to be less stressful on a GPU than trying to set game settings to match the 2560-by-1600 of a 13” MBP or Air.
 
Last edited:
  • Like
Reactions: altaic and jdb8167

diamond.g

macrumors G4
Mar 20, 2007
11,437
2,665
OBX
This I don’t necessarily disagree with. It’s clear the MBA’s passive heatsink will saturate given enough time under load and cause problems. But the evidence so far doesn’t demonstrate that the load is higher than the fan’s ability to dissipate it in the 13” MBP and Mini.



The fan isn’t running at high speeds because it doesn’t need to run at high speed. It’s the same cooling solution as in the 2018 Mini. The difference here is that the M1’s total wall consumption is much lower than the 2018. Max 40W at the wall, versus 122W. Max 133BTU/h vs 417BTU/h. The i5/i7 Minis will stabilize at around 65W for the CPU alone (although I’ve seen an i5 go higher), giving us a rule of thumb for what the cooling can actually dissipate. And the M1 won’t even draw that much from the wall at max clock speeds, so there’s headroom to keep the M1 from thermal throttling.

I was running some encodes (not using VideoToolbox, so it was all on the CPU cores, not the accelerators) during the recent heat wave here in the PNW. M1 Mini was sat in an ambient temperature of around 105F (~40C). SoC sensors reported peaks of ~70C. A couple core sensors did get into the 80s, with others in the high 70s, but clock speeds remained maxed out without spinning up the fan. Spinning up the fan manually would have brought the temps down noticeably, but since it wasn’t thermally throttling in a heat wave, I didn’t bother. My i7 though… that is another story.

Now, when Apple shrinks the Mini again, and puts a higher end SoC in there, you’d probably have a point about that one. Apple’s going to consume that headroom to make things smaller. They certainly ate up the headroom to make the MBA passively cooled. But if anything the M1 Mini as it is today is “over engineered” for the power it consumes.



Then at least look at the data to date on the thing you are commenting on, because all the reviews have been pointing out the low power consumption under load of the M1 compared to the i5/i7 it replaced in the Mini. Apple posted power consumption numbers for both models (which tend to be surprisingly conservative compared to real world measurements of those two models). Folks have done tests around clock speed and thermals. And none of it lines up with your claim above.



Unfortunately, the common usage is the second definition for a forum like this one, while the first is mostly in certain engineering circles (not common even on my dev team for example). Based on bobcomer’s usage of the term, the second definition is being leaned into in passages like what I quote above. So it’s fair to continue the conversation in that context IMO.

——

Getting somewhat back on topic though. One thing that I am surprised hasn’t been brought up is how systems like the PS4 use tricks to gain performance that aren’t (yet) common in PC gaming, and are dealing with lower resolutions like 1080p compared to a Mac. Although that will change, and these tricks would help Mac gamers especially due to the high resolution displays Apple embeds in their laptops.

Internal scaling and dynamic resolution were pretty common on both PS4 and Xbox One IIRC. I remember discussion around the dynamic resolution tech for Halo 5 a while back for sure. Being able to render at something as low as 720p and upscale to 1080p or 4K for output is certainly going to be less stressful on a GPU than trying to set game settings to match the 2560-by-1600 of a 13” MBP or Air.
I've talked about image reconstruction on Apple hardware before. As it stands there is no native hardware support for it (think DLSS). FSR may be a (poor) solution (it isn't reconstruction it is just upscaling). UE (I am not sure about Unity) offers temporal upscaling (which is closer to DLSS than FSR), I believe they call it TAAU as it is based on their TAA implementation. Though to be fair I am not sure if TAAU is available as an option on Apple hardware.

The consoles mainly use checkerboard reconstruction (which I believe is temporal as well). The only PC game that I know of that kept the option and added DLSS is Death Stranding, so it allowed for easy comparisons.

There are two downsides to rendering at lower resolutions, 1 you get CPU bound pretty quickly and 2 you get muddy textures, and nobody wants muddy textures.
 

Krevnik

macrumors 601
Sep 8, 2003
4,101
1,312
The consoles mainly use checkerboard reconstruction (which I believe is temporal as well). The only PC game that I know of that kept the option and added DLSS is Death Stranding, so it allowed for easy comparisons.

There are two downsides to rendering at lower resolutions, 1 you get CPU bound pretty quickly and 2 you get muddy textures, and nobody wants muddy textures.

Don't necessarily disagree with your assessment here. But when people are asking why the M1 isn't keeping up framerate-wise to a PS4 or an Xbox One, despite having a more powerful GPU, then it's worth pointing out the tricks consoles use to lighten the load on the GPU vs PC/Mac where those tricks are rare, as you point out.

And if your goal is to target a consistent framerate rather than a specific graphical quality (i.e. you need the quick responses more than the fidelity in a particular game), then I'd argue the trade offs are sometimes worth making. Or in the case of the Mac, where running the GPU at native resolution on a HiDPI display for heavy games is a fool's errand most of the time, and there's some benefits to at least providing some mechanism to provide something closer to the optimum resolution to work in to hit the framerate target without the player having to do that tweaking themselves. That said, I think Leman pointed out some other possibilities for Apple GPUs here that would probably be better than the more brute force approach consoles have used this last decade. My main point was more that the Mac with high resolution screens could benefit from any this sort of trick to lighten the load and make it easier to sustain 60fps.

Although when it comes to getting CPU bound, I'd assume if you are targeting a framerate, you'd mostly see getting CPU bound above that framerate (based on target hardware, yadda, yadda, yadda). With Microsoft's dynamic resolution tech at least, they were trying to keep things in balance to stay near the target framerate. Spend roughly the same GPU time per frame, cranking up fidelity on easier frames, and down on harder ones. So I'm not entirely sure what scenarios you are thinking of when you bring this point up.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.