I have to say, I was expecting them to beat Intel handily, but fall short on the GPU front. The efficiency of those GPU cores (something that was less discussed than CPU efficiency before this event) is pretty amazing! Almost made NVIDIA look bad!
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Geekbench score for M1 Max. It's great
- Thread starter Kung gu
- Start date
- Sort by reaction score
since you are talking about 16" , then yes, you are right, since that has an dGpu also......now that im thinking, you are totally right...the Intel 16" has 96W also and no issues at all at full load
Good grief. Glad the forum engineers are here to point out all of the things that the actual engineers couldn't possibly have accounted for already.
You think that the 14" will have the same cooling capacity as the 16"? That would be... unusual, if it were true.
You know that’s not possible. But we don’t know if these systems are thermally limited at all (like old Intel laptops). Maybe both chassis can hit the max frequency without becoming thermally limited.
To be honest, there is a bit of a question mark on if the M1 Max will throttle. I’m not worried about it being power starved, but rather that the Intel 16” MBP starts to sweat with these 50W+ loads, which the M1 Max doesn’t really shrink a ton here. It shaves off the worst power behaviors of the Intel CPU, but keeps the design power in a similar ballpark, instead opting to shove as much GPU grunt in there as possible.
So you have:
- Smaller chassis.
- Slightly lower heat potential to Intel 16” MBP.
- Improved thermal management, but unknown exactly how good it is.
So, I’d think it fair that there is some doubt as to exactly how well the M1 Max sits in the 14” chassis. I hope Apple has done a good job here, but I don’t think we have enough information to remove all doubt, either. Apple did make the M1 MBA fully passive, knowing that the M1 would scale back after 15-20 min of heavy load. But even if the full-fat M1 Max can’t quite run at full tilt sustained, I wouldn’t be surprised if the 24-core version could. And for bursty loads, the 32-core could still be useful over the 24-core.
That's a bit of a chicken and an egg problem
Apple usually designs their chassis to provide exactly as much cooling capacity as the chips will need (at some reasonable sustained performance level they have defined). I think M1 Max in 14" and 16" will work a bit like M1 in Air and Pro — the larger laptop will have higher thermal headroom and thus be faster overall. But one can't really speak about "throttling" here as the system operates as intended.Agreed. Although Apple generally targets a low fan noise operation too. I think it will be slightly faster on the 16” anyways, but it’s a bit different than the M1 Air/Pro comparison since the Air obviously can’t spin up its fan to compensate. Maybe the 14” M1 Max will just be a bit louder under load, or a bit slower, or a combination of both. But I doubt it’s going to be significant.That's a bit of a chicken and an egg problem
My train of thought is that they were, unlike previous generations, designing around the 14" and then scaling to 16" once performance goals were met.
To have just that small difference, it should make a lot of 14" users very happy...i mean the difference M1 macbook air vs Macbook pro is not that huge...That's a bit of a chicken and an egg problem
I think the fans could as big as the 16"..and the free space to be filled with 30 more watt-hour lithium-polymer battery and bigger speakers ..who knows, i want to see the inside of both 14" and 16" with the m1 max..please man, show me the inside, i know you have them already...
No, i9 is slower. There is no i9 with 12500 multicore score. Maybe overclocked and used as space heater.
People are kinda skipping over the fact that the CPU cores have access to this 400GB/s memory bandwidth as well as the GPU cores. This is unprecedented. Only dedicated GPUs have access to this kind of memory bandwidth. CPU/RAM bandwidth on traditional high end PCs caps out at under 100GB/s.
You can now use the CPU cores for data intensive processing like you would the GPU cores.
You can now use the CPU cores for data intensive processing like you would the GPU cores.
The CPU cores will still be limited in how much data they can fetch (since they don't have the width of the GPU), but already M1 is beating x86 CPUs in many workloads that rely on memory transfer (e.g. stuff that involves searching in any way). I am exited to learn how much bandwidth will be available to individual CPU cores. It should make my work with data analysis much speedier indeed.
But what's even more important is that these chips allow you to do truly hybrid algorithms which until now have not been practical due to high CPU-to-GPU latency.
Agreed. Even having the option of getting a dGPU in the 13” MBP has been missing for what, a decade or so? So having some actual GPU oomph in the smaller MBP is a welcome change, and making the choice a bit easier for folks, rather than forcing people who need more GPU performance into the larger laptop.
One can hope. But I’m not really sure it’s a huge problem if the 14” cannot go all out for hours and hours on the full M1 Max. If it still helps noticeably for bursty loads like Lightroom, Photoshop/Affinity, and the like, it’s good the option is there. But one impression I get from this update is that Apple is not wanting to leave performance on the table, and is more interested in riding the limits of the thermal envelope they have. Which considering the lack of dGPU on the Intel 13” MBP, surprises me.
There is an i9 with a higher single core score and slighty lower mulit-core (11XXX score)
I have a intel i9 16 and it won't run my video editing software, OBS and chrome (evercast) without a AC at 18ºC. If I raise the AC to more comfortable levels of 22-24ºC, the system will drop my clock to 1.8-2 GHz, basically making my editing software to crawl. All of this magnificent performance on top of a full blast fan noise floor.
It's one of the worst machines I have ever own... much prefer my 2011 cheesegrater.
I edit in HD proxy, by the way.
Well...this is not about the Mac Pro...but still....
No, but then again 13" M1 Air without cooling is barely under-performing compared to the M1 13" Pro with a FAN, so i don't think there's gonna be a problem. And the fan on the 13" barely turns on anyway.
Fastest notebook CPU from Intel is i9 11900HK, single core 1652, multicore 8480 (newer mind that notebook must be on power cable, nor that it runs hot and noisy and battery life sucks). You can even find better scores as long as room temperatures is low enough or it is first run of benchmark.
Anandtech has nice article about it here: https://www.anandtech.com/show/16680/tiger-lake-h-performance-review
I think you are comparing to 11900K, that one really is faster, but that is desktop CPU with 125W at base frequencies that is running 5.3GHz in turbo mode and consumes 200W of power while doing that.
Edit: There is 11980HK, with 1673, 10648, that is a bit better, but at expense of power consumption.
Well, all Intel cpus starting from Skylake were just minor iterations. New hope will came from Qualcomm, that might get Apple some competent competitor in 2022.
Last edited:
The CPU reads data as fast as the L1, L2 and the SLC allows, and that’s usually at CPU core clock. I’d think Apple’s engineer would have designed their memory controller to fill up the SLC as fast as the memory modules allows. So CPU saturating 400 GB/s memory bandwidth should be possible. Hopefully Anandtech will do an in-depth analysis of the M1 Max.
I think Andrei mentioned that there were some limits on data transfer (from L2 to L1), which are probably still there, but these chips should have a faster fabric internally. I wonder what the SLC bandwidth is on these chips. If I read the graph correctly, M1 was limited to 100GB/s bandwidth from cache per core (which is still very very much).
The 2020 Mac Mini Unleashed: Putting Apple Silicon M1 To The Test
Depends on how large each cache line is for L1, L2 and SLC. If each cache line is 512 bits, at 3.2 GHz, my calculation is just under 200GB/s.I think Andrei mentioned that there were some limits on data transfer (from L2 to L1), which are probably still there, but these chips should have a faster fabric internally. I wonder what the SLC bandwidth is on these chips. If I read the graph correctly, M1 was limited to 100GB/s bandwidth from cache per core (which is still very very much).
The 2020 Mac Mini Unleashed: Putting Apple Silicon M1 To The Test
www.anandtech.com
Do you own either device? When cold the MBA M1 performs lower than cold MBP M1 and loses ~34% performance due to throttling under sustained load.
Last edited:
except when the power goes out and then all your work is done.